Real or AI? How to Spot AI-Generated Photos, Videos, Voice Clones & Deepfakes in 2026

1. Intro: Wait. What's that AI generated and this? What about her? We live in an era where S is no longer the same as believing. And that's super weird and very overwhelming. Right. For most of human history, our eyes were the most trustworthy source of truth. If you saw a photo of someone doing something, you assumed it happened. Of course it happened. But now a single AI prom can create a world that never existed. Photo that looks like it was shot by a national geographic photographer, except it was not. It was eye generated. A video with shocking news that everybody believes in a fashion campaign with a shirt you plan to buy. The good news. Even though those tools are getting better and better, our eyes are still smarter than we think. You just have to know what to look for. And that's why I created this course. I want to share all my knowledge, how to spot AI generated photos, videos, news and I voice overs and how to stay safe and aware. In today's a little bit strange reality. So, yes, I will spill the tea. And I divided the course into four sections, four practical sections, and I hope each one has many things that will surprise you. There is a lot to discuss, so let's go. 2. Class Project: Class project. After watching the cars, I want you to use what you've learned and train your eyes and spot those little inconsistencies that reveal an AI generated photo or video. I want you to see and I want you to notice the thing that most people miss. And you will be amazed how your awareness sharpens once you start looking for the clues. Okay, but what's your class project? What I need you to do? A viral or not viral AI video or a photo. If you prefer video, choose a short clip that's clearly AI generated. You can find plenty of those on TikTok, Instagram, YouTube shots or X. If you can't find any, just search AI video Deep fake or AI influencer. Be, for example, a fake celebrity interview I fashion show or an AI generated cinematic scene, anything that looks almost almost real and watch it slowly and more than once. Play it at a normal speed first. Then rewatch it in slow motion. Notice the little shifts that fell off and look for those visual inconsistencies and the clues we will be discussing in the course. And when you spot them, freeze frame and take screenshot or screenshots. Pause at the exact moment, something strange happens. The exact moment you notice something strange and take the screenshot. And then circle mark the details you noticed or write small notes just like I did to show you. Then, of course, share your findings, upload your screenshot in the class project gallery right here and include a short description of which video you analyze. You can also share the link if you want. And what inconsistencies you found with those arrows and notes on the screenshot itself. So I'm really curious, please share it with me. I'm really, really excited to see what you notice because truly my goal, the goal of this course is for you to leave leave the course, finish the course with stronger visual literacy and deeper understanding of how to question what you see online right now. 3. What AI-Generated Content Is (+Deepfakes, Voice Clones, AI Art): AI generated content is defects, voice clones, and AI arch. Let's go back for a second to 2020 or 2021. Simpler times, right? Back when the most artificial intelligence, most of us interacted with was quietly running in the background. For example, recommending us videos, rerouting us around the traffic of flagging suspicious payment. It was when I was working mostly behind the scenes with stuff we didn't really see. But do you remember that video with the deep fake of Morgan Freeman? I am not Morgan Freeman. And what you see is not real. Well, at least in contemporary terms, it is not. What if I were to tell you that I am not even a human being? Would you believe me? What is your perception of reality? Is it the ability to capture, process, and make sense of the information our senses receive? If you can see here, taste or smell something, does that make it real? Or is it simply the ability to feel? I would like to welcome you to the era of synthetic reality. Now, what do you see? It's far from being perfect because, yeah, it's from 2020. And at that time, I used to talk with my colleagues about the future of technology and AI. And my secret weapon was fake videos of Tom Chris. What's up TikTok? You guys cool if I play some sports? I love it. More for the audio experience. As much as the momentum. Hey, listen up, sports and TikTok fans. If you like what you're seeing, just wait till what's coming next. Highly realistic face swap videos posted to Tik Tok. Creator Chris met plus and impersonator, showing Tom Cruise doing everyday things widely shared and flagged the moment the effects become convincing enough to four casual viewers who was scrolling through TikTok and found it. Well, funny. So when I was showing it and discussing it with my friends at work, it always got people's attention. It was the kind of thing that made people go, Wait, What? What is happening? And suddenly everyone cared about machine learning for like 5 minutes because it scared them and fascinated them at the same time. I'm going to show you some magic. The real thing. But back in 2021 and 2020, generative AI wasn't a big thing yet. This was before the rise of generative AI before we could just ask a model to take a photo, generate a photo, a song, or entire video out of FNN AR. I remember discussing about what would happen when this technology got into everyone's hands. And most people I know imagine a nightmare future, fake presidents, fake news, fake everything, and so many things that try to deceive us. And then well, yeah, it happens. And now we are swimming on AI images, AI videos, I songs, stories, fake news. All of it everywhere. AI is now part of our phones, our feeds, and our culture. Our favorite ops now have very advanced AI powered features. And you probably notice your social fits getting flooded lately with very realistic AI videos the kind that make you stop scrawling and think it's a little bit strange, but wait, wait. Is this real? The clumsy Ai giveaways we used to laugh at. The strange fingers, frozen smiles and awkward scenes changes. They are almost almost gone now. The latest AI generated videos move more smoothly. The faces look alive, and even the voices carry emotion that feels convincing to so many of us. And they are realistic enough that many of these short clips are already spilling into Tik Tok, Instagram and YouTube, blurring the line between what's filmed and what's made up. So, in short, yes, the next wave of AI video and photos is here, and this time, it actually looks real and is super high quality. Generative AI went mainstream almost almost overnight. And suddenly all of our previous W is conversations turned into, Okay, what now? So now we live in a time when reality is very remixable. AI can generate people, voices, and pictures that never existed. And unfortunately, they are getting harder to tell apart from the real thing. So, to begin with, let's unpack the three big ones deep fakes, Vice clones, and AI art. Deep fakes videos or images where someone's face or body has been digitally replaced using artificial intelligence. The deep part comes from deep learning, a type of AI that trains on thousands of images of a person until it can realistically map the face onto someone else's body. And the result, a person can appear to say or do something they never did. You can make them say or do anything you want. Yes, for example, my favorite example from the past Tom Cruise on TikTok, hyper realistic videos of Tom Cruz doing magic tricks and golfing made by YVX artist and an impersonator, totally fake but also totly viral. Also, another example that was very viral in the past, though it was ages ago, a fake Obama video to warn people about defts and misinformation. You're entering an era in which our enemies can look like anyone is saying anything at any point in time. Even if they would never say those things. So, for instance, they could have me say things like, I don't know. Kill Monger was right, or Ben Carson is in the Sunken place. Or how about this? Simply, President Trump is a total and complete diphat. Now, you see, I would never say these things, at least not in a public address, but someone else would someone like Jordan Peele. This is a dangerous time. Moving forward, we need to be more vigilant with what we trust from the Internet. That's a time when we need to rely on trusted news sources. It may sound basic, but how we move forward, and the age of information is gonna be the difference between whether we survive or whether we become some kind of ****** up dystopia. Thank you. Stay woke, *******. Also another example from ages ago from 2020, South Korea's I news anchor when broadcaster Ji ha had a defake twin deliver real news on TV. Viewers were told, but it raised ethical questions anyway. So why do we need to learn to identify them and pay attention more than ever before? Because defaks can be funny or creative, but they can also be super dangerous, although we often don't realize it at the first thought. They are used for misinformation scams and non consensual porn. I think they really challenge how we define truth in today's digital spaces. Okay. So now, what are voice clones? So voice cloning uses AI to mimic a person's speech, the rhythm, accent, and tone, everything that makes the way we speak hours using only a few seconds of their real voice. Yes, this is very scary. And think of it like a deep fake but for your voice. Once your voice is cloned, someone could make it sound like you are saying anything. Anything. Really, anything. Real examples of it? Okay, let's go to the past because it all started a long time ago in 2021. For example, that year that year, a big corporate scam happen when fraudsters use a cloned voice of a company CEO to trick an employee into wearing 243,000 Dollis. Also, there are many fake celebrity endorsements. AI voice clones of big celebrities like, for example, Emma Watson or Tom Hanks have been used in fake fake arts. Viral AI generated song using clone voices of Drake and the weekend without their permission. Okay, and now the third big player AI rt. So what is? AI art is when algorithms like Dali, Magni or stable Diffusion generate images and videos from text prompt. You type, for example, a cat wearing gucci sunglasses on Mars and wala you have your cat. A cat appears in seconds. And to debunk the maps, as you probably already know, it's not copying a single artist because it lends for millions of images online to create something, something new. And artists use it as a creative tool, but others see it as plagiarism, since it learns from human work without consent. And what are the loud examples? For example? Yeah, yeah, I was also ages ago, but it was very loud, so we need to go back in time to that. Theater Opera special, a Colorado art fair winner that turned out to be AI generated sparking the debate about what counts as real art. Is it art? Should it be band? Should it win? The debates were endless? The next example, the Willie wonk experience Mm and event promoted with AI generated visuals that misled Attendee. So as you can see, the free big players, it's all about something a little bit different, but they also have similarities. So here you can see what they are in a shortcut in a nutshell. 4. Real-World Consequences: Misinformation, Scams, Reputational Damage: Word consequences, misinformation scams, reputational damage. The consequences of AI-generated videos and photos are such a broad topic. So I think we need to talk about it a little bit longer because the topic really deserves it. Many times, we don't realize how huge the topic is and how serious it gets because we are not seeing the real consequences in real life, for example, in our lives, but it doesn't mean they are not there. So as you already know, AI has a shadow sight, and a the flakes are surely it. They are the kind of technology that makes your stomach drop a little because videos, photos, and even voices that look and sound perfectly real except they're not. Hey, here we go. A there, Swifties and cyber Warriors. Taylor here, I'm so excited to announce that the weight is over. The special No B four Edition album just dropped. Twist. I'm not actually Taylor. I'm Perry. And this is a deep fake. Yeah. This was a lot of fun, but deepfakes are not just fun. Deepfakes can be extremely dangerous. As you already know, they can make anyone appear to say or do something that never actually happened. And while the concept isn't new, because as you already know, it was also doable in 2020. The high quality of it is. What used to look like a bad lip-sync or a little bit glitchy video filter now looks really believable. Too believable. And that's the part that scares most people, and honestly, it scares me a little bit too. Yeah, that's why I'm creating this material for you right now because the same creativity that makes fun and inspiring and so good to play around with also gives us tools that can bend reality in ways we've never had to deal with before because a deep fake isn't just a prank or a digital trick. It can be weaponized. It can spread lies faster than we can **** check them. Ruin someone's reputation completely overnight or quietly shift public opinion before anyone realizes it's happening because then it's too difficult to stop it. And one of the biggest worries, of course, are elections. Imagine waking up to a video of a political candidate saying something totally awful totally disgusting and everyone believes it because it looks and sounds exactly like them. Even if it's proven fake later, the damage the damage is already done. Trust is broken and the lines blur and democracy which run on trust and truth takes another big hit. For example, my grandma or my mom would totally believe any fake they saw. And they would start telling their friends what a terrible liar that candidate was. And that's how this little circle of personal misinformation begins to spread quietly but very powerfully reaching food than we can expect. And that's what makes feces feel so unsettling, not just that they exist, but because they remind us how fragile our shared sense of truth really is right now. And because of that, it's a strange time to be living in, I admit, a world where, as I've already told you, s is believing doesn't quite hold anymore. Also, other disturbing versions of the effet are very intimate. No consensual explicit de fakes often called fake porn, make up the overwhelming majority of the fake videos online. Nearly all of them target women. Many of them public figures, celebrities, influencers, or entertainers. It's hard to even talk about this because at its core, this isn't a tech story anymore. Of course, it's a case about abuse, humiliation, and yes, power as well. Also my reminder for you that behind every AI break through headlines are real people often being violated in ways that they never consented to. A few months back, I joked in a post about AI and Hollywood strikes that most video generation tools reminded me of early motion capture movies like The Polar Express, fascinating but deep in the uncanny valley. Fun to look at, but nowhere near real. And yes, I have to take that back now because Open AI's newest version of Sora in a wave of generation models have completely shifted the landscape. These videos aren't cute demos anymore. They are cinematic. You can type a sentence, and the result looks like it was shot by a full film crew on a real set. You can also create a very realistic talking head video just from a single photo and an audio clip complete with natural facial movement and perfect lip-sync. And of course, details include responsible AI disclaimer say it's for virtual characters not impersonation, but you can probably imagine how fast the line could get crossed. And that's the thing we have to worry about. The same tools that make beautiful art can also make terrifyingly real lies. And of course, the world hasn't fallen apart. Yes, because in the past, a good deep fake was really hard to make. And now it's getting easier and easier each day. The ones that look cinematic often took huge amounts of computing power and some advanced technical skills. Now, the tools are very beginner friendly, and they make it so easy for almost everyone to access the technology. Also, the process is getting cheaper than it was before much cheaper. Of course, fortunately for us, some social platforms, for example, YouTube kind of care. Even if they are not deleting AI videos, they're quietly down ranking AI-generated stuff. And some AI videos got taken down fast, and other companies seem to be preparing for whatever laws are coming their way. But we can't rely on social media platforms or laws to regulate that. Need to learn to be super observant and more analytical to survive in the times that are coming because when you start to analyze and observe things more carefully, it turns out many deflects aren't that perfect and aren't that convincing yet. But you have to analyze that on a deeper level, and that's exactly what we are going to discuss today as well. For example, I once saw a fake video of a European leader admitting defeat. Creepy idea, yes, but believable? For me, no. But for so many people, even the people I know. Also, the slightly older people who aren't used to that kind of fakes. Yes, very believable. So while the video was very quickly debunked, it proved how the fakes can be weaponized in war to undermine morale and spread the confusion. Or if you want another example, look at the May 2024 Pentagon Explosion image hook. Fake AI-generated image of an explosion near the Pentagon went viral on X, shared even by verified accounts. The stock market briefly dipped before the image was debunked. And that's really scary how even a single AI image can shake public confidence and move markets in, like, minutes. So most of us worry about defects in the context of politics or celebrity scandals and rightly so. But, oh, boy, the risks don't stop there. One of the places that keeps me up at night is also healthcare, a field built on trust, accuracy, and truth. Imagine what happens when those foundations starts to shake. Do not stress you so much. Here are just a few unselling stereos that are unfortunately entirely possible in today's world. Public health disinformation, for example, a fake video or a well known doctor recommending a strange cure. Fabricated government announcement about, for example, a pandemic. And as you already know, misinformation spreads so fast and facts could turn it viral. And what I also worry about manipulated media recalls, altered scans or fake reports could be used for insurance fraud or worse, leading to misdiagnosis or harmful treatments. And destroyed reputation are also a real danger because imagine a fake clip or respected surgeon, or respected healthcare specialist saying something racist or a fabricated video of a patient revealing confidential health information. Even a few seconds could ruin a career and destroy trust and move the market. And also, the new technology is giving so many opportunities for fishing because summers already use deep fix audio to impersonate CEOs, so it's not very hard to imagine them pretending to be a doctor, your therapist, or your pharmacy. And, of course, this is also a very real concern. Fake science because the fakes could be used to manipulate research footage or interviews, crafting the illusion of for example, scientific breakthroughs that never existed. You know, that breaks thrust not just in a scientist, but in science itself. And unfortunately, none of those are hypothetical anymore. We are already seeing early versions of it, and I fear, yes, it's going to get more intense, unfortunately. So could it be that we are already living in a world where the best fakes are undetectable, unfortunately, in my humble opinion, yes. Because the next generation of AI video models is already there. You've probably seen things like open AI Sora, the newest version, especially or Google's video, tools that turn plain text into realistic video scenes. They are not exactly deepfakes, but they cousins in the same strange new family of syntatic media. And they are behind the sudden leap in quality. For example, the newest version of Google Video is a video generation system from Deep Mind. The newest version of video is designed to create fully realized video scenes from just a text to prompt. And what is very important, it doesn't just generate the visuals. It can also produce dialogue, ambient, sound effects, and music that sync naturally with what's on screen. And yes, it understands how light motion and gravity interact with output with free K preview available through vertex AI. So now anyone with Google AI account can experiment with it, though the number of clips you can generate depends on your subscription level of cars. And even on the entry plan, you can experiment with the possibilities that Tools gives you. So when AI video gets this realistic, We are stepping into a strange new chapter. One way video, the medium we've trusted for decades starts to feel completely unreliable. You can't believe your eyes anymore. So how do you protect yourself? To answer that question, Google released something alongside the newest version of video called Sync ID, a kind of digital fingerprinting system for AIMdia. It can scan a video and tell you it was created using Google's only Google's AITols. So here's how it works. Every video produced by the newest video carries an invisible watermark, a hidden signature embedded directly in the files data. So you can see or hear it, but Google's Sync ID can detect it instantly confirming whether a clip is genuine or AI-generated. Is this real or AI-generated? Sometimes it can be difficult to tell. For hundreds of years, humans have used watermarks to prove where content was created. One challenge is developing a mark that's hidden to viewers but visible to those looking for it. Enter Synth ID by Google Deep Mind. It can produce a digital watermark that's imperceptible to humans and works across Google products to tag AI-generated images, video, audio, and text. The watermark can withstand common editing techniques like reordering and trimming, adding noise, compression, cropping, and filters. How? For images, it can be embedded directly into the pixels. For video, it also marks frame. For audio, SynthiD converts the signal into a spectrogram, a visual representation of the sound waves into which it embeds the watermark before it converts everything back into a waveform. You can't read hear or see the watermarks, so you can fully enjoy your creations. SynthiD technology is already used across generative AI, Google consumer products. Synth ID is just one tool we're using to make sure generative AI tools are built with safety in mind from the beginning. And I know, it sounds reassuring, right? Well, it shouldn't shouldn't entirely because here is where it gets really complicated. Synth ID only works for videos made with Gols owned products. If another company say MIT, Idobe or any independent developer builds a model without that watermark. Sync ID can detect it. And just like we've seen with AI-generated images, people will eventually find ways to remove or mask those watermarks altogether. And it's not hard to imagine where this is going. Social platforms like meta or TikTok will probably roll out their more advanced detection systems, but history tells us how this plays out. Every new layer of detection will be quickly met with a walk around. I think it will become an endless back and forth game like a digital cat and mouse game. But that leads to the real question. What happens when reality itself becomes negotiable? When a political speech, a celebrity acting or a breaking news clip can be generated in seconds, and the tools meant to expose fakes can no longer keep up. That's not science fiction anymore. That's the information battlefield we've already walked into. And the truly dangerous part, sometimes those videos don't have to fool everyone. They just have to reach the right person at the right time. So maybe the most interesting question isn't about the technology at all. It's about people, how we choose to use it, share it and believe in it, and how to survive in these times and take care of our well being as a society. And I think to do that, to survive in today's world, you need to understand how it all works and why it works like this. 5. Understanding AI Image Generation: Standing AI image generation. There is a special kind of magic in typing a few words or a few sentences, for example, a fluffy cat wearing sunglasses on a surf board at sunset and watching it appear on your screen like a thought made real, like a thought made visible. And the first time I used one of these AI artles I actually laughed out loud, not because the output was perfect, but because it felt like I had just whispered an idea and suddenly it appeared on my screen. Super strange for me, suddenly, I had this power. Suddenly I had access to some sort of white magic or came into the Harry potter world. But it's not magic. Tools like Majury Dolly, all stable diffusion, and all the rest have turned imagination into something instantly appearing on our screens and for better or worse, they've made creativity feel like collaboration between human and a guy. But beneath the poetry and the magic, beneath it, beneath it. It is a fascinating and also unsettling. Reality. The systems, the tools don't understand what they are generating. They don't know what a card is or what a sunset feels like. Be artificial intelligence and the models, the LAM modules, they are not inspired, they trained. So let's pull the curtain back a bit. Art of prediction at the simplest simplest level, these tools work by predicting what pixels should look like, the pixels you want. When you time a prompt, for example, say, 1960s Paris cafe seen and watercolor, the model doesn't paint it like an artist would. Instead, it starts with a noise, a random static field image, and then gradually removes case one small step at a time until it becomes something recognizable. And I know you already heard this word a lot. This process. Listen, this process is called diffusion. And you can imagine it. You can think of it like unblurring an image that never existed. And in training, the module has already seen millions, sometimes billions of pictures along with the text that describes them describe them. For example, it learns the patterns that cafe often includes tables, cups, worm lighting, and it also learned that watercolor means softer edges and faded hues. It also learned that 1960 Sporis has a certain kind of fashion and architecture. Over time, it learns the visual language of our human world. So not the meaning behind it, but the shape of it. Then when you give it a prompt, it translates your word into that learned visual grammar. It doesn't create from memory. No, it reconstructs based on probability. Embeddings and tokens. Here is where it gets a little bit nerdy. When you type words into those models, they are converted into what's called embeddings, like little mathematical vectors represents meaning. In the models world, dog, puppy, and canine live close together in this abstract multidimensional space of human wealth. That's how the model understands that those words are related, even though it's never seen them next to each other before. And the visual sight works similarly. Every color, every line, every shadow and texture also exist in this enormous map of relationships between words. So when you ask the model for a stormy see scape, it searches its memory in a statistical sense for where stormy and see scape overlap and then starts to construct an image that fits. So no, to debunk the myths, it's not pulling old photos from a database. It's imagining the most likely combination of visual details that match your prompt. It's a bit like how your brain completes a sentence before someone finishes it. Except that AI is doing that with pixels instead of words. The free big players. Each of the big names you probably heard of and probably used Mejoni Dolly and Stable Diffusion has its own vibe, kind of like three artist painting the same scene in different styles. In my humble opinion, Madgeny often feels like a driver because it outputs its images cinematic, moody, and often more surreal than images from the other ones. It runs on this chord, which somehow makes the whole process feel communal. Med journey uses its own diffusion model, but it's fine tuned for aesthetics rather than photo realism. It's less what would this look like in real life and more what would this look like in a beautiful dream? Dali through opening eye is the pragmatist of the group. It's the one that tries to get things right. The module has been trained with more moderation and more filtering, and it has built in safeguards against generating real people or explicit content. It can also easily edit existing images, adding, removing or re imagining parts of a photo, while keeping everything consistent with the lighting and perspective. Then there is stable diffusion durable, and unlike the others, it's open source, meaning anyone can download it, modify it, and use it however they like, and they want. That's both its strength and its danger because it's been used for stunning oral art and also for less savory things like fake porn or propaganda. So what they are actually learning. So here is the part that feels almost existential. These models don't see images the way we do. They don't know beauty or context or emotion. As I've already reminded you, they only le patterns. They learn that things called mountains often have jacked shapes. That sky is usually blue. That sadness correlates with grey scale stones or downward gazes. They don't feel af when they render a sunset. They just statistically predict the pixel colors that correspond to sunset. And yet the results can make you feel something. And isn't that strange? The art itself is fake. It's syntaxic. But our reaction to it, what it evokes in us, it can be real. The ethical site. Of course, the conversation doesn't end with how these modules work. It quickly turns to how they're used. They are trained on huge collections of offline images often scraped from the Internet without the original artist's consent. And when EI generates something in the style of a living artist, it's not inspiration, it's imitation. The model learn that style by starting artist, that artist's work line by line, color by color. So that raises huge questions about ownership, credit, and creativity because who is the artist in that case, the person who wrote the prompts, the engineers who built the model or the countless often unnamed artists whose work trained it. And as you can guess, there is no clear answer yet. There are so many perspectives we have to take into consideration, and the question itself matters so much. These questions matter so much, and I think it's so important for us as a society to discuss them, to think about them, to think about our personal approach, how we see it. And what I personally fear about when it comes to this world of really good generative art is how fast it's changing what really looks like because every image, every photo we see online now carries a tiny invisible. Question mark. Was that painted, photographed or generated? And does it matter? I think it does matter. And it's so easy to romanticize AI or monise AI. But maybe the truth lie something in between. Mid journey, dolly, table diffusion. They aren't magic. There aren't monsters either. They are just mirrors that learn to mimic human work. 6. Common Use Cases: AI-Generated Art, Ads, Memes, Propaganda: Common use cases, AI generated art, ads, memes and propaganda. I think every new technology finds its voice somewhere between play and power. And of course, AI image generation and video generation is no different. When tools like Megury dolly and stable diffusion became accessible to everyone, the first thing people did wasn't create political ads or the fake propaganda. They made memes. Cas in space series, biblical scenes starring Shrek. Politicians turned into Disney princesses. And that's how most new things start, of course, innocently, with curiosity and human. But as the novelty of the same technology that made us laugh and that made us laugh, also start shaping what we see, what we buy, and what we believe in. So now let's analyze and let's discuss what people are actually using these tours for the good and the bad. And I prepared some interesting cases for you to see. I generated art. Yeah, AI art has created an identity crisis for the creative world already. On one hand, it's the most exciting thing to happen to virtual culture in decades because a 14-years-old can not make images that look like they belong in a gallery. Any poet can visualize their words and experiment with the interpretation, the visual interpretation. A fashion designer can prototype entire collections in a day. It's kind of democratize creation in a way that feels both liberating and so deeply overwhelming and concerning, right? Of course, the beauty of it can be the accessibility. Art that used to require years. Years of hard training or expensive tools now starts with a sentence. Whatever sentence you wish, but it's very complicated. Because when everything is art, when everybody can generate art, does art lose some of its meaning. And when a model trained on other people's work produces a beautiful image. I've already told you about the question. I think it's the most important question. Whose art is it really? And this is the question I can't stop thinking about because I've seen stunning pieces of AI generated art that genuinely moved me. I've also seen others that felt very hollow, like the visual equivalent of an empty calorie. I've consumed it, but I felt nothing. So in my eyes, when it comes to art, AI can produce beauty, but it can produce intention. As. Yeah generated art becomes a business tool. So the Os wild was one of the first fully embrace generative visuals, because of course it was. Advertising is, after all, storytelling with a sales pitch, Instagram ads, Facebook ads, YouTube ads. Now, instead of hiring photographers, stylists, and entire production crews, rents can pm their way to a perfect campaign. A young couple on a rooftop in Barcelona, Golden Hour, holding iced coffee. 30 seconds later, there it is. It's almost too pretty to drink. Yeah, but it's too good not to. AI generated advertising images and videos are cheaper, faster, and endlessly customizable. Need your model to be older, more diverse, wearing a different shirt? Well, that's a prompt way. But there is a cost, of course, not only in jobs, but in authenticity and customer trust because these are images that don't come from lived experience. They are composites of our collective idea of happiness or desire. So we are entering an era where the smiling faces in billboards may not even exist. But they look real, right? Say some dreams are too ambitious, but we've never believed in limits. Challenges didn't hold us back. They pushed us forward. Every obstacle, every doubt shaped us into something stronger because progress isn't about the past. It's about what's possible. And what's possible is everything. Memes. In my eyes, personally, art is about expression. Arts are about persuasion, and memes are a way to connect with other people, and to connect with humor and something that is relatable. And Yi has turned me making into something wild, honestly. You can now generate hyper specific humor that feels tailored to your exact niche, for example, mid 30s millennials who love oat milk and has existential crisis and had the job. And sometimes somehow the AI gets it. And I to have made me creation more surreal, more layered. And yes, more absurd than ever before. Because whole me formats now emerge overnight because someone figure out how to generate endless variations of the same joke. It's fun, but it's also how cultural narratives for now because, of course, it was always like that. Memes spread faster than facts. They shape perception in a way. They become shorthand for opinions and identities. And when AI is creating them, that process speed up tenfold. So we used to say the Internet never forgets. Now it feels more like the Internet, never stops remixing the reality. Propaganda, the darker use case. And, yes, yes, this is where my mood shifts. The same technology that can make beautiful art and laughable memes can also make manipulation and oh boy, such serious manipulation. And AI generated images and videos are being used to spread false stories and stir division right now. For example, a photo of a World Leader cut. Cut to doing something scandalous. A protest scene that never happens. A fake disaster photo designed to drive panic or sympathy. These images travel so fast and S unfortunately, much faster than fact checking or truth. And propaganda has always existed, always existed, but this new version of it looks real enough to bypass our skepticism. Sometimes it's very coordinated. State backed campaigns trying to influence elections or international opinion. Other times it's more like decentralized calls, a random post from an anonymous account that goes viral before anything anyone knows is fake. Either way, the result is the same confusion and confusion is so powerful because when everything could be fake, even real things start to feel suspicious. That's the scariest part, I think. Not that we will believe lies, but eventually some of us will stop believing anything. So yes, these tools are creating art and us and memes and propaganda. But what ties them all together is something distinctly human. The desire to communicate something. So people use these tools, good people and bad people use these tools to express, persuge, connect and influence, just like we've always done. The medium is new, but the process isn't. The same technology that makes a political deep fake that is so dangerous can also make a birthday card. The same algorithm that bulls in ad campaign can illustrate a love poem. 7. Signs of Fake Images: The Visual Clues: Sign of fake images and photos, the visual artifacts. These are AI generated images using the same prompt one year apart. Very impressive right. The progress is almost unbelievable. But even now, AI generated photos and images often look perfect at first glance until you look closer. Then you start to see the strange details. And that's the good and bad news. The technology goes on, and it is definitely going to be near impersistb as this tech moves forward. But right now there are still some little artifacts, some little mistakes that AI makes. Start with text. Sometimes AI still struggles with language in images and photos. A street sign that says stop might come out like this. A T shirt logo might morph into something unrecognizable and very strange. That's because the model doesn't actually reach. It just knows what text looks like. One simple rule first. Always zoom in. AI loves to mess up the tiny details. And here's what to look for when you zoom. Text in the background. Street signs, posters, book covers, license plates because AI text often looks almost right. Letters melt into each other. Words don't make sense or fonts randomly change. If something looks fine from far away, but starts falling apart, when you zoom in, yeah, that's usually AI. So always look at signs, posters, billboards, license plates, shops, shop names because AI hates these. And I will also show you how it works in practice because I have so many interesting examples to exemplify this better. And what usually goes wrong? Gibberish words, letters look real, but the words make zero sense, like it almost spells something, but doesn't. Melting letters, text bends, stretches or blends into the background. Straight lines suddenly become wavy or soft, inconsistent fonts. One word, three different font styles, or letters randomly changing size or shape. Impossible clarity or none at all. Either the text is weirdly blurry compared to everything else or unrealistically sharp, anyway, that feels fake. Text that avoid being readable. AI often keeps distant signs just clear enough so you don't question them until the Zoom, right? And real photos usually have boring, readable text, even if it's slightly blurry. And AI tries to look convincing but collapses under Zoom. Then there is asymmetry. Humans are asymmetrical. Gloriously so. Our faces, hands, and even our smiles all carry tiny inconsistencies. But AI, of course, tends to create balance. It over corrects. So look closely at portraits that seem to perfect where both sides of the face mirror each other a little bit unnaturally. This perfection problem runs deeper than aesthetics. AI generated influencers like the virtual model Iana pus or Lil Miku often appears too symmetrical, too smooth, too consistent. They are very attractive, yes, but in a way that feels engineered. Humans are drowned in perfection because, well, it's a sign of realness. On the flip side, that's funny, you will sometimes see the opposite strange shapes or misaligned jewelry, earrings that don't match, or backgrounds that curve where they shouldn't is amazing a texture, but it still is clumsy as geometry. I think it's the artist equivalent of a kid drawing a beautiful house with the windows floating a little too high. The eye generated fashion portraits often have melting earrings or asymmetrical jewelry. We'll also spot hands with hiding finger and we don't know where it is or wine glasses bending in straight directions. At the ice. A more subtle giveaway in AI generated photos is often hiding in the ice. If you look closely, the reflections can seem too sharp, too white, almost like it was painted on. The pupils sometimes have this perfect little white circle of light, something that really happens in real life outside the photography studio. For example, when I compared an AI generated image of myself to a real photo of myself, look, the difference became obvious. In the AI version, my hair looked unnaturally sleek. Like every strand had been ron flat and the light in my eyes looked strangely uniform. Look, this perfect opec circle that didn't feel live. The real photo, you can see how daylight is visible on my skin, on my face, but also in pupils, softer, gentle highlights rather than pure white spots. To be fair, the AI image was trying to imitate a studio photo with artificial lighting and studio setup can sometimes create those shiny bright reflections in pupils. But AI tends to overdo it. The highlights become too round, too bright, and too perfect, giving the whole face that slightly plastic to polish look. When you look at the details, for example, in pupils, light shadows and anatomy. A tiny but important giveaways. Right now, lighting can be one of the easiest ways to catch a fake and AI generated photo because real light behaves consistently. It always has direction, tone, and depth. And AI light, it gets confusing. You might see a person standing under the sun with shadows that fall in two directions or reflections that don't match what's nearby. So it's very important to pay attention to the reflections, also the reflections in mirrors, glasses, water. They often reveal a completely different reality. Sometimes there is a disorder hand in the mirror an object that doesn't exist on the first plane or even a second phase that AI forgot to erase. And when it comes to anatomy, Yeah, I know. You might think about the famous AI hands, fingers that multiply like heat rows, arms that bend and at impossible angles, and teeth that blur into white cals. Ears that fade into earrings. Right now, those things are almost gone because these modules are learning fast. So now AI is doing much better with hands and fingers and toes, but sometimes from time to time, it still makes these mistakes. So look at those two portraits. One is AI generated and one is real. And what do you think? What tiny imperfections give the real one away? What do you feel when you look at the AI version? And nothing screens a eye at first glance. It gives itself away in the micro details. So look where the lighting doesn't fully come in because eye lighting often looks cinematic but directionless. Feels evenly nice across the face, lacks harsh shadows where they should exist. So what do you think? And when I ask JTPT to judge which photo was eye generated and which one was real, it got it wrong. And JTPT said the photo I personally took my friend was definitely AI generated, which, of course, it wasn't, and those were the arguments it gave. Because, look, the photo on the right is real, 100% real. It's a real person and an almost raw image from my photoshoot edited only lightly in live room. So if you are judging a eye based on how pretty or how polished a person looks, you're already in trouble nowadays. And I think this is where the real lesson. There is no single visual giveaway anymore. Anyone claiming a eye always looks like this and this is oversimplifying. High quality photography can look more artificial than AI nowadays blur. Is the background super blurry? In many AI generated images, the background is very often either too blurry or too sharp or hyperreal. Real photos can, of course, have a shallow depth of field. But no, AI blur looks different. It's exaggerated, spread evenly across everything behind the subject. I actually got this one wrong during the experiment lately. I followed the media lapse checklist, taking the person facial hair, skin texture, light, and their eyes. And when I was looking at a photo, I think everything looks authentic to me. The skin tone was natural. The reflections made sense. I was super convinced it's real, but it wasn't. And the background gave it away. Look, that op done superfect blur was the little clue I missed. So when you try to identify whether the picture, the photo you are looking at is AI generated or not, ask yourself, does the image look too real? Because if blur is one extreme of AI photos, hyperreality is the other. Sometimes AI images look so sharp and polished that they start to feel unreal. As if every color and texture has been turned up to maximum. In the photo here, there are several clues that the photo might be AI generated. But what really gives it away is how perfect, how perfect everything looks. The leaves on the trees are almost glowing, unnaturally crisp and luminous, and the line of trees is weirdly symmetrical. Like someone arrange them with a ruler. It's the kind of balance you might see in a manicured hatch, not on an ordinary city sidewalk. And that's the giveaway of AI hyperreality when the world looks clearer, brighter, more intense and colors, and more precise than the real one ever could. So look at those pictures, look at those photos, and ask yourself, do they look real to real? And also, look, if you look at the spot where she's holding her mother's hand the details melt together in a way that doesn't quite make sense. Context plus when the story doesn't add up, even when an AI photo looks flawless, technically, there's often something off about the story it tells. Look at the details in the background. The soldier might be wearing the wrong era's uniform. A building might feature architectural styles from three different continents or a hand might rest on an invisible table. It's because AI still has trouble with logic. It knows what objects look like, but know how they interact. They will put a person sitting comfortably on a chair that doesn't have flex from time to time. Even facial expressions can feel mismatched. Like everyone in the image in the photo is reacting to a slightly different moment. And that's because they are. Each phase is generated independently. So the emotional rhythm never quite sings. Sometimes the background looks very washed or generic. It's meant to support the main subject, but not tell a story. So AI doesn't pay attention to generate so many details. Learning to see differently. Yeah, I have to tell you the point isn't to become paranoid because we can easily get paranoid, right? I tell you all of this because I want you to become aware and more observant. Will you I need your pinky promise. Because learning to spot fake photos and videos, it's not meant to be about distrusting everything. We just have to look a little bit longer than we did before and use our human intuition in our eyes, especially when a photo sparks strong reaction in us, for example, outrage PA. We need to take a breath before we consider sharing and passing it further. Zoom in, check the text, to delay the background. Because, yes, the truth is those photos are getting better by the week. The visual scenes are closing, details are shrinking, and soon the difference between real photos and fake might not live in the pixels at all as all the Muriels get better and better. So maybe the most impressed skill in this new age, because I think we can call it that way, the new AI age won't be about technical literacy. Maybe it will be about emotional literacy, the ability to feel when someone feels right and real looks right and real but doesn't feel right and real. 8. The Massachusetts Institute of Technology (MIT)’s Media Lab List: Researchers at the Massachusetts Institute of Technologies Media Lab decided to do something powerful. Instead of just writing academic papers about deep fakes, they wanted to help ordinary people, everyone see them and to experience firsthand how tricky it can be to tell a from the real thing. What did they do? They built a website called Detect fakes, an interactive project where you can watch photos, some AI generated, some not. You can try to guess which is which. The idea is simple. The idea is that by practicing on real examples, you will start noticing the small visual details that AI generated Media Away. The project builds on a massive research effort called Kegels Defect Detection Challenge, a global competition backed by AVS, Facebook, Microsoft, and the partnership on AI. And the challenge offered 1 million price to teams who could build the best machine learning models for detecting defects. So it's pretty big. It was pretty big. And MIT's approach was a little different. Rather than fine tuning the perfect algorithm, they wanted to explore something more human. How can awareness and practice make us better detectors of AI manipulated media? So their hypothesis was that if people are exposed to AI generated media and to deep fakes and learn to spot what feels off, they will become better at recognizing deception across all kinds of AI generated digital content. So as I've already tell you, they launch detect fakes website using thousands of curated media from the FDC dataset. And the latest version of the site features 32 short images, and we can click through guess which are real and instantly see whether we got it right. So that's really exciting because we can check it. We can check how observant we are in real time and get the answers and the clues right away. As you can see here, the project also shares subtle clues like little sign. You can look for and trying to decide whether this photo this image might be fake. These aren't hard rules, but rather patterns that often reveal the manipulation. And MIT also publish the list with the eight signs of a deep fake. So I curious what they share here, what we need to look for, and what are the most common signs? It's really good to know them. Focus on the face because most deep fakes revolve around facial transformation, swapping one face for another or sadly adjusting expressions. So you need to study analysis there. Look closely at the cheeks and forehead. Does the skin look unnaturally smooth or strangely wrinkled? Do the facial features match the person's age? Deep fake sometimes combine a yellow full face with older eyes or hair creating this very odd mismatch. Check the eyes and eyebrows. Lighting and shadow, as I've already tell you, how to fake. So you need to analyze. Are the eyes reflecting light naturally. A shadows falling where you expect them to. Because the fakes often miss the real wet physics of light in depth. Pay attention to the glasses. Glare or reflection on glasses can look wrong. Sometimes there's no glare at all or too much of it and it doesn't shift properly when the head moves. Analyze facial hair. Birds side burns or stable can be added or removed digitally, but AI still struggles to make them look realistic. So headlines in some pods might blur off liquor slightly. Notice facial moles are small mocks Because deep fakes can forget them entirely or make them move slightly from frame to frame, and you need to analyze that because it's a good way you can distinguish what's AI generated and what's real because real skin features stay consistent. And AI generated media can you know, the moles, or the freckles can change the structure, be in different position, and it's very often consistent, for example, when you have a few photos of a similar person. Watch the blinking. Older defects famously forgot to blink. I'm here at Harvard University. Just an extraordinary place goes back to the 1600s. Always wondered what it would be like to go to a school like this. Pretty nerdy stuff. Serious nerds. But nice. You will sometimes overcompensate. Either way, natural blinking pattern still can be a clue. Study the lips, because lip sinking is still one of the hottest part to perfect. Out of mouth movements sync perfectly with the wits. Does it feel slightly off like a doubt movie? Pay attention to D. Y'all be right there, Steve. Hey, what's up? TikTok? Look, I do a lot of my own stunts, but I also do a lot of industrial cleanup, okay? It's important. So, obviously, you keep your hands clean, but you need that exfoliating product to really cut through the grime, okay? Just another tip for you talkers, or the tip to the tip took tips. I'm getting too old for that. I don't High quality deface can still fool even trained professionals. But with practice, your brain starts catching the small inconsistencies, lighting that doesn't match skin that looks a little bit too perfect or has too much contrast on some parts, like, for example, on this example. And when it comes to MIT, the detect face project is built on that exact idea. The more we train our intuition, the harder it becomes for AI generated media to manipulate us because we are very observant and very careful. So you can try to experiment yourself right here, watch the photos, make your guesses, and see how easily our perception of truth can be tricked. And now let's try to identify a few together. It will be fun. That's what she said. And I have to be honest when I first try those tests and was so sure I'd be great at recognizing the fake ones. 100% accurncy no problem, I will do it greatly. Easy. Or at least that's what I told myself before jumping into the MIT media detect fake experiments to see if I could tell an AI generated image from a real one. I came very prepared. Bring it on, i. And I was like that. I was confident, maybe even a little bit cocky. And then I got fooled. Yeah, I got fooled again and again. Oh, now let's look at a few examples together and break down what's worth paying attention to. As I mentioned earlier, so many times, it's always worth zooming in on the details. That's where the clues usually hide. Some of those images are unfortunately low resolution, which makes it harder to zoom in properly, I know. But, for example, here, the license plate looks okay at first glance. However, once we look at the car model badges, more doubts start to appear. Image editing makes the task coder, of course, but the markings on the ground also feel inconsistent and not very coherent. So let's take a closer look at the wheels and the car door handles because details like these often reveal the most. This is this wheel looks like it was generated imprecisely. The seat reflected in the window also appears unnatural to me. And I want to emphasize that I'm fully confident this image is AI generated. For me, what I found, the things I've already mentioned, it is already sufficient and very clear evidence. So, am I sure? Yes, I am? And the verdict? Well, yes, this is artificial AI generated image. So on to the next, at first glance, this one seems less streaky. The cabinets all have identical handles which already suggests this is a real photo. On the counter, we can see an menu that holds topographically. It respects spacing, enlightenment, and clean lines, which, again, points toward this being a real image. The background details like the items in the display case, also support the idea that this is the real photo. The only thing that makes me pause is that yellow object and why the guy in the blue shirt is holding one on the counter or stove. But the resolution is too low to fully figure it out, I think. And for me, this is a real photo. Let's check it. Voila. It is a real photo. That was easy, right? And how did you vote? I'm really curious. Let me know. Let me know in the discussion section. I'm really curious. Now let's take a closer look at this image. What do we have here? A fancy restaurant, but it looks like an older image because the flaws are very obvious. Let's look at the fabrics, faces, and tree leaves. In a very small size, this photo might pass as real. But at this size, when we zoom in, everything is clear as day. This is definitely I generated. So how did you vote at the beginning? It's worth going through many more examples because they really vary in difficulty. These ones were fairly easy. There were big red flags, but I've also come across cases that were much less obvious. So in the next chapters, we will go through many more of them together because even if these seem easy, I thought so too at the beginning. But I will show you a little bit later that it's not always as simple. And you can experiment with this project, with this site and test yourself and you spare time and let me know in the discussion section how it goes and if you are getting better and better at noticing patterns and noticing the fake ones. I think people are really ply to wear a couple. I think most people believe story of Miller. 9. Time to Put Your Skills to the Test!: Another way to put your skills to the test. We've already spent some time discussing how to spot AI generated images, but as you already know, theory is one thing. I have another very fun and very practical but very educational way to learn to distinguish and decide what's real and what's not so now it's time to see how well your eyes actually work in practice. And I love this. I love this. I have to tell you I'm a little bit addicted to this. So I think you will get addicted to it, too. So, listen, a creative redid user who's been experimenting with my journey since the early days had the same questions as we did. How well can we really tell the difference between real photos and AI generated ones? So he built a game called Reality Check to find out. In this simple yet very addictive task, you will see two photos side by side. One real and one generated by AI, and you will have to decide which one was taken by human. After each round, you will see the correct answer, along with the source of the real image and the AI prompt that produced the fake one. It's fun and it's very humbling, and it's a great reminder of how convincing AI tools have already become. So let's try it together, and it's time to see how sharp your instincts really are. The author of the project set as the modal started getting scary good, I started wondering, how well humans can actually tell AI generated photos from really ones. Everyone, I know, thinks they can spot AI generated content, but can they really? I've noticed it too in my colleague circle that everyone says they are great at spotting AI generated media. Sometimes their awareness is sleeping and they forget to stay careful because they are so cocky and so short that they can't be fooled. No, every one of us can be fooled nowadays because AI generated content is so good. So the times has changed, definitely. So let's test your spotting skills and see how sharp your I is. And, of course, the website is free, requires no registration. Every few days, new photos appear on the side with a fresh theme. The challenge is always changing. Right now, in the moment of recording this material, the topic is messes, actual messes, spilled drinks, cardic rooms, and general human disasters. This theme expires in about an hour, by the time you visit the website, by the time you watch me here talking, the focus will almost certainly be something completely different. 10. Test Your Eye! AI or Real Photo? Grocery Shopping Edition: Cast your eye. AI. A real photo. Grocery shopping addition. Okay, I got feedback that what you enjoy the most is going through more examples together. And you said that it's kind of addictive. So let's do it. More examples and more practical tips on what to look for. And good news. In the meantime reality check platform. Has been significantly improved. You can now go for sets from the past and choose specific themes. That means you can explore all photo categories and play past batches. So for now, let's go for the grocery shopping category together because I think that's a great warmap. Why? Grocery stores are full of texts, labels, and tiny details. So we will have plenty to point out and show you. Let's go. Here we've got two supermarket oils. At first glance, both photos looks a little suspicious. No, but no, only one of them was actually generated by AI, so we have to find it. Let's look closely at the text on the boxes because that can tell us a lot. And remember, clean, consistent fonts and lots of detail are usually signs of real photos. Small inconsistencies or weird looking letters still show up very often in AI generated images. So as you can see at first glance, everything here looks seems fine. Lots of details. Nothing jumps out at me yet. It's always worth checking whether the same products are truly identical, because that's where AI often slips up and gets confused. So I'm going back to the text. Yeah, we are checking those. And now I'm going back to the text on those first boxes because they can be a good point of reference. Nothing obviously suspicious. Do you see anything? Do you see anything? Now, let's move on to the second photo. And again, we scan the details and the text. This is exactly how we look at everything slowly and carefully to sharpen awareness and learn the patterns. There are lots of tiny details here, but they are so blurred that it's honestly really hard to judge them. But let's go back here for a moment. Oh, I think there is our first clue. The same Zoop, the same product, but the graphics are a bit different. Do you see it? Sure. These could be different flavors, but the ones next to them also have subtle differences, and that's already very suspicious. Here, for example, this one, okay, probably a different flavor, but there shouldn't be any differences in the company logo. Now I go back to the text with extra caution, and here too, the company logo is slightly different, and in real life, those differences don't exist. So there we have it. We've got our clue. This one is generated. And let's move on to the next one. Will it be trick here? Okay, let's start by looking at the details first because they are the most important. Are these pasta packages actually identical? We obviously have to check. That's how you look for the first clue. We keep investigating step by step. Now let's check the water bottles. The cups look different. These are a few small red flags here and on the labels, too, but they could also just be turned in different directions. So that alone isn't a verdict yet. Different cup thickness on products from the same brand definitely raise suspicion, but it's still not enough to say for sure. So we keep looking now these texts, unfortunately, already look very yogish. They're way too sharp and crisp compared to the rest of the image. It feels like someone just told the model and the prompt exactly what the text should be on the shelves. What do you think? So we move on. I keep searching, scanning everything, and oh, man, there it is. The same product, but one of them has clearly disserted letters. Do you see it? That basically screams a eye. And right next to it, we've got another confirmation, I think, on the milk. Totally worth digging for clues. It's always really satisfying when you find them. And just so it doesn't seem like I'm ignoring the photo on the left, even though I'm already sure about the one on the right, let's check whether everything here looks okay. Lots of details, believable details. The font isn't overly sharp and the same products don't have glaring differences in their logos. Yeah, I'm certain now. The one on the right SAI, honestly, 100% certain. Okay, on to the next. And so, okay, both of those grocery stores look pretty unusual. Which one doesn't actually exist? We will figure it out together, okay? So one on the left one, that scale looks pretty unusual. And my first thought is, did someone mention this in the prompt and just get a little imaginative? Or is this actually a real thing in that store? These days, when you look at unusual places, you automatically become extra alert. Do they really exist? But the details on the scale seem fine. The numbers on the dial look correct. No red flag so far. And here we've got some kind of menu. Even though it's only partly readable, it makes sense. There is no Ai generated fake Gibberish text. It mentions something about coconut flavor, cocoa flavor. Sounds tasty, right? So we keep scanning more details because you never know where something might be hiding that suddenly screams, This photo is i. I honestly have no idea what this product is, but everything else still looks fine. It doesn't trigger any alarms. So what about the one on the right? What do we have here? First, weird thing, the same product, but different labels and different local colors. Let's search the story a bit more. See what's hiding deeper in the background. The man's hand looks a bit unsettling, but he's in motion, so some artifacts could appear naturally there. Maybe the lamp will be our clue. The way the cable is mounted on that lamp is pretty strange. I've never seen a construction like that before. I think I need to take a closer look. Mm. For me, that's already a big question mark, to be honest. So we keep going, and it's always worth checking door handles and grips. And yes, something weird pops up right away. Do you see it? They are different and they shouldn't be. Bingo. This one is a y. Onto the next one. So at first glance, I'd say both of those looks like AI because they feel pretty artificial. But that's probably also just the lighting and the photo processing. Only one of them is actually AI. So details, details, that's what matters most. We have to look at the details, okay? Here, not all the onions or maybe licks have the typical ending you'd expect them to have. But I guess they don't always look perfectly defined like that, so this isn't a red flag yet. So let's check the prices. Everything here seems fine. Clear, readable, realistic looking numbers. In the photo on the right, unfortunately, nothing is readable, so we can't really judge anything based on that. So next question, do the vegetables looks real? Do the details here give anything away? In the left photo, there are lots of details that AI usually doesn't pay this much attention to. Like the markings on the labels, for example. In the image on the right, though, something is already bothering me. The ends of the lettuce look really strange. Look, almost unnatural. I've honestly never seen them look like that. I keep looking for other clues, but yeah, the price tags aren't giving us anything useful here. Still, nothing on the left raises any suspicion, so the one on the right is a eye for me. Bingo. What do we have here? Two shopping carts in a store, both with packed meat among the groceries. Both packs of meat look fine, so we need to look for other clothes. There's a lot of the lorry brand here, but the logo looks pretty much the same on every package. But here, yeah, this is not good. This immediately screams, AI couldn't handle brand details. So we've got our biggest red flag right here, and instant certainty right away. That one was easy, I have to admit. Look, just from the green leaf brand alone, you can tell it's AI because the name wasn't generated correctly on every package. And here, you can even start guessing what kind of prawn was used to generate this. I was curious whether the creature told it to generate products from the green leaf brand, but clearly not this time. So on to the next. And once again, we've got the healthy section of the store, veggies, colourful vitamins. So what could possibly be suspicious here? Unfortunately, for medios labels are way too sharp, way too uniform, and in that very recognizable lately everywhere iPhone. So I'm almost certain already. The vegetables themselves actually look fine, but the text makes you think this photo wasn't taken with a phone or camera. And, of course, on the lattes, we've even got those kind of leathers, blurry, smudge shapes that pretend to be text. Very, very much in that classic AI style. Now let's look at the right side. Here, everything feels more natural when it comes to details, countries of origin, more information on the price tock, important stuff for a real customer. Different colors, more variety nuance. Oh, especially this packaging here. It really supports the idea that this photo is real. Meanwhile, back on the left, if we keep checking, look at those mushrooms, different texts and different illustrations on the same product. Yeah, come on. That's obviously a eye. Obviously. Okay, this time we are in a less healthy aisle, but hey, who doesn't crave lace every now and then the taste of childhood? The empty shelf and the photo on the left makes me uneasy, but with this many heavily branded products, I already know the details. We'll give away which photo is A. So let the investigation begin. And the photo on the right, there are details that immediately boost its credibility. For example, the pool sign on the door. That's a good sign. So now let's zoom in on the chips on the left. Oh, no, oh, no. We've already got our first red flock. Different illustrations of potatoes on the same flavor. I told you, details expose the AI version fast. And on this one package, something is clearly messed up in that very typical AI way. Meanwhile, in the photo on the right, everything is consistent. The same products have the same texts, colors, and graphics. And on the left, some of those other items don't look suspicious at first. But those lace backs and their illustration already told us everything we needed to know. I'm sure of it. That one is AI. Okay, on to the next. Okay, so now we are looking at the world from the perspective of a shopper pushing a card, and here those little details on the card at exactly the kind of complexity that as you already know, AI sometimes fails to handle properly. And these are tons of details here, brand names and logos that are easy to read. Nowhere Gibras leaders and show some things are a bit blurry because of the resolution, but you can still decode them without any trouble. So, honestly, I don't see any red flags at first, do you? And does that mean they are hiding somewhere? Let's look. You have to scan everything. Oh, and this photo even captured details like the condition of another shopper's hair. Hey, but look here. Don't those shopping cart wheels look weird, different heights. That's a big red flag. For me, already that already means this is AI. And, yep, there we go on this card, too. So we've got those irregular classic AI errors and no mistakes. The mesh barriers near the checkout. This one looks like this, and that one looks different. Yeah, no. We could keep looking, but we already know what we need to know AI. Okay, here we go again. The same thing. Probably a mistake, but hey, every repeat is a good practice for sharpening our eyes and our awareness, right? So do you remember which photo was a eye? Do you remember which details exposed the one that was pretending to be real? I think, of course, you do. Alright. Let's keep going. Okay, now we are at the deli. One of these photos is style to look a bit older. Maybe that's meant to lure us into lowering our guard. What do you think? Let's start on the left with the prices. Yeah, forget. We are we are looking at everything. We are examining everything, absolutely everything. And right away, we can see there are lots of details in this photo. Now on the right, Mm, wait a second. Peaches and fresh peaches. And the cheaper ones, what exactly are they supposed to be if they look exactly the same? That's super suspicious. Let's take a quick look at the surroundings, too, even though I'm basically already saw, but a little confirmation never hurts. Observation is something you have to practice. The rest seems fine, but those peaches, yeah, I think I'm shaw. What about you? Bingo on to the next. Okay, once again, we are back in the healthy aisles. This font here looks a bit like it could be e, but I also know it's not a font that belongs exclusively to AI. So next up prices and details. Unfortunately, the details are pretty far away, which makes it hard to be really confident. You just can't see that much. The lumps differ slightly from one another, but that could easily be a perspective thing. So it might be a clue, or it might not. We don't have any handles here, suddenly or identical identical products placed next to each other to compare. And as you know, that's exactly where AI often fails. So I love looking at the products which were supposed to be identical but orange. But here we have nothing to compare. On the left, these fruits raise some suspicion for me because some of them are very irregular. And this right here, how is that even lying there without falling out of the basket or without customers knocking it over as they walk by? There's no way that wouldn't fall, no chance. And here we've got a lot of really huge onions. I actually start wondering if they are too huge. Onions bigger than apples. I mean, okay, technically possible, but I honestly haven't seen onions like that in supermarket in a very long time, let alone that many of them all that massive. The rest of the vegetables are too far away to really tell us anything, but those onions definitely made me think made me stop and think. So let's keep looking, checking for other clues. But for now, the onions are clearly telling us this photo is much more suspicious. And this fragment here also raises doubts. Okay, I'm sure now. What about you? At the next. Now we are moving to a completely different part of the world. Open air markets, not a modern supermarket. So what good stuff do we have on the left? Had written prices, lots of different details, a kitty. A box labeled fresh carrot. Honestly, I can already kind of guess this won't be AI. A sign like fresh carrot probably wouldn't show up this clearly in an AI photo. It would likely be blurry or unreadable. AI usually doesn't like dealing with this many details. Still, we keep looking detail by detail. Do we see anything suspicious? Is the roof construction, okay? That's another place where AI often messes up. The lumps and other details look fine, too. No. So that must mean the photo on the right is our suspect. See those baskets with clothes, almost like I don't know, blankets. That's a strange thing to have at a food market, especially displayed like this. And basically only next to this one very vegetable focused stall. For me, that already screams AI because Ai doesn't really understand context. The vegetables themselves don't raise suspicion. They are generated convincingly, and honestly, the veggies look believable in both photos. But in the photo on the left, we even have super realistic details like a cash register. And in the photo on the right, on the other hand, a few things just don't add up. Those baskets in particular. And there we have it. Okay, so we are back in the supermarket, but this time, it's definitely not eco friendly. So much plastic rob. And what immediately catches a eye. Yes, that fresh sign. On the right, on the right, to me, it's once again that font eye really likes. But let's take a closer look at the tomatoes. Maybe they will tell us more or maybe not. On the left though, we've got a much more polished, believable, fresh sign, and all those details for each product, price tags with photos and descriptions. I know I sound like a broken record, but these are exactly the things AI often messes up. And here, everything looks correct without the inconsistencies that are so typical of AI. Everything is clearly labeled a solid, well organized store store. Even the details on the fridge here looks very real. So, yeah, I think we can be sure now. Now, something for Apple lovers. Lots of packed apples. Let's take a closer look. Right away, I can see that on the left, the details look fine at first glance. The meshbg, the way the light reflects on the apple packaging, the text on the apple stand. Yeah, everything looks legit. And on the right, very clear, very simple text. And as you already know, you already know what I think about that. AI does this all the time. Simple text is easy for AI to sharpen, so it ends up too perfect, too sharp, too clean, too readable, while it often struggles with more complex stuff. These here are simple, which is why they are super sharp and almost too readable. On the left, the details feel much more convincing. And look here, some of these apples are really misshapen, almost not apple like at all. Plus, the background is heavily blurred while the sal orange dot sale text is razor sharp. That doesn't follow the rules of optics. Let's keep looking closely and really alize it. But my confidence is already growing, especially because on the left, even though I really tried, I couldn't find a single red flock. So I will show you a few more details so you can look for yourself and judge. But I think I'm already shop. So, yeah. So yeah, I'm already clicking. Okay, we are moving to Asia now. Here, the product names don't tell us that much, but we can still judge whether things feel real, at least as much as we need to. And remember, a great trick with photos like this is checking whether the same products are actually identical. AI struggles with that a lot. On the left, the teas and water look flawless. Some products, same products, same details. The Coca Cola logo also looks correct, which is something AI often messes up. There are tons of details here. Even the clock looks right. Every single element feels solid. I think nothing raises suspicion. So let's grab a magnifying glass and head to the right. Here, the packaging is less visible and harder to judge. Still, you can tell the details org as refined and on some packages like this one here, it looks as if if the elements are melting into one another, and those bottles, the cedar shapes, some of them look like they weren't generated quite properly. So we keep analyzing everything piece by piece. And no one else I see the mirror shows an incorrect reflection, which is very common in AI generated photos because there is no way a store fridge would would look like this in real life with that kind of store layout. Do you see it? So now I think I'm really sure. This one is AI 100%. That's the one. Okay, both photos look like they came straight out of a store catalog. So they are very symmetrical and dominated by artificial lighting. At first glance, let's be honest, both looks a bit suspicious. So let's take a closer look at the text. To me, the lettering the photo on the left gives off less of an AI vibe, but that's just a first just a first subjective impression. We need to keep digging and go deeper. Of course, we also examine the fruit checking for any strange shapes or any anomalies. And we look at the packaging, too, because we know AI most often trips over the details. So in the photo on the left, even the legs of the display stands look real, look regular, well made, which, for me, supports the idea that this one is a real photo. So now let's check the next on the right. Such a nice market, so many healthy things. But wait, here we supposedly have the same product, the same label, the same price below, and yet they are completely different in color. One looks green inside, the other yellow. The rice mushrooms above also differ a lot, even though they share a single price stock. The avocados don't raise suspicion. I'd happily eat them for lunch today, honestly. As always, I also have to take a look at the lumps. I wouldn't be myself if I didn't check whether they make sense. We've also got some kind of list or menu here, but it's very blurry, and it's strange that part of it is in a thick font, while another pot is thin and kind of smirge. For me, this photo is definitely a eye. Well, okay. And we are back in the sweet potato ale. And once again, we've got a big onion smaller than before, I think. What do you think? Of course, we start by checking the price tock. They can tell us a lot, and I hate to say it, but unfortunately, I'm seeing that sharpened form that Ahi really loves. And look at this. Organic red onions twice. And one of them isn't red at all. Sure, a sore employee could made a mistake, but this definitely puts me on high erg. On the right, we've also got bar codes. Something AI usually messes up. Here they actually look okay, which supports the idea this photo might be real. The text is readable, which we like and it doesn't feel AI generated. Let's take one more look at those price stacks. Yeah, it still screaming at me. And now let's check the background. There's a heavy blur back there, so we get fewer concrete clues. But look at this. I find that display near the entrance to the back area really strange. That's a weird spot for a stand with fresh produce. There usually isn't space for something like that there. Now let's double check the right side to be sure there are no red flags. But I even recognize some of the rings. This is definitely a real photo, and that means the other one without a doubt, is AI. Let's check. Yeah. Yeah. Oh, wow. All the spread. S with all kinds of ads in. I'm guessing somewhere in Italy, so many textures, colors, and flavors. I honestly wouldn't even know which one to choose. On the left, we've got much more elaborate, thoughtfully labeled varieties. On the right, do you see this? Risotto a porcini the same sign in several places, but the rice itself looks completely different each time. And up there, Alter tufo but wait. That looks like it has humeric ginger and saffron in it. And above that, supposedly a i porcinieporcin, again. No way. On the left, the textures really match the descriptions and feel way more convincing. The one on the right is AI, I'm sure of it. Okay, back to fresh fruit. I love this vibe. The first thing that jumps out at us is the sign above the shop. The one on the left, shop very clean, super readable font, and on the right, older, visibly worn, but still readable. And here's the first moment of doubt. I'm not sure AI would handle that kind of wear this will. Because every leather and number looks perfectly embosed. We've also got a ton of details here, barra codes, bananas hanging on hooks, details on imported packaging, an important one. You can pay by card, Mastercard and Visa, and AI really doesn't like those icons and often messes up the colors. But here, they're correct. At this point, I already know the photo on the right is real. And on the left, the fruit might look nicely generated at first glance, but look at this. Would those plums really sit like that without falling? I don't think so. So for me, that one is 100% AI. Okay. Okay, and Asia, again, I can always hear the boss and chatter just looking at those photos, but only one of them is real. Which one? The sign look very sharp, very crisp. And my first thought is, this could be a eye. But that alone isn't proof. On the left, the box is packed with details. That's always, always a good sign. The products themselves also have lots of fine details. The colorful packaging and its texture did make me pause for a second. But the rest of the details and especially those little icons on the cardboard cardboard are very well done. The hands on the right look fine. But look at those jaws. AI messed up the texture there. For jars to deform like that, they'd have to be plastering and sitting in direct harsh sunlight. This one is AI, unfortunately, I'm 100% sure. 11. Why We Fall for Fake Images and Fake Videos: Why we fall for fake images and fake videos. I have to tell you sometimes I really think about how hard it is right now to be a human brain. Because we've crossed the tuning point, AI images and videos can look so photorealistic. We've spent thousands of years evolving to trust our senses, to believe our eyes, to follow our instincts, to recognize danger and comfort by sight alone. The eyes were always so important. And now suddenly the world is filled with images and videos that look true but aren't. But we are still programmed by default to believe And look, we are emotional before we are logical. Here is the thing. We don't process images and videos like computers do. We don't scan, verify and cross reference. We feel first because we are well, humans. So when we see a photo of baby animal, we heard instantly melt. We see a picture of destruction or tragedy, and we stomach drops and we instantly feel sad and we feel the empathy for the people who could be harmed in this strategy. This instinct is a part of what makes us human empathy through seeing. But it's also what makes us so extremely vulnerable to AI manipulation because AI-generated images know how to press those emotional buttons. And people who use AI, be people who use AI know how to do that because a fake protest photo that aligns with your political beliefs, yes, you are more likely to believe it. Fake miracle Story frame with an inspiring picture, you might share it before you Fox Check. And no, we don't fall for fake media for AI-generated content because we are foolish, not at all. We fall for them because we care about the things we see. I see it that way because we want things to make sense because we want the world to look like we already believe in it. And unfortunately, our modern brains love shortcuts. They are efficient but sometimes very lazy. Psychologists call these shortcuts cognitive biases, and they quietly shape how we interpret everything. There is confirmation bias, which makes us trust information that agrees with us and doubt anything that doesn't. That's why two people can look at the same fake photo or fake video and draw opposite conclusions. Each see only what reinforces their world view. Then there is emotional bias. We are more likely to believe something that makes us feel something. Outrage, joy, fear, the stronger the emotion, the faster we want to share the photo or the video. And then there is authority by us where an image stamped with an official looking watermark on we banner suddenly feels legitimate to us, and AI tools have learned this trick, by the way. They will sometimes generate fake CNN logos or BBC style graphics, so we need to be careful. And we like to think we are rational, but the truth is emotion always enters the room first. Logic shows up late and tired, asking what happened. AI-generated visuals are so polished that they often feel more real than reality. The lighting is impressive. The colors are cinematic. The composition is exactly what a professional photographer would choose or maybe what a professional photographer wishes they could choose. And our brains eat that up. We've already drawn to visual harmony to beauty to things that fit. Something looks that good. Our instinct is to trust it. I often think when I observe people, for example, in the park, waiting for a taxi, that we scroll so fast that our brains basically don't stand a chance to notice anything pange. In the time it takes to really study a photo, we've already seen five more. The platforms we live on are designed for fast, very fast consumption, not reflection and pausing to analyze what we are really seeing. So when a fake image appears, our brains do what they've been trained to do, react not reflect. It's that we don't pause long enough to apply what we know about AI-generated content. I think the solution isn't to stop trusting the sources we see, but we have to start looking deeper and analyzing more. And it can be learning to pause for 2 seconds longer than usual to ask simple questions. Who shared this photo or video? Where did it come from? What's the purpose of this image existing? Because as the line between real and fake blurs and it will blur because the technology is getting better and better, truth won't announce itself with a label. We will have to notice it and distinguish it ourselves. 12. Tools to Detect AI Images: To detect AI images. Let's be completely honest. Most of us don't have time to zoom into every photo we see online and analyze the lighting on someone's left ear. The Internet and well, our whole lives moves too fast for that. And our attention span and our focus aren't really in good condition right now. But on the other hand, as AI media, AI images become harder to spot with the naked eye we are going to need a few digital alleys to help us separate real images, real photos and videos from generated ones that try to fool us. Thankfully, yes, there are tools for that, small but mighty ones created to help us. But hey, unfortunately, you can't be fully relieved because these tools can replace our intuition. They are more here just to support it. Here are a few methods to tell if that photo you are looking at was ever really a photo at all. You've ever had that gut feeling, wait, haven't I seen this image somewhere before? You are already thinking like a detective, and that's a good thing. A reverse image search lets you upload a photo or paste a link and see where else it appeared online. The two best known tools are Google Images and TN eye. Here's how it works. You drag an image into the search bar and the tool scours the web for visually similar versions. If it finds an identical photo from years ago, say, a stock image from 2017 that's suddenly being shed now as breaking news with a few changes, you just caught a fake news in the act. And yes, reverse searches aren't very helpful for completely generated content for content that was generated by Mangione, for example, but reverse searching especially helpful for spotting recycled content. And we've seen a lot of these, a lot of recycled content these days, real photos modified a little bit using AI, used out of context or AI-generated ones posing as documentary evidence. And if the photo is truly new and doesn't appear anywhere else, that doesn't automatically mean it's fake. But it's still worth raising an eyebrow. Sometimes this silence can speak volumes. So we have to analyze where the photo comes from and what website or what person published. Every photo carries hidden data within it, a kind of secret diary called metadata. It recalls details like when and where the image was taken, what camera would use, even what setting captured it. Tools like accept tools can open that diary for you. They show the raw data behind a photo time stamps, geo location, device information, editing history. And here is the catch. AI-generated photos usually don't have that metadata. They include something generic like created with Stable Diffusion instead of a camera model and all the specific data about setting. So if you find that an image is missing, it's usually digital fingerprints, the usual metadata. That's your first clue. It might never have been taken by a real person holding a camera. But this is very important. Metadata isn't foolproof. A lot of real photos lose their metadata long before we ever see them. Websites often strip it out automatically to make images smaller and faster to load. The same happens on social media platforms and messaging apps. For example, when you send a photo for messenger WhatsApp on Instagram, the active data is wiped to protect privacy and, of course, to reduce file size. So if you open a picture and find no meta data at all, that doesn't automatically mean it's fake and it's AI-generated. It might just mean it's been compressed, optimized or shared online a few times, and the metadata just was, you know, deleted along the way. So metadata can be a clue, but unfortunately, not always. AI image detectors. Now there are tools that are designed to spot other tools. Several platforms are working on automated ways to tell whether an image was AI-generated or not. Unfortunately, they work as digital lie detectors. They aren't completely reliable. They are very imperfect, but getting smarten every day, so there's hope. And for example, hive scans photos for telltale AI signatures, tiny statistical irregularities and textile patterns that human eyes can see. Reality defender goes a step further combining multiple detection methods, including metadata and visual analysis as well to give you a probability scope not just real or fake, but how confident we are that this is real or this is fake. But as I've already told you, unfortunately, none of those platforms, none of those tools are flawless because AI models constantly evolve and detection tools have to sprint to keep up. And I'm afraid that honestly, no amount of technology will save us if we don't learn how to recognize manipulation and how to spot that something is wrong ourselves. Why am I repeating it so many times? Why am I so serious about this? Because, look, sometimes it's incredibly easy to fall for AI-generated media photos, videos, especially if you are not looking at them with a critical eye. While some of these scams and some of AI-generated media seem almost super funny in hindsight, the people who get caught in them are real and so are the losses. For example, in January 2025, a woman in friends lost nearly 830 k euro. After Scammers pretended to be Red Pitt line. I know it sounds hilarious, and at first, I didn't believe it. But yes, it really happens. So it sounds funny at first, but the rest of the history isn't funny. And it began innocently a friendly message from an account claiming to belong to Brad Pitt's mother, which later turned into private conversation with Brad himself. And the impostors told her that Brad was undergoing treatment for kidney cancer. And that his money was temporarily frozen due to divorce. And to make the story more believable, they sent her AI-generated photos of Brad Pitt lying in his hospital bed, along with edited video showing him recovering. And the photos looked convincing enough to fool even careful viewers. So eventually, the woman transformed large sums of money to help Brad Pitt and only later realized she had been scuped. Spokesperson for the real Brad Pitt confirmed that the actor doesn't communicate with anyone on social media and French police opened investigations. Nur screen, fake AI-generated Brad Pitt photos that duped a real woman into a foe, not real love relationships. Scammers posing as Pit swindled a 53-year-old woman in France out of get this $850,000 in an unbelievable catfishing scheme alone the fake Brad Pitt supposedly needed for cancer treatment. The woman even divorced her own husband. The hoax prompting the real Brad Pitt to issue a statement. It's awful that scammers take advantage of fans' strong connection with celebrities, but this is an important reminder to not respond to unsolicited online outreach, especially from actors who have no social media presence. Sadly, this case isn't unique. Similar AI assisted scams have appeared all over the world, not only in France. Because AI-generated content can feel so deeply personal, and when it plays on empathy, loneliness or argency, even the smartest of us can be fool. 13. What About Deepfakes? How Deepfakes Actually Work?: What about defects? How defects actually work? E D fakes were very clumsy, faces flickering, skin tones mismatched, eyes drifting out of sync. Now they can be extremely realistic and they are getting better every week. And at their core, D fakes use machine learning to mimic the human face, how it moves, how it reacts, how it breathes. And algorithms analyze and map a person facial expressions, and then recreate those expressions with a different face layered on top. Under the hood, a lot of this happens through something called a generative adversarial network called GAN. It's basically a battle between two AIs. One AI is trying to create the fake, the other trying to spot it, and the faker keeps getting better until the detector can tell the difference. That's how it works. The result hyper realistic fabrications that can fall even the most skeptical eyes. That's why it's such a clever system and a deeply dangerous one. If you've ever used Snapchat or Instagram filter that makes you look like a puppy, smooth your skin, or swaps your face with your best friend. Congratulations. You've already met a baby version of the technology behind DfX. And while filters are just for fun, thefts are intense over achieving cousins. Instead of cute dog ears, we are talking about videos where someone appears to say or do something they never did, and it looks completely real. So how does that even happen? How is that possible? Let's discuss it in detail. The first step, the machine learns your face. And it's hard. The fake is powered by machine learning, which is exactly what it sounds like. Teaching a computer to learn from examples instead of just following instructions. Let's say the AI wants to learn what your face looks like. It this data, of course, thousands of images or videos of you turning your head linking, smiling, frowning, talking, laughing, sneezing all of it. The more the better. And then the algorithm studies all that footage and starts to notice patterns about you. How your eyebrows move when you are surprised. The shape your lips make when you say, Wow, how your eyes crinkle when you smile. Of course, the more data it gets, the smarter it becomes. It builds a digital mask of your face, one that can move and react just like the real thing, just like your real face. The second step, the great AI face swap. Now that AI knows your face, it can start swapping it onto someone else's body. Okay, I know. Imagine the algorithm as a super advanced digital makeup artist except instead of foundation counter, it is his mouth. It takes person's A face, for example, Tom Cruz and person's bee bee movements. Say a stand in actor golfing. Then frame by frame, it paints Tom Cruise's expressions over the actor's face, syncing perfectly with every blink, every smile, every hear flick. It's so good at mimicking how light hits skin or how muscles move that your brains get strict. You see it and think Okay, yeah, looks real. The first step the AI battle, meet the JANs. Okay, here is where things get science fiction call. Most defts use something called, as we've already told you, a generative adversarial network called JN and it's two AIs. Locked in in an endless competition. You can imagine it like this. The first AI is the artist. Is job is to make the fake, to create the most convincing video it can. And the second AI is the detective. Is job is to look at that fake and shout fake every time it spots something off. They play this game over and over each time the first AI, the artist gets a little better a fooling and convincing the detective. The detective in turn, gets sharper at spotting flocks. They keep pushing each other until the fake one, the fake version, the deep fake becomes so realistic that even the detective can tell it's not real. That's when the video graduates into a convincing deface. Tilted a bit. This idea came from two computer scientists, Ian Goodfellow and Joshua Benju who came up with JMS in 2014, originally as a creative experiment, not a tool for deception. And now the fourth step, the finishing touches. Once the EI creates the defak editors might clean it up just lighting smooth edges or match the sound to the mouth movements, sometimes with AI voice clones. And at this point, the fake is nearly flawless. You could play it side by side with the real person, and most people on notice the difference. That's why it's both fascinating and frightening as a technology, a masterpiece of mimicking that can also destroy trust in what we see and hear. So as you can see, deep fakes aren't just a tech issue. They are a human issue. They touch on everything that makes us who we are, our dignity, our privacy, our ability to trust what we see on here. And yes, they can be used to humiliate, to deceive, to control. They can twist elections, ruin reputations and erode the shared reality that democracy depends on. So I think we can argue with that. Defects aren't just violating human rights. They are fretting our shared sense of truth, as well. And as with most things in this EI era, this technology isn't inheritently evil. It's what people do with it that defines the story as defines how the world will look like. 14. AI-Videos and… How Fast It Progresses: EI videos and how fast it progresses. EI video models have evolved so much over the past years. Remember that in famous Will Smith eating spaghetti clip from April 2023. The one that looked like a fever dream of melting noodles and disorder faces, it was made using an very early version of stable diffusion, and at the time, it was both hilarious and horrifying. A funny showcase of how weird AI video could be. But it was, you know, it wasn't realistic, and we were all laughing at it. So if there is one video that perfectly captures how far AI video generation has come, I think it's this famous Will Smith eating spaghetti club. The clip went viral, not because it was good, not by any means, but because it was so bad. It highlighted exactly where AI video stood at the time. Yes, impressive in concept, but completely convincing in execution back then. And for months, Will Smith eating spaghetti became a running joke and unintentionally a benchmark. Anytime a new video model was released, people compare it to that clip. Okay, but is it better than Will Smith eating spaghetti? And that's the funny part. It became a kind of cultural measuring stick for AI progress. And today's models like Google's video or Open As Newest Sa produce videos that look cinematic and really livelie. And when you watch them side by side with the spaghetti video, the difference is huge. So in less than two, three years, we went from very unrealistic soap because to footage that could almost fool the human eye. What started as a Mm has quietly turned into a historical market, proof of just how fast AI generation is evolving. Now, fast forward a few years once again and compare that early cost to Google's view two demo or the recent Sora video by Chad Nelson from Open AI. I think the difference is jaw dropping. Movements of fluid, lighting feels cinematic, and the characters faces almost pass for very real people. So this leap in quality is unlike anything we've seen in media technology. Still, of course, it's not perfect yet, because a few key areas still give the illusion away. For example, realist, we are crossing that invisible line where AI humans stop looking creepy and start looking believable. In the vOEMO, for instance, the DJ and the doctor looked impressively livelie until you notice the subtle sniffness in the expressions. Humans are incredibly good at detecting even the smallest irregularities in faces. We've been wired to read them since we were cats. Interestingly, most demos is quick cuts or short clips, so we don't see much emotional range. And, no, that's short clips, and the way of editing isn't a coincidence because holding longer shots, longer videos would make the imperfections more obvious and more visible. So quick cuts are just safer. Audio and lip sync. Sinking realistic mouth movement with natural audio is challenging. Tools can align facial movements with dialogue or reference video, but they are not always seamless yet. The timing often feels a bit, a little bit off, or the lips move in ways that don't quite match human muscle behavior. It's very close to perfection now, but not always there yet. So let's dig deeper, shall 15. How to Spot AI-generated Video: To spot a generated video. There's something about video that is to feel sacred so far. For generations, cuddle camera was the ultimate proof, the thing that silenced all the doubts. A photo could be edited, modified. A story could be exaggerated. But a video that was evidence. Come on. That was pure reality, but just record it. And now, that certainty is being taken away. Those realistic clips of people doing or saying things they never did have changed their roles. They can look authentic, sound, convincing and travel across the Internet in seconds. But as you already know, if you know where to look, you can still spot some signs of AI generated videos. Let's discuss them. How long is the video? Here is one simple but surprisingly reliable clue when you are trying to figure out whether a video was made by AI or not, the length of it. Even with the incredible progress of tools like Google Video free, US version of Open AI, Sra, nano Banana, and the rest, most AI generated clips are still quite short, very often under 15 seconds. Why? Because generating video isn't only about creating a few realistic frames. It's about maintaining consistency across hundreds of thousands of frames in the row. The longer the video, the harder it is for the model to keep everything stable, lighting, physics, motion, facial expressions, and continuity from one frame to the next. So if you've ever noticed an AI video that looks perfect for the first few seconds and then starts to drift a little bit, maybe the background shifts objects, like, for example, earrings melt a little or someone's clothes starts to change that's exactly what's happening. The model loses track of what it created earlier. Long continuous shots also require temporal memory and understanding of what happened a few seconds ago, and current AI model still struggle with that. Instead, they are trained to generate short bursts of coherent action, which is why so many demos and promotional clips use very quick cuts of montage style editing. Because each short segment hides the modal weaknesses by resetting the scene before errors build up. So for now and big emphasis on for now, if you come across a single unbroken few minutes video where someone moves naturally, speaks fluidly, and the lighting and background stay consistent, chances are is real footage. But of course, this is changing pretty fast. Both Google and Open AI have said their next generation modus will aim more than a few minute long continuous videos. But for the moment, duration remains one of the biggest gas check tools, for potting, AI generated media. So short and flow less or longer, but editing and including many different cuts maybe AI. Long and continuous, most likely real. I think it's a small clue, but really a powerful one, especially as the rest of the visual clues will slowly start disappearing. Watch what changes when someone leaves the frame. Here is another subtle but powerful clue. Watch what happens when a person leaves the screen and then comes back. AIV models often struggle to maintain this visual continuity across cuts of scene transitions. When a character disappears from view, the model temporarily forgets what they look like. So when that person reappears, small details are often inconsistent. So look closely, and you will start to notice things like their clothing suddenly changes color of fabric. A blue shirt tends a little bit green or a jacket becomes much solder. Jewelry that mysteriously swap sides, vanishes or reappears, for now earrings can be a big giveaway still. And also slight differences in age or facial features. The same person might suddenly look suddenly younger, older, or have a different hairstyle after each cut in each scene. Even small props like a cup, a phone, or a necklace can shift position or shape between scenes. You already know that. That's because most current III models don't yet have a stable memory of the characters they generate. They create each new scene from scratch, guided by a prompt or context, but not by continuity logic the way human filmmakers use. And as I've already told you, this is exactly why many AI generated demos videos rely on those quick edits and cinematic jump cards. It helps hide these glitches. The moment the camera cats away, the modile gets to reset and doesn't have to perfectly match what we came before. In contrast, in a real film video, these tiny details stay consistent. The same outfit, the same lighting, the same person, aging only in real time. And that natural continuity is surprisingly hard for AI to fake for now. So the next time you are on show where the eclipe is real or AI generated, pay attention not just what you see right now, but to what changes quietly in between the cuts. The lips are the hardest part to fake. As I've already told you in the previous chapters, even with cutting edge technology, AI still struggle sometimes to perfectly synchronize mouth movement with speech. This ocean, it's a force, a wild, untamed mite. And she commands your w with every breaking light. The result, lip things that are always right, but sometimes not quie. Sometimes you will notice words that seem split second off from the sound like a badly dubbed movie. Other times, the shape of the mo doesn't match the sounds being made. Oh, boy. I love accents. Can't wait for summer, eh? Need to get inside. It's so cold. It's freezing up here. I just love my voice. Wow. This place is amazing. You might see rounded lips during E or still lips when someone's supposedly talking fast, so that would be not possible in real life. Those signs can be very, very subtle, but once you start looking for it, you can spot it from time to time. And if you ever watch a video and you notice someone's speech seems mechanical or detached from their facial expression, yeah, you need to dig deeper and stress that feeling and look for more clues because real emotion shows up in the muscles around the mouth. Not just the words that come out of it. According to this old sea chart, the lost island isn't myth. We must prepare an expedition immediately. Blinking. Eyes are the windows to the soul, and we apparently also to AI generated videos detection because one of the earliest giveaways of fake videos was, of course, the blink rate. For a long time, AI generated videos didn't blink at all. Yes, it was the time when yes forgot the humans close their eyes sometimes. That has improved massively since, but blinking still sometimes look unnatural, too infrequent, too stiff, or too perfectly timed. The reason is fascinating. Blinking isn't just a mechanical action for humans. It's an emotional one. Because, look, try to notice that that week. That's a very funny thing to observe. We blink more when we are stressed, and we blink less when we are very focused. And we blink irregularly when we are nervous or lying. And AI models don't understand that nuance just yet. They just insert blinks where they think they belong. You can also watch for strange micro movements, eyes that don't track naturally pupils that seem to vibrate or gases that feel a bit detached. AI videos and people in AI videos often feel like they are looking through you. That scary a bit? Shadows light and the physics of reality. Light is very honest. It follows the rules of physics. It always has direction and consequence, and in AI generated videos, it often doesn't. Watch how shadows fall across a face or background. Does the lighting shift logically when someone turns their head? Do reflections match the light source. Does the skin glow unnaturally like it's been smoothed by invisible makeup or invisible soft box? These are very small clues that the AI has blended multiple images or fail to stimulate consistent light. Sometimes you will even see shadow go, sad outlines where the face doesn't perfectly align from frame to frame. And it's not just the face. Clothing, folds, hair movements, and jewelry reflections can all give AI away. The light behaves very inconsistently or super weirdly, that's your clue to stop and dig deeper. Even if the face looks convincing, the rest of the body might betray the fake. For example, de fake sometimes render bodies that move slightly out of sync with the head. A tell that's too sharp, a gesture that ends too early. So you need to pay attention to hands and shoulders as well. If they seem very stiff, or the emotion feels a little bit too floaty, you are probably looking at a face layered onto someone else's frame onto someone else's body. And here is the other giveaway. Emotion. Real humans move when they talk. They move us. The emotions appear. For example, anger makes us more tense, sadness makes us you know, shrink, we shrink our shoulders and softness makes us smaller. The effects can mimic movement, but not intention and not spontaneous emotions. 16. AI-Generated Audio & Voice Clones: Voice Cloning Technology Explained: I generated audio and video clouds. There is something magical and slightly embarrassing about hearing your own voice played back to you. I think we all had that moment. Do I really sound like this? Wait, do I really sound like that? Now imagine the same video, the same voice saying words you never said perfectly, confidently with your tone. With your rhythm and even your laughter. Yeah. That's the strange wonder danger of AI voice cloning. In just a few years, synthetic voice technology has leaped from aerobotic monotons to human like replicas that can whisper, cry, joke, or sutter. Just like we do. It's the kind of leap that feels straight out of science fiction, except it's already here in your podcast, your Tik Ts, and sometimes, unfortunately, your phone calls. So let's talk about how it actually works. Before there were voice clones, there was text to speech, the original AI voice. You've heard it everywhere. Si Alexa, GPS directions, automated customer service menus. These voices read text aloud by using pre recorded phonetic samples stitched together into words and sentences. Text to speech systems were functional, not emotional. They sound Well, fine. When they are giving you directions when you try to reach the nearest Starbucks for your favorite, Mo karate or okay, there is no such thing as more karate, I think. I drink only American lately, so I don't know. So those voices sounds okay when they are giving you directions or reading your calendar, but they don't feel like anyone. They are voices without history, without any deeper personality. Voice cloning, though, is very very different. It's not about reading text. It's about replicating identity and little details about someone's voice that makes the voice theirs. So a clone voice is built from a real person's speech patterns, all the patterns, their pacing, tone, accent, and all the tiny queries that make their voice and their way of speaking unique. All have those things that makes our way of speaking very unique. So it doesn't just copy how you pronounce words. It learns who you sound like when you feel different things and different emotions as well. So where text to speech asks, what should the words sound like? Voice cloning asks, how would you say it? So to clone a voice, and IIMdal needs recordings of that voice. Sometimes even a few hours, sometimes just a few seconds can be enough. The clearer and more natural the audio better, of course. The model then studies it in microscopic detail, pitch, tone, breathing, timing, mouth shape, even the faint grain in device, that soft imperfection that makes each of us sound like ourselves. So it breaks all that down into a blueprint of your speech. Then when you feed it new text, the AI uses that blueprint to generate audio that sounds like the same person only speaking well they've never spoken or saying sentences they never spoken. So under the hood, this involves complex neural networks that analyze and reproduce waveforms, patterns of sound energy that makes a voice a voice. But the simpler version is this. The AI lends your voice the way a talented impressionist would. Are now dozens of platforms that make voice cloning accessible to anyone. Some of them are good, some less so. For example, even loves has become one of the most popular and realistic voice generation tools. It can replicate a person's voice from just a short clip, adding convincing tone, pacing, and even emotion. Their technology is breathtaking and also terrifying in its accuracy many times. Sherlock Holmes' quick eye took in my occupation, and he shook his head with a smile as he noticed my questioning glances. Beyond the obvious facts that he has Sherlock Holmes' quick eye took in my occupation, and he shook his head with a smile as he noticed my questioning glances. Beyond the obvious facts that he has, at some time done manual labor. Alright, so get this. I'm at this new cafe downtown, trying to act all sophisticated with an espresso, you know? Alright, so get this. I'm at this new cafe downtown, trying to act all sophisticated with an espresso. Right now, this script is also pretty popular. It offers an over dup feature for creators and podcasters. You can record a few minutes of your own voice and then edit your audio just by typing new words and the clone fills them in seamlessly. Then we have resemble AI, and it focuses on custom synthetic voices for film gaming and accessibility, allowing brands of storytellers to create voices that sound consistent, familiar, and expressive. There are many, many others, too, of course, from open source project to major AI LB. Some prioritize consent and ethical use. Others don't ask many questions. That's the tension of the movement, unfortunately. We've built a tool that can speak in anyone's voice, but we haven't yet figure out how to protect those voices from being misused. So voice cloning isn't inherently bad. In fact, it's already doing remarkable things. People who've lost their voices to illness can now reclaim them through synthesized speech built from old recordings. That's amazing. Filmmakers can restore dialogue for damage audio. Language learners can practice pronunciation with hyper realistic examples. Even accessibility advocates are using cloneed voices to give emotion and new ins back to digital communication. But for every beautiful and wholesome use, there's a docky one scam calls using cloneed voices to impersonate family members, fake political speeches, deep fake podcasts. So I think it's very true that when you can no longer trust the sound of someone's voice, someone's voice, for example, when they call you, the world can become a scary place. In the past, we had the cartonage Nigerian prince or Queen Diana wants to send you monthly emails of the early 20s. But now we have the cleverest scams. The ones that don't just steal money, but also trust. Scammers today don't knock on your door or lark in dark alleys. They live in your inbox, your text messages, your direct messages. Because now, thanks to AI, they can sound like your bank, your boss, your mom. They use the fake clone devices and perfectly written argent scenarios designed to make you panic just long enough to click or tell them something you definitely shouldn't. Some pretend to be customer service reps, other post as charities after a tragedy, and some even create entire fake store fronts that vanish the second your payment clear. It's a strange kind of mod of intimacy. Those people who know just enough about you to manipulate your emotions. They will know you will rush to help if someone says your friends stuck overseas or that you will click if the subject line says in voice overdue. The technology has made it easier than ever to deceive. But the hasn't changed much. They are still just counting on your kindness, our fear, our instinct to trust. And that's what makes it so unsettling because scams aren't just taking money from people, from random people. They are about using the softest parts of us, the parts that want to believe the parts that want to help. And I think a lot of us believe this is exaggerated. Just fear mongering until we actually witness it ourselves. For example, I haven't witnessed voice cloning scam myself, so we fought happily. Please, I hope it won't ever change, but my uncle unfortunately got a phone call where a voice that sounded exactly like his sons said, terrified that he'd been in an accident and needed to quickly pay 10,000 K to get out of a dangerous situation. The caller even gave details, graphic convincing details. And as you can imagine, my uncle was terrified. He thought his son had really been in an accident and was hurt. And, of course, it's it's his son, so he wanted to transfer the money as fast as possible. Luckily, he hung up for a moment and tried calling his son's real number, and his son picked up completely confused because, well, of course, he hadn't had any accident or didn't need any money. But the scomers using clone audio and the fake audio, had hoped my uncle would panic. Make the transfer immediately during that fake phone call. So unfortunately, this can happen to anyone, to random people, to ordinary people, people like you and me, even those who aren't active online, even those who aren't posting on YouTube or very active on Instagram and other social media and never share the voice or videos publicly. My uncle's son, for example, isn't an influencer or a content creator. He's an engineer and a lecturer, and yet his voice was still a clone. 17. How to Spot AI Voice Fakes: How to spot AI voice fakes. I think there's this very strange intimacy to the human voice because before we even learn to read, before we can understand language, we know the sound of care, anger, warm, and fear. The voice carries emotion long before it carries any specific meaning. Maybe that's why AI generated voices. These so called voice fakes feel so scary. They imitate us. Voice cloning has gotten unbelievably good in a short amount of time. And what used to require hours of studio quality recordings can now be done with a 30 seconds clip from a YouTube video or a voicemail. That means anyone's voice, you all mine, your boss, your mom's can be cloned and used to say things we never said. And just like deep fake videos, these synthetic voices aren't just a technical miracle. They're an emotional manipulation tool. But sometimes, if you listen very caffllly, there are still some ways to tell when a voice isn't real, and when there is no real human behind it. Speaking try to listen to the way I speak for a few minutes and notice the tiny details. And you'll notice that real human speech breathes a lot. It rises, falls, stumbles sighs. It has changing grad and texture, warmth. AI voices, sometimes even the advanced ones often sound just a little too smooth because they miss the micro imperfections that make us more believable. The quiver of love, the way our pitch drifts when we get too excited. The tiny hesitations when you think before speaking a sentence that is brave and AI tends to keep everything more balanced. Similar tone, similar pacing, similar emotional temperature throughout. And if you are on the phone with someone who sounds emotionally flat or a little bit robotic in some sentences, especially if they are asking for something urgent, trust your intuition that something is wrong. A real person's voice always carries stress, excitement, or fatigue. Fake one clearly carries the illusion of it, but sometimes that illusion crocks. AI has gotten great at sounding natural, but it still struggles with some parts of mimicking the natural rhythm of our language. I need a drink. Hey, mate. Want to come to the pub? We should get something to eat now. Hey, mate, want to come to the boozer? Liverpool are the champions. Champs. Welcome to Tokyo We are the best in the world. Oh, howdy there. Mm hm. Guess what accent this is. You might hear weird pauses mid sentence like someone is reading from a teleprompter that's skipping frames or an odd emphasis, for example, pronouncing finance as fine as or stressing the wrong syllable entirely. At first, my voice may sound perfectly normal. But if you listen closely, some of my pauses arrive just a moment earlier than you expect. Certain words carry a brightness that doesn't quite match the mood of the sentence. As if my tone forgot what I meant to say. You might notice that I breathe at strangely convenient times, almost like I planned my breaths instead of kneading them. If you pay attention, the rhythm of my speaking stays a little too steady, like I'm following an invisible metronome. Emotion comes through, but in a way that feels rehearsed, like I'm imitating a feeling rather than having one. My laughter, when it appears, ends a fraction of a second too cleanly without the messy fade you're used to hearing from real people. Every now and then, I glide over a consonant too smoothly as though I were more breath than body. When I get excited, my pitch rises, but never quite unpredictably enough to feel spontaneous. Sometimes it's very subtle, like a half second delay before after or a pause that feels a bit too long. And these little timing errors are often the only signs that you are dealing with a cloned voice because while AI can replicate sound, it doesn't understand the emotional meaning of the words. It doesn't know why something is funny or why your voice softens when you say someone's name. Ironically, one of the easiest ways to spot a fake voice is what you don't hear. Real audio and real conversations always carries a bit of the world around it. Room town, breathing, background hum, the faint echo of space. AI voices often exist in a kind of sterile vacuum perfectly clean, perfectly silent. It's like hearing someone speak in a soundproof room, and most people don't have podcast studios in their houses, right? On the flip side, some poorly generated voice clips include background noise that doesn't match the situation, like a consistent static *** that doesn't react to speech or mismatched reverb that makes a phone call sound like it's happening inside Cathedral. So always be careful during phone calls, be observant and use your ears. The most powerful detector you have isn't any fancy piece of software. It's still your intuition and your senses. If something about a voice message feels off, for example, the pacing, the tone, the energy, don't hesitate. Dig deeper, listen to that feeling. Because scammers very often weaponize urgency. I need your help right now. Don't tell anyone. Send me the money immediately right now. We will talk later, okay? Those phrases are red flags even without AI. Add a cloned voice on top, and it becomes a dangerous cocktail of trust and fear. So when you get a phone call like that, take a breath before reacting. Call the person back on a verified number like my uncle did. Ask a question that only the real version of them would know how to answer. You can discuss that question with your loved ones. So when a strange conversation, when a weird phone call like that happens, they will you will know what question to ask. Think about that before. Because, yeah, we are living in strange times and it's really good to always be careful, and you are always careful when you are prepared and when you are very aware of the things happening around you. Then you don't have to be stressed so much because you know you know how to deal with such kind of situations. And we can't stop technology from evolving, but we can slow down to react more carefully. Voice fakes are one of the newest tricks in the disinformation playable and also one of the most personal, I think, because they sound like someone you know, someone you love. And when you hear that your loved one is in danger, of course, you panic. And scammers know it. Scammers know it very well. They know that this is the moment your rational thinking can disappear, and you are more likely to give them the data they need, your password, your money, your codes. By the way, do you remember This website that I showed you in one of the earlier chapters, the same person on Earth is an interactive project from MIT's Media Lab created by the same team behind detect fakes. But instead of focusing on photos and faces, this one is about voices. The idea is very simple. You listen to short audio clips and decide whether they were spoken by the same person or maybe by two different people. Sometimes the voices are real and sometimes they have been I cloned or digitally altered. While we were shooting it, the set burned down the mountains in Malibu, at the exterior locations. But all these buildings, you know, all those metal buildings. All that was left was little puddles of aluminum on the ground. While we were shooting at the Set Burn down in the mountains in Malibu, the exterior locations. But all these buildings, you know, all those metal building. All that was left was little puddles of aluminum on the ground. It's great. Yeah, it feels good to be involved with something that is really advocating kindness in a time where, like, our world seems to reward hate in a bunch of weird ways, right? I'm most excited just to hang out with those guys again. Really? I had so much fun on set, and I really miss that. Kind of freedom. And, of course, the goal isn't to win call points. It is to train your ears. You start noticing subtle things about how real human voice sounds and how to distinguish it. You start to notice tiny pauses and briefs, emotion and tone shifts. The way pronunciation changes with mood. How AI voices sometimes sound too smooth, too flat, or too consistent. And I think after a few rounds, you realize how easy it is to be full even if you are sure you want, you want to get caught and you won't fall into the trap and how sharpening your attention helps you spot when something sounds weird. In short, the same person or not project can help you or maybe you can help your parents, someone from your family, someone from your friends or your loved ones become more aware listeners in a world where AI can already imitate how we look and how we sound. 18. Cross-Verification & Critical Thinking. Source Verification Techniques: Cross-verification and critical thinking. Source verification techniques. Nowadays we scroll through a lot of photos, videos, and catchy headlines every day, sometimes every hour. And in the rush, it's so easy to forget the simplest, most grounding question of all. Where did this come from? In the world where anyone anyone can publish anything and where AI can make that anything looks look so real. The art of verification has become kind of modern survival skill. So, start with the uploader. Who shed this first? Every piece of content has an origin, even if the Internet tries very hard to hide it. If you come across a shocking image or a viral video, take a moment to see who posted it first. Was it shed by a credible source, a trusted journalist, a verified news outlet, a reputable organization? Or did it seem to appear out of nowhere? From an anonymous account with no history. Click through to the profile. Look at the older posts, older things they share. Does this person usually share original material, or do they mostly repose sensational content from others? Do they have followers that look authentic or do they appear newly created. It sounds simple, but most false or AI-generated content spreads because people share it without looking at the source. Real sources have a trial, and fake content, AI-generated content usually don't. Check the upload date and time. Every photo, video, and post carries a timestamp, and those little numbers can tell you a lot. If the video claims to show breaking news, check when it was actually uploaded. Sometimes the same clip will resurface years later rebranded rebranded to match a new crisis. A wildfire video from 2019 becomes footage from this morning, maybe only modified a little bit. A protest from one country becomes a protest in another country. Compare with credible news and local sources. When a major event happens, real journalists leave many footprints, articles, broadcasts. Eyewitness quotes time stamps. If something online seems extraordinary but isn't being reported by established outlets, that's a reason to be hesitant and find it extremely strange, I think. So we to look for multiple perspectives. Local news often has the most accurate details early on. If international headlines don't match, don't match what's circulating on social media that disrepancy might point to misinformation or at least exaggeration. Of course, I'm not necessarily telling you to choose one trusted source and sticking to it. But in these signs, you must be cross checking the strange things you read to not fall into the traps. Follow the context clues. Every genuine piece of media carries invisible signs and context clues. Listen for accent. Or background language in a video. Look for recognizable landmarks, weather patterns, or license plates. Compare them with reliable coverage from established outlets. If something doesn't line up, if the flood in Paris shows palm trees or the snowstorm in New York has green grass. Trust your eyes, context rarely lies. Verification groups like Bell and Cat, BBC Verify, and Reuters fact check often use this method, what they call open source intelligence to confirm whether an image truly belongs where it claims to. The same skills that uncover global misinformation can help you verify the authenticity of a single viral post. I truly believe the slower you go, the sharper you see. I truly believe it's true. I know fact checking can be boring. It's not fast. It's the opposite of the online tempo and super productivity we've been trained for. But slowing down, even just for a few seconds can save you sometimes, save your time, energy, and even safety and money. So before you believe that powerful video or emotional photo, many times, it's better to stop panicking or stop from reacting and ask, who made this? Who posted this? When was it made? Where else has it appeared? What might someone gain from believing it? 19. Real or AI? Can You Tell What’s Real Anymore?: Practice. Real or AI-generated. So can you tell what's real? Let's take a break from theory for a moment. You've learned about deepfakes, voice clones, synthetic photos, emotional manipulation, all the ways that digital world can bend reality right now. But this knowledge only becomes wisdom, and it only becomes powerful when you practice noticing more carefully. So in this chapter, I want to walk you through a few real world examples and invite you to analyze them slow down, look closer and notice what the little clues are telling you. And to do that, first, we will discuss some of the very loud, very popular examples because they are a little bit funny, but they are also very good for learning because I will tell you why they were created. So we will start with them. We will start with all these ones, the ones that you probably heard of or you probably seen them. But there's a reason they went viral and we need to discuss it. Okay, so let's start with the simple one. If you went online in 2023, you probably saw this one. P Francis looking unexpectedly stylish in a huge valentiaga style white puffer coat. Image spread everywhere, earning both laughter and confusion. Yeah. At first glance, it felt believable for some people. The lighting was soft and cinematic. The texture of the jacket looked flawless, and the poets expression seemed, I think, natural. But they were tiny tails. The hands look slightly warped. The shallows didn't quite match the fault of the fabric and the background was a little too crisp. Verdict, of course, you already know it's AI-generated. This image became a cultural milestone, not because of who was on it, but because of how easily people believed it. Yes, this one was funny, harmless, and a little bit absurd. And that's what made it work. It didn't look like a scum. It looked like an innocent photo taken to show people that Pope is a person like you and me, and you can also be a bit interested in fashion and appreciate good puffer coat. When I set of photos showing Donald Trump being tackled and arrested by police went viral, they sparked instant calls online. News outlets crumble to confirm whether it was real. Look closely, though, and you'd notice small inconstants Dissorted pols badges, a bit asymmetrical hands, strange facial blurs and lighting that didn't match across frames. The composition felt like a movie still. Too dramatic, too cinematic to be real. Verdict, of course, you already know AI-generated. It wasn't just a photo. It was kind of revolutionary storytelling. And that's what AI is learning to do frenly well, create virals that tap into our cultural imagination, not a way not a logic. A Ukrainian soldier holding a cat during what. This picture circulated on social media showing a young soldier crushed among rubble holding a tabby cat close to his chest. His expression equal pause exhaustion and tenderness captured hot worldwide. This one is trickier. The lighting is consistent, the shadows fall naturally the cat looks real. Nothing screams a ice straightaway, but zoom in. The uniform insignia and inconsistent. The fingers around the cat are slightly blurred and mismatched. You already know it's AI-generated. It's a perfect example of how AI-generated pictures photos now operates in the gray zone, not to deceive in an obvious way, but to evoke emotion. It plays to our empathy, using our kindness as the hook. This one looked scientific and objective, doesn't it? A color coded aerial view showing massive flooding after a storm. Tozens of accounts shed it as a proof of the events scale. At first, it seemed legit, data driven, neutral, but researchers later track the image back to a synthetic dataset created by an AI model trained to visualize hypothetical environmental disasters. The metadata showed no satellite ID or time stamp. So, this one is very important because it shows how misinformation doesn't always start with bad intention. Sometimes context just gets lost along the way, and it makes the difference a huge difference, a dangerous difference. AI create something educational, someone screenshots it, and by the time it hits social media, it has become breaking news. Netflix baby Render poster, featuring actor Richard Gatz with hunting eyes and well known spread widely online prompting spectaculation, that it looked too AI like. But in this case, it's Oh. 100% human made. A professional portrait shot by a photographer, edited with Sony tools, not AI tools. So the verdict is real. It's not AI-generated. What's fascinating is that our perception has flipped. We are now suspicious of reality itself. Authentic photos can look too perfect, while AI art can feel emotional and real. So as a result, we will have generations of viewers who doubt everything. Okay, we'll continue the exercise in one of the next chapters because I got the feedback that you really love doing it. So we will discuss many more examples, and I will zoom in on those examples to show you the details you have to analyze and you have to notice. More of this story. The next time something viral crosses your feet, a photo, a video, a screenshot. To for 5 seconds and ask yourself. What emotion is this trying to make me feel? Does the light, texture or background make sense? Would I bet money? This is real. Can I find where this photo or this video appears first? If the answers make you hesitate, dig deeper. Don't scroll past. Don't share it. Because spotting AI-generated content or fake content doesn't show our cynicism. It's necessary. To stay awake in a world that profits from our reactions. 20. Ethics, Safety & Future Trends. Ethical Use of AI Content: Ethics, safety, and future trends, ethical use of AI content. How to create share and innovate responsibly. The conversation around AI often swings between two extremes. Breathless excitement and being totally anti AI. But somewhere between those two extreme approaches lies the real work, learning how to use this technology responsibly because AI isn't going anywhere. It's already woven into our art, our writing, our marketing, our classrooms, our lives. When people ask what's allowed in AI, they usually mean legally, but legality. It's the floor, not the ceiling. What's permitted isn't always what's right. So legally, AI generated content lives in a murky middle ground. Copyright law, for example, is still catching up to a world where AI models can create. The model lens from millions of online photos, including copyrighted ones. Who owns the output? The user, the company that's trained the model. All the artists whose work the model studied. So far, courts and lawmakers are still debating that, but morality often answers faster than law. If you are using AI generated art, that clearly mimics another artist's style. Without permission, it may not be illegal yet, but I think it's still a little bit exploitative and it's not fair. So the ethical test is simple, I think. If a human made the original and they didn't consent to being imitated, you owe them credit, context, or compensation. For example, one of the most common prompts people use in M journey and Sora is created in the style of studio gibl. For example, people use prompts like this or this one. Is it illegal? General style, soft palettes, painterly big runs usually isn't protected by copyright. But specific characters, logos and distinctive designs are like, for example, in this case, Tortora, no face, the Gibbly low type. So if your image looks like it features recognizable Gibbi character, or a near copy of a signature scene prop, you are drifting into derivative work territory. Trademark, false endorsement. Using Gibb's name logo or look like characters to promote product or event can imply endorsement and may raise trademark or passing of problems in the future. Is it moral? The Gib style and feel comes from decades of human craft, layout, background painting, hand drawn motion, mimicking it to sell something without permission or credit may be legal in many contexts, but in many eyes, it's not fa. Because let's deconstruct the vibe, the ingredients we see in Gib films. Low child height camera angles, tactile food, steam shine and weight. Background with watercolor texture and atmospheric perspective. Ma, purposeful, quiet, and wind as a character, moving grass laundry hair. Corner stories, the saturated grains browse for nature, warm lantern lined at dusk. So it's very specific, very on brand. Gibley's work carries specific cultural values, quietness, respectful nature, gentleness. Copying the surface without honoring the spirit can cheapen the thing people love and spend so much time polishing. So these are red flags and don't do this. Slapping Gibb's character lie. B great forest spirit, a suit, spirit, no face clone on merge. Using studio Gibby style poster as marketing copy for a commercial product, making fake quotes or trailers that imply Gib made or endorse them. My tip, we can always use safer Pn instead of asking the model to generate something gibbly style. For example, we can ask to hand painted landscape with soft edges and atmospheric depth, a cozy hillside town and golden hour, focus on wind in trees and laundry lines, gentle realist, warm, lived in textures. Or Child eye view of a kitchen, big wooden table, steaming soup, sunlight dust moods, painterly brushwod, calm, domestic magic, no existing characters. Because AI makes it so easy to replicate voices, faces, and styles, but just because we can doesn't mean we should. Voice cloning someone without permission, even as a joke, crosses very serious personal boundary. Using a celebrity's likeness in an AI video without rights clearance might break intellectual property law. But even outside the legal context, it violates something deeper autonomy. So my proturb to feel good and write morally before using AI to create something that imitates someone's work, or isn't someone's style, even hypothetically, ask yourself, would they be comfortable with this? Would I be okay with this if the roles were reversed? Label what's AI generated. I think today transparency can become the new trust. If you create or share AI generated content, be clear about that. A simple caption, like a node at the end of your caption generated with majoni or created using AI does more than protect you legally. It helps keep the digital ecosystem more honest. This matters, especially for content that looks realistic and could be mistaken for documentation. A fake photo labeled as art is creative expression, but a fake photo labeled as evidence is manipulation. Unfortunately, not every brand audience, big or small, luxury or personal is comfortable with the use of eye, of course. For instance, I recently came across a premium fashion brand from my home country that started using eye to showcase their clothes and different style looks as inspiration. Their pieces aren't cheap, quite the opposite. And the brand positions itself as high end and its customers expect that level of authenticity and craft. But when the brand began posting AI generated images and videos of their collections, many clients were upset. They felt like the brand was trying to save money on professional photoshoots and real models, which made the brand feel less genuine, less lovable. On the top of that. The AI generated looks and clothes didn't look exactly like the real ones. The textures and details were off and it was almost impossible to tell how a piece would actually fit or drape on a real body. After all, an AI model doesn't understand fabric, thickness, movement, or how a Material behaves under light. So it's a good reminder that even if AI can make something look beautiful, authenticity and realness still matters, especially when people are paying for quality, craftsmanship and trust. 21. When AI Can Be Used for Good: AI can be used for good. But hey, I don't want to turn pessimistic. AI, when used ethically, has incredible potential. It can amplify creativity, reduce barriers, and help people express themselves in new ways. If you read live examples, of course, Accessibility. AI-generated voices help people who lost speech due to illness communicate again. You already know that. Education. Generative tools can visualize complex topics for students, making learning more inclusive. Restoration. AI can restore damaged audio and film footage that might otherwise be lost, mental health, AI journaling and emotional companion apps, when handled responsibly can support reflection and therapy. The key word is, of course, responsibility because for every positive use, there is a mirror image for harm. Accessibility versus impersonation, creativity versus plagiarism. Is a very, very complex topic, as you can see. Human touch. Every ethical framework for AI eventually comes down to this idea. There must always be a human in the process. AI can generate, assist, and suggest, but humans must review, approve and take responsibility for what gets released into the world. That's true for artists, teachers, journalists, marketers, and policymakers alike. When you use AE to create something, you are still the author. Which means you are also the one accountable for its impact. The question isn't just can AI do this, but should I do this with AI? If AI can generate a photo or even create a complex photo shoot and video, art, or even a whole movie. What's left for humans to do? I see it that way. AI can produce and generate, but only we can choose. We choose what deserves to exist, what aligns with our values, what s meaning to deny. It all lies in intention. So I believe every time we create something, we should ask ourselves, Is this helping or is it harming? Clarifying or maybe confusing, serving or exploiting. 22. Detection vs. Generation: Detection versus generation, the future of truth in a world where everything can be fake. Unfortunately, it is very much true. Every time you think you've learned how to spot fake, how to spot AI content, the technology levels up. AI generated content is improving fast, faster than our ability to detect it. Each new tool, each new model, new version of existing models that promise hyperrealism, better faces, cleaner voices, more fluid motion pushes through food into the gray zone. And for every advance in generation, another team somewhere is racing to build the even better one. Generative models like So apica, runway, and many more are already creating videos so realistic they could pass for al footage from a film studio in some cases. Meanwhile, photo generators, like M jenny and stable diffusion can render lighting, anatomy, and reflections with near photographic accuracy. So soon the imperfections we once relied on to spot fakes like strange shadows, strange texts will totally disappear. They happen more rarely right now. A I will learn to replicate physics, emotion, even camera lens artifacts. This means that detection tools can rely solely on surface clues more. They have to look deeper and deeper also into the data itself. But this data can also be modified, of course. As I've already told you before, one of the most promising solutions to the AI generated content and deep fake crisis is water marking. Embedding hidden signature into all AI generated content that proves its origin, like a DNA stand for media, invisible to the eye. But traceable with the right tools. Companies like Opening Eye, Google Deep Mind and Adobe are already experimenting with watermarking systems that label content at the point of creation. Adobe's content credentials program, for example, tags images with meta data showing who made them, when and which tools. But watermking only works if it's universal and respected. If half the Internet uses it and half doesn't, we will still be swimming in this uncertainty. Plus, even as tools advance one fact remains. Detection doesn't work without trust. A Watermg means nothing and people don't believe in the system behind it. A real AI flag is only useful if we trust the platform to apply it fairly. News outlets, governments, tech companies, and everyday users will have to agree that very fine content matters more than sharing it fast. And in this economy, it's very hard to believe that it's doable, right? But this is a very complex topic for a longer discussion as well. So, what the future might look like? Here's a glimpse of where things could go. Every digital media, for example, image, photo, video and audio file created by NII model. Carries a traceable signature by default. Social platforms automatically display AI generated labels or authenticity indicators. Instagram already does it, but it's not always correct. Many times it displays AI generated content next to the content that wasn't AI generated. And many, many creators struggle with that. Detection algorithms run quietly in the background, flagging manipulative or misleading content before it trends and goes viral. And also YouTube already does that. Governments introduce deep fake transparency laws requiring public disclosures in us media and political content. But even in that optimistic scenario, AI generated content and fake content will never disappear because technology will keep evolving. There always be someone trying to stay one step ahead. And that's exactly why we need to constantly educate ourselves and keep up with the technology because the more people know how this technology works, the less power it has to deceive. And to be honest, I know, I know it can get overwhelming. I'm with you and I'm here for you. 23. Spot the Difference! AI or Real? Baked Goods Edition: Positive difference. AI or real baked Goods edition. Are you ready to test your eyes again? Now we are starting a series that will definitely make those of you who are hungry a little bit more hungry or even a lot. I recommend not watching this chapter completely on an empty stomach because it can be dangerous. And I think baked goods are a difficult topic because there won't be many captions. There won't be lot of details like in a grocery shopping edition, we really need to be very careful and we need attention to detail. So, shall we go? Let's begin. So two cakes. At first glance, both are very candle like, very overly sweet. But let's look at the details. Here, what immediately catches my eye is the texture. It seems like you see it has those air pockets that, on the one hand, look natural. I feel like I've eaten cake with a similar texture before. There is definitely something strange on the fork here. Let's look at the tablecloth details because that might give something away. It's okay. Okay, what else? What else is worth looking at here the confetti. You see something strange is happening with some of the shapes. Let's see how it looks on the one on the right, and here the confetti doesn't change its shape that much. But here, some of them look kind of bent. Okay, let's take another look at these details. And now I think this tells me the most. The ice cream texture. You see here the texture is very, very strong. And here it's similar, but even though they're separated by very little in terms of depth of field, here the texture is still different. So, in my opinion, this will be AI-generated because the ice cream cone has different textures, and I assume they come from the same package from the same manufacturer, so they should be the same. Here we have text, of course, unreadable. This is a photo with a very, very shallow depth of field. And let's take a look at the pistachio. Let's analyze the details. But what immediately stands out to me, especially in this photo, is this plastic like form. Do you see it as if the pastry was made of plastic. Totally like it's plastic. Plus, it has suspiciously many layers. I know this type of cake has many layers, but here, there are unrealistically many of them. And I think this one is hotter because there are far fewer details. Here we have two textures, actually, even free because here the cake is on decorative ray. And here, there are also really free textures. But this angle is very strange. Do you see it? I'm trying to imagine how you'd place it, how you'd achieve this angle. But then everything else wouldn't be at the same angle. There is something very unnatural about this angle, yeah. So let's take one more look. At what lighting was used and move on. Okay, let's take one more look at what lighting was used and move on. Okay. Here we have two gingerbread houses, and of course, we are looking as always for small details and small inconsistencies. Here you can already see a very big difference in thickness. Plus, I think it would be very hard to make something like this with icing, especially looking at the lower layers. Plus these figurines here, do you see it? They kind of have a very undefined, very, yeah, very blurry undefined shape. These are fine, but then when we move to the right, something strange starts happening while here, everything has a defined realistic shape. Same with the ornaments. Do you see it? As if they were poorly generated and the shapes got lost. This is definitely generated. Okay, this one is harder because at first glance, I would say both are achievable in real life that both photos look real. The devo field here is very, very realistic. I'm wondering what this is, but it could just as well be soap or some kitchen tool. However, the lighting really makes me wonder because here the light is very harsh. And the cupcake doesn't show such strong reflections. Plus, here I see inconsistencies on the napkin. And as we know, those very rarely appear in reality in real photos. Here the texture is uniform. So for me, that one is AI-generated. This one is difficult because one of the photos is really poor quality. As you can see, the one on the right. And these photos are so stripped of context that both could have been AI-generated. So here, I'd honestly prefer to skip this because this example won't tell you anything. The only thing we can look at is the regularity of the confetti, but really the only strange thing is this. I think that in product photography, there wouldn't be such clumps of cake and confetti, so I'm betting on that. But like I said, this is not a very good example. So let's move on to the next one. Let's move on to this appetizing photo. I probably said on another occasion that AI very often struggles with corners in photos are with details, specifics and contexts in corners. And here we have a price or product number. Plus, we also have a tool for filling them with cream. I think that without a prompt, explicitly mentioning it, this tool wouldn't be here. And if there was such a prompt, it would probably appear more in the foreground. So this makes me think it's a real photo. Let's also look at how they're filled, which ones are filled in a way that would definitely sell to a customer. Here, everything seems fine, even on the ones further away. And here it's done a bit more carelessly, plus they vary greatly in size. They see it for example, this one is much smaller than this one, than that one and this as well. So as you can see, the ones here in this area are very tightly packed. I think this wouldn't really be sellable. So, in my opinion, this is a high generated and this blur here as well. Yeah, I think there are many signs pointing this to being a high generated. Let's move on. Here again, we have the same situation I mentioned earlier, product corners which AI often struggles with. Here, it's hard to tell what exactly this is. Although the cake with cherries look fine. And in this one, we have some kind of cheesecake, I think, although, wait, does it really look fine? Because now I'm starting to wonder if this is actually an artifact and mistak in AI-generated photo, especially when we look at this reflection. And the reflection here, do you see it? On a photo on the right, the reflection is realistic. While here, the reflection is in completely wrong place, at a very wrong angle. So my initial confidence that everything was fine with these objects lowered my guard. This reflection is totally unrealistic and just as I thought. Yeah, just as I thought. Okay, this is tricky again because there are very few details here. I generally don't like product photos. I find them very hard to analyze. There are a few details here. I'd say this croisson has a very strange shape that in product photography, they wouldn't choose such an irregular croisson to present. To present. I saw it pronouncing everything with French accent. No, I don't know how to get rid of it. Okay, let's take a closer look here. I think the texture gives us the most clues here which texture could be real. Here, one thing immediately stands out. This one is defective, and definitely no one would sell such a combination. Plus, the pattern is also reflected a bit incorrectly here. While all of these are really elegant. So for me, there is absolutely no doubt that this one is a i generators. Yeah. Let's keep going. When I look at the one on the left, the first thing that immediately catches my eye is that you see the price tock, and here the names are printed at different heights when you look at it from the site like this. And that kind of makes me think it might be AI-generated. On the other hand, the details of these canals are beautiful and everything else makes sense. But here I'm surprised that on display, these cookies are so irregular, as you can see. Take a look here. I don't want to zoom in, but look, they are very irregular. Plus, this one, for example, looks slightly like it's been bitten into these habitat, too. One is shaped like some kind of ghost and the other is normal. Plus, I have no idea what this is in the background. It looks very messy, especially compared to the one on the left. So, yeah, in my opinion, this is generated. Yeah. Okay, now we are taking a closer look because for sure, the details will decide our choice again. The one on the left seems fine. Everything here is regular. It could be a photo promoting some cafe, I think. But what immediately caught my eye is that this one here has a completely different shape. Is much taller and has much more feeling. I know such irregularities can happen, especially when someone isn't making this in a cafe, but, for example, for personal use. But it seems to me that this reveals a lot. Plus, the cream has a slightly different color, even though it's the same flavor as the other as this one here. So, yeah. For me, it's AI-generated. Okay, here we definitely need to look at the details, and there are a lot of them, but earlier, I was complaining that there were no details, so it's nice that here they are. First glance, everything is symmetrical here. When we look at these blobs of cream here, as you can see, yeah, everything is fine. And here I have no idea. Okay, now I see now I see one thing that is strange. It could happen, but I think it's strange to me. The one with cucumber with such a slightly misshapen cucumber. Also, it's a very strange choice to put cucumber there. Let's also look at the berries. These two are a bit suspicious in shape, I think. Just like the perspective of this donut here, but let's take another look. There are a lot of details here, and in my opinion, AI would definitely mess up in at least one detail here. So that makes me think this photo is real, and this one is AI-generated. Okay, this photo definitely looks very Scandinavian. Let's take a look at the texture of these baked goods. This texture worries me a bit. It kind of looks like a i. And on the other hand, what stands out to me here is that this part, this part here looks unfinished. And here it looks really strange, like Ai already took the easy way out. These two are definitely supposed to be the same as this one. The first cinnamon roll also has a very strange shape without its typical spiral. So I still bet this one is AI-generated. Okay. Okay, let's look at the shapes of these bagels because shapes will tell us the most, I think. Here they are very irregular. Such irregular shapes usually don't happen with this kind of bagels. Okay, they are not bagels. They are pretzels. I don't eat them. Very often say, Yeah, my bat, forgive me. But look, those are very, very irregular. And I think such different shapes usually don't happen with pretzels. Here, on the other hand, everything is very even. And that makes you think this one is definitely AI-generated. Okay. Okay, on to the next. Let's take a look at the nuts. Everything seems okay for me here. But let's also check the texture of the plate. It's also okay. But what immediately catches my eye is this very misshapenns here or the fact that this should be sharp in this place because the one next to it is sharper. And I also don't understand why this one is overexposed. Yeah, for me, this is a very clear sign that this has to be AI-generated. Okay, I have to admit this one is difficult, but I think I would actually say that both of these both of those photos could be easily AI-generated. There are photos where it's very hard to point to one specific thing. This could be lit with some kind of studio lite, for example. And when I look at it, this one seems more ish to me, but as you can see, it's really very hard to tell. And like I said, book could be AI-generated or book could be real but very, very edited. Okay, let's check. Okay. Okay. Okay. Here we have a very good looking, I think, Austrian straddle. I know such vintage spoons exist and have shapes like this. But you wouldn't be able to eat this stroddle with that spoon. So I'm betting that this one is generated. Okay. Onto the next. Here, let's look at the texture of the confetti. Okay, here I think there is nothing very alluring yet. What is definitely alarming in this photo is this gingerbread cookie with icing here, which is much thinner, much thinner than the others because I think such a thin gingerbread would be impossible to make. Correct me if I'm wrong, but this is a very thin one. You can see that it's much thinner than this here. Plus, this tool is also very strangely benched. And here we see kind of very strange thickening. Plus underneath it, it looks like broken glass. Wow. Yeah. Yeah, for me, this is AI-generated, for sure. I don't believe someone would intentionally create such an illusion of broken glass. This is very unsettling. Yeah. Okay, and the last photo in this round, unfortunately, again, from the category I don't Like as you already know, I think my first instinct was that this one has a very uneven texture. You can see how much is going on here. I think it's very hard to achieve such an effect naturally. Okay. And yet, no. So that was my first mistake then in this round. So let's take another look at it. As you can see, because of the small number of details, this was very, very difficult. That's why I prefer the ones where there are much more details and it's easier to distinguish. And what do you think about this round? Was it difficult for you to distinguish which photo is AI-generated? Let me know in the comment section in the discussion section because I am really curious. Let me know. 24. A Visual Test: Red Carpet Events: Visual test from the red carpet events. Okay, the next chapter is red carpet events. There will definitely be a lot of details here. A lot of cameras, a lot of elements that will test our perception. So I think this is an exceptionally cool main theme to go through together. Let's see how sneaky these pairings will be. Are you ready? Let's go. Okay, just like I thought. There are a lot of details, but also only one photo has text. I always prefer will obviously with text and a large amount of text and details because it's easier to judge. The less there is, the harder it gets. So let's take a look. Let's zoom in on this photo. Here the text, I think there is nothing suspicious in it. Nothing is blending together. No leather looks like it's titled. Now let's look at the geometry here of the carpet and the gates. You can see it's a photo taken with flash. Let's look at the second one because I think it will be able to tell us much more. Here I can already see the signs, the little signs, the little artifacts that shouldn't be there. Like, for example, here, something is kind of badly generated. Of course, we can point a lot to the lower resolution of the photo. But when you look, for example, at these faces and this arm, you can see that something here is not generated correctly. Same here. This nose, this hand So at this point, there's not much more we can say because there are too few details, but we can also see that there are no real leathers there. There are just kind of placeholder blobs that AI often froze into the background, so it doesn't have to refine the details. So right away, we have proof that this photo looks good here at this size. But when we zoom in, we can see a lot of badly generated things, a lot of artifacts. So we can say this with full certainty, this is AI generated. Onto the next. So here at first glance, both photos look like they were generated by AI. In my humble pinon, there are a lot of red flags in both. So let's take a closer look. Flawless skin and its texture overall, but this could be a clue or it might not be because as you probably know in this part of the world, people really do have very good genes and beautiful skin, so this could be completely natural. Fingers with children, sometimes proportions can look different, of course. So the fact that this thump here looks like this could also just be a matter of perspective, nothing conclusive yet. The eyes look natural. The details as well. This could be a well generated AI photo, but it could also be completely natural and completely real. Now let's look at this one. Right away, I noticed some inconsistencies related to the shoes. Do you see that? As if they either have a very strange shape or are badly generated? And for me, although now that I'm looking at the sole, one shoe has a completely thin sole here at the back, and the other has a high one, and they are very asymmetrical. Same here with the pens and here the endings, the ending of this cape. And when we look at these shoe laces, you can also see they are badly generated. AI took the easy way out, and I can say this with full certainty that, yeah, this is AI generated. From the faces, it's not that obvious. Although when we look at the eyes here, we can say a bit more. But in this photo, I think it's the shoes the shoes give away the most. So yeah, here we have two actresses, which one of them is real? Let's look at this one. It's always, I know, I am repeating it like a broken record, but it's always worth comparing logos and text because when there are multiple instances of the same text, AI often has trouble reproducing them. Mm hmm. Let's also look at the hand. The hand looks natural, I think. AI also often struggles with intensely repeating patterns. Here, this bear. This bear, as you can see, is slightly different in size, but the text, for example, this E is finished in exactly the same way. So that looks real to me. If this was AI, I think there would be a few more clues, a few more wrongly generated details. So I'm betting this one is real. But now let's look at this. And the foreground theoretically, everything is okay, although here I have the impression that something went wrong with the eye, but sometimes at this resolution, that can also happen in real photos, so this is not decisive yet. But look here, something clearly went wrong with Face generation. I think AI assumed we wouldn't look that closely that at this size, it looks okay, but look, here, you can clearly tell that AI took the easy way out, yeah. So even though the foreground might look partly believable, could be believable, the background definitely gives away a Eye. Okay. Yeah. Let's move on to the next. These are harder at first glance. Hands, look at hands. Hard to say. This hand belongs to this woman, and here only the hands this handshake belong to our main foreground characters. This photo doesn't have great resolution, so we can't say very much about it. But the details still seem believable. The depth of field, considering this is a flash photo, as well. Okay, let's look at the second one. Here, at first glance, this hand looks very Ish, like it was generated to me and this finger yeah, but let's look further. Here, the eyes definitely has the AIS vibe. Yeah, I have no doubts here, I think. And look at this hand. I have no doubt. From a distance, it looks good. It doesn't raise suspicion right away. But when zoomed in, you can see large red locks and wrongly generated details, and we can be 100% sure, I think. Okay, here a lot is going on. So what should we look at first? I always like to look at logos, as you already know, and the sponsors, whether they are generated correctly. And actually, at first glance, without meaning to, I already landed with my click on an AI generated photo because Look at these faces and look at these poorly generated sponsor names. This is the same logo as here. Well, it's not the same. You can see how inaccurately it's generated. When the photo is small, it looks okay. It can lure a careful eye, but it's enough to zoom in to realize that everything is wrongly generated and that there are really a lot of badly generated details, a lot of artifacts. So here I don't even need to look at the second photo, maybe just a quick look. Here every person is real. We can already tell that, and there are many more details. So this one is definitely AI generated. Onto the next. Okay, here we have two actresses waving at us. One photo is blurry. So right away, we know it will be harder to judge here. This hand is also in motion, which normally could be a clue, but unfortunately, the quality of the photo leaves a lot to be desired. There is a very strong blur. We are not able to assess it well. But the bracelet looks real and detailed. Here the earrings, the hand looks okay. This area here makes me wonder a bit, but it could absolutely be real. So let's take a look at the second photo. Okay, here I have the impression that something is wrong with this camera and same with this one. Do you see it as if the cameras were blending together. So even though our actress looks believable, the finger also gives away a lot. And same as now. I see it here the shoes. Do you see that? You can clearly see red flags here. And that finger is still in the frame. Yeah, this will be generated. Okay, two beautiful Asian women, and only one of them is real. Let's look at the one on the right. Of course, everything is blurred. Here we see some text, but we are not able to say much about it. So, unfortunately, this part it's too blurred to judge. Here we can wonder why this finger is darker. This nail is darker. The clutch, this hand on this hand, we don't see this dark nail polish, so that might be the first red flag. But the face looks believable. The endings of the earrings are the same. Here, the top also looks believable. Okay, let's look at the second one. Maybe there will be more pines to latch onto. Okay. And right away, we see that even though the photo looks amazing, like a vogue photo, something is wrong with these glasses as if they were blending together. Do you see that? Plus this ring? Is it onto fingers? Or is it also catching onto this one? Something is wrong here, just like with this back. Like it's a badly generated channel back. Yeah, I think this will definitely be AI generated, even though the rest, for example, from this perspective, looks very real. But you can also tell by the earrings. Look, do you see it? There are no identical earrings. This is a very, very common AI mistake. Okay, let's move on onto the next. Here, we have two very characteristic celebrities here. Let's start with the one on the right. And here I instantly see that the text is very well generated, not generated. It just looks very good, very believable, like it's real. So it might be real. For example, here, it was probably Amazon. And here you can see the are the same. So even though the face looks so perfect that it could easily be accused of being AI generated, the text of sponsor is super precise. Do you see that? They are aligned under each other. So this looks like a real photo to me, definitely. Okay, let's look at this one. And right away, this lanyard raises my doubts because who would go to an event with so many people with such a worn out lanyard, one that's about to break. Here, the endings of the hands also raise my suspicions a bit. Let's also look at the lapels. You see here there is a seam, and here there is no seam. Same with this button. The fabric is finished very strangely. But the face doesn't raise suspicion. The face looks okay. But look here, here the faces of these people also look kind of badly generated. Maybe you can see it here in this area, but, for example, here and here is very visible. So, for me, the verdict is already clear. This one is AI generated. Okay, again, we have two actresses. One is smiling with a very, very Hollywood smile. The other went for a mysterious, more shy smile. So what can we say here? Which photo at first glance looks more AI generated? I'd say this one, but let's check whether everything here is actually okay. This camera looks a bit strange. I have to admit it just hanging there. Someone could just as well have it on a strap, but it looks weird. That raises my suspicions a bit. The rest seems okay, but this photographer here looks suspicious to me. The face looks believable. We don't see the second earring, so we can't say much here. But let's look at this one. Here I immediately see that a lot is wrong with the faces. This see this, this and this, and all that. Here it is like the face is in the wrong place. Here, the hands also blend into the camera. This face, definitely based on this, this suit is also undergenerated, and same as this one. From a distance, these photos can lull away careful eyes. But unfortunately, this is clearly eye generated. On to the next. Okay, here we have what I like the most because it says a lot, the license plate. Let's look at it. It looks very believable. Check. The current logo is also well generated. The people inside also look well generated. I mean, not generated. This is probably just a real photo. The logos on the balloons also look the same everywhere. We don't have any red flags on the faces. So let's look at the second photo. Of course, they are out of focus, so we can't say much. Here I can already see that something with the camera is deadly generated. This person here looks more like they are holding a weapon than a camera. So let's look at the details. Here, it also doesn't seem like these are real faces, these shapes, these fingers. Fingers and hands tell us a lot here, and I think the verdict is already clear. Okay, let's start with this one. It looks a bit like a selfie, but it's not a selfie. The people in the background seem real, especially this screen, as well. I'd say there's nothing off so far. The sign is also fine. All the palm trees have regular leaves, the hands look natural. The shoes have details. I'd say this seems real to me. So let's look at this one. Here we have some kind of brain. In the frame. And here I already see a badly generated face or a motion blurred face. Let's look at it more closely. Oh, wow, here are the structure as well. It's very strange. Do you see it. From a distance, it looks good, but they wouldn't hold any Gala with such a structure because everything could collapse. Yeah. When we look at the detail, so we see a lot of very inaccurately generated things. You can see AI took the easy way out, so here the case is easy and the verdict is already clear. Okay, let's move on. At first glance, of course, they seem more AI generated, but you know these photos can also be very sneaky. So let's look at the details and at the hands. Yes, I think these hands are AI generated. But on the other hand, the shoes don't look too bad, although I'm wondering what fell of them here. And, yes. Yeah, by this face and by this camera, I can already tell this is AI generated, unfortunately. Same here. By this wrongly generated detail, this artifact, by the inaccurate generation of these faces, do you see how these faces blend together here, how they are very badly generated? Here at first glance, what made me wonder was that the cannon logo is reversed, but I no longer remember how it is on mine, whether it's like that, too. I thought it might be an AI RO, but here everything is actually generated correctly. I mean, not generated. This is a real photo. Let's check that. Okay, on to the next. So let's start with the photo that has text. So with this one, it looks a bit suspicious, but is it very suspicious? Is this a sufficient clue? I think that here, the text itself is really suspiciously bad, bad, bad, really bad and really inaccurate. I also think that this face here is very telling because black eyes, black lips, I don't think someone looking like that would be allowed into this part of the event. You can also see that the face you see here, this silhouette. A lot of details are missing here. I think I'm already 100% sure that this is AI, and, yeah, the lies are telling as well. So the photo at this size looks completely real. But when we zoom into the details, we instantly see badly generated elements without details. So this photo is real. Although at first glance, we might also have some doubts, but here everything checks out. So the verdict is clear. Okay, again, we have a lot of cameramen and photographers. Let's take a closer look at them. Here the text raises my doubts, just like this leg here. I think I started. Okay, these legs. I think they are completely wrongly generated. As you can see the details, here are very everything blends together just like these hands. No way. No way this is real. Same here. Yeah. Let's also look at this photo, but I'm already practically sure that one, the first one I say I generated. This photo is motion blurred. But when we look at the details, the camera looks good. Nothing is blending here either. In the background, we can see different brands. Yes, I'm 100% sure, especially later if you go through this category yourself, remember to look at shoes, legs, whether, for example, such a skin texture or such a gesture is actually possible. Okay. That's the verdict. Okay, here we have to divs. Let's start with the one on the left. Here the text is a bit hard to see. The fingers also give me small doubts, but they could just be very slender fingers. The face looks believable. These earrings here seem a bit different, but that could just be as well, a reflection. Let's look at the people. Unfortunately, we are not able to assess the shoes because they are very, very blurred. But the handbag. Oh, I accidentally marked it. Oh, you see, I rushed too much, and I didn't want to mark this at all, because to me, this also looks real. So don't mind it. Let's take another look to discuss. Yeah, we need to discuss it. Now that we accidentally already know that this is indeed real, I'd still look at the handbag details. And now let's think about why this is a. Do you see these hands here we would actually instantly look and already know that the nails is generated, that the nail is generated in the wrong place. And what about the people? Okay, there are red flags, too. For example, this man is generated well, but here something seems off. And here the lens protection wasn't generated. So that would be also my verdict. This is definitely a i. So now on to the next one. Okay, let's look at the people. They have unusual expressions. Cameras are different. But here we can see this is probably cannon or Nikon. Here, these details are visible here as well. The shoes have details. I already think this will be this is a real photo. But let's look at the second one and zoom in on the people. Do you see this backpack? No way. Even if it were some very, very original design. And these shoes, one shoe is much thicker than the other. Here there is also a missing piece of face. Here is something happened to the glasses. Here, what kind of lens is this? Who is holding it? Where did this lens come from? This huge hand here. These fingers definitely generated. Okay, now let's start with analyzing the one on the left. I instantly see that the text is consistent. Can you see film festival? Here something is motion blurred. That could be a eyes do in, but these seem identical. Can you see? All of them except this one, actually. Sign here a bit hard to read. Here I have dubs. But the number is good. No concerns. The rim also looks real. Let's also look at the camera. The fingers seem okay. Let's look at the second one. Okay, I think I accidentally zoomed in on something that tells us a lot. This hand and the logo, which has different colors and shapes from every angle. Plus, here, this shadow also seems to have the wrong shape, and with this hand, I think we already saw. Besides these shoes, the longer we look at it, the more red flags there are, even though at this size, it looks good, right? Okay. Here I instantly see the consistency of Audi and Audi Audi logo. In the background, we can take a look. It appears everywhere, and it is the same. Here we also have the name of the same hotel in the background here as well. The details are blurry, but you can see that, for example, the sign here is fine, looks very real. So yes, this photo looks like a real photo to me. But let's look at the second one. There are very few details here, but this photo also looks, this photo looks very like ish. You see, these spikes have different shapes, and I think that never happens. Same here. There is a different length of this space and the spike. The shoes also look badly generated. Here it's the same. Yeah, I think at this point, we can be certain that this one is AI generated. Okay. The second to last one in this category in this round, let's start here with the smiling actor. Let's start by analyzing what I love analyzing the most the shoes. Here, the shoes seem to have all the details and be the same of equal thickness. Here we also have a lot of details. Here as well, these handshakes seem real, I think. Here in general, yes, there are a lot of details. The roses also look real. So let's look at the second photo. We are looking for something suspicious here. Yes, I think this pace is already giving AI away here. And as you can see here, there are also wrong distances. At first, AI keeps this distance, and then there is no proper distance between these gates. Here, we also see red flags, especially on these shoes. And unfortunately, you can also see it on this hand on this wrist. So I think we already have the verdict. Unfortunately, this photo is AI generated. Okay, and the last one in this series in this round, let's see if it is the hardest. Here we have very few details. So I would start with that. Okay. Here, on the contrary, there are a lot of details, a lot of cameras, faces. Even though they are out of focus, we can still see some details. For example, glasses. And also, there are notes we could look at, but even the microphones keep their shape. Same with the phone, have, this will be a real photo. I can already feel it. And here we need to take a look. Okay, it was enough to just zoom in to see shoe details and shoe mistakes, shoe artifacts. Shoes are part of the styling that has a lot of details, a lot of shapes, and they are also hard to reproduce realistically in motion, so they can tell us a lot here. And because of that, I think we can be, we can be really sure. This is AI generated. Okay. We made it. We made it. And another round is over. I'm really curious which set of photos was the hardest, the most difficult one for you. Let me know in the discussion section. I really, really, I need to know as always. And are you ready for the next round? Let's go through it together. 25. Practice: AI or Real? Tattoos Edition: A real tattoos edition. Okay, now I'd like to go through the tattoos round together because tattoos have a lot of details. What matter here is the linework, the background. Yeah, all the details. So I think this is the perfect perfect theme to go through together. So let's see how difficult this round will be. Are you ready? Let's go. So let's start with the left photo. At first glance, the hands seem fine here. Nothing on the nails raises my suspicion. The skin texture, the earrings, everything is okay. Only the texture of this ring makes me wonder a bit why it has so few details, but that's too little to state anything for now. So let's look at the second photo. And here, unfortunately, you can already see this hand. This finger goes in and suddenly ends here. Yeah, unfortunately, definitely, this is AI generated. Besides that, compare the width of these tumbs here. And these rings you see that. The details the details as always, as usual. And as this size, it could pass. But once you zoom in, you can see it black on white. Okay, let's look at this one. A very unusual west design. However, I see handwritten style font leathering here. And as we know, AI doesn't really handle those very well. This watch looks a bit suspicious, but maybe I'm just being super suspicious already. Still, the tattoo itself, its details, its form, the irregularity of the letters make me almost certain this is a real photo. So let's look at the second one. Here, I immediately wonder what this fragment of the orange vest is because it doesn't make sense for me. I'm also wondering where this irregularity of the lens comes from. You can see here that on every floor, they look different, on every floor, they look different. But let's get back to the tattoo. The blouse looks like it's worn back to front. Something is also wrong with this button. Those kinds of details unfortunately very often reveal a i. Yes, something is definitely wrong with this button. And I think that's enough. Here the texture of the blouse also doesn't look entirely natural, entirely real. I think whereas these railings, we can be almost sure this is AI generated already. Although I have to admit it's quite good quality here, but, yeah, I'm certain it's time to it's time for the verdict. Okay, let's start with the one on the left. At first glance, the arms muscles seem fairly symmetrical. The rose also has proper details, I think. You can see the fingers, so that makes it a bit harder to judge it. But these buttons seem fine, the shot texture as well. So let's look at the second one. What doesn't sit right with me in this photo is that when someone has pale skin, I think they have pale skin everywhere on the hands and then on the arms. But here, the skin is much darker than this. Or rather these details. Here the sleeve is directly on the skin. There are no lighter pigments introduced. And here as well, these details. Can you see it? As if something here wasn't generated quite right. And same here. This cuff is much larger than this one. I think I'm almost sure this will be a generated. Oh, I already clicked accidentally. And, yeah, I accidentally clicked somewhere in the meantime, but this is AI, so we will write. So let's look at this one now. The fingers instantly stand out. Unfortunately, even though the surroundings are fine, you see this thumb here, the endings of these fingers and also the way this t shirt sits on the shoulder here could be a red flag. Unfortunately, based on these facts and here as well, these holes in the shirt, these are details that indicate this photo is generated. But let's take a look at this one as well. Here the hands and the details. We are even able to read them. So yeah, unfortunately, this photo these hands, these strange calls details on the shirt clearly tell us this photo is AI generated. Okay, here, at first glance, I'd say this one. But because these stretch lines on the hand really surprise me. But let's look at this one. Here again, the ring is in a very strange place. And if this were a tattoo, it's too saturated compared to this. This one, I have to admit is more tricky. This thump also has a bit of a strange shape. I even have to look at my own thumb. So let's look again. Okay. The ring is in the right place. You can also see hairs, which a guy sometimes struggles with. What surprises me is the difference that here they are the hair is so dark and here they are already gray. But yeah, this is tricky, I have to admit, because I bet that this is AI generated. As you can see, Okay. Okay, that was the example where something is wrong in both autos, and I say, both are hi generated, and yet only one is. So it's simply a matter of a very specific photo and very specific lighting and details. And that's why the shadows fall the way they do. Let's look at these details. Here we have a lot of details, so there will be a lot to analyze. Is something here badly generated? I might admit it's strange that this is out of focus here. This ship's bow here also surprises me a bit. Let's look at the second one. This one has kind of more details. It also looks a bit more realistic to me. Here we can also see nice shading. Okay, I think this is an AI artifact. Here the details all seem to connect with each other. There is a dragon here, then this flower, the curve itself, the light. It could just be a very heavily edited photo. What makes me wonder is this chair, these metal metal elements. But let's look at the second one. The background is also very strange, but here it seems to me that this is generated well, but this here. And this is, in my opinion, a typical AI artifact, AI wrongly generated this. Also, this white shading. Yes, I think so. In this photo, there are lots of handwritten details, handwritten text. They are not complete, so we are not able to judge whether they make sense. At first glance, let's look at this cross. Well, here, the details of the cross are a bit misleading. As you can see, if it's barely holding together. It's very suspicious, but let's also look at the second one. This tattoo very strangely kind of goes into the bird. Here, there are also a lot of regularly generated elements. Also, quite a strange bird shadow, which is larger than this one bird itself here. Okay, so despite this raising my suspicions, the handwritten text, the handwritten font, and apart from this cross, nothing raises that many red flags. I said, This is ai tolerated. Okay, let's start with the fingers because they might tell us something. Does anything strongly raise suspicion at first glance, surprisingly, no. So let's look at the second one. I think this yellow element of the shirt here is very strange. Also very strange shading. You can see as if there was blood on this tattoo. And this red here looks very disturbing, especially this. After all, such elements, I think, don't appear on tattoos. So for me, this is already a sign that this might be AI generated. Let's see the verdict. Okay. Oh, surprisingly, no. Now with this one, all the details, all the details. This fumb looks suspicious. So let's look at the shirt. Here the ending of the shirt and this fold look very AI like, very Aish. But let's look at the second one. Here the details are generated well. All the tattoo elements make sense. So here, both this lump and the shirt fastening and this strange fold, yes. Yes, we knew it. We knew it. Okay, here we immediately see that this calf is bend in a very unnatural way. Let's take a look at this. Do you see this as if the shoe is going into the skin? As a sketch, it's also unnaturally saturated. And you can also see creases. I'd say this has to be a eye, but let's look at this one as well. Here you can immediately see that this is a pen sketch. You can also see smart degrease. It's quite a specific tattoo, but maybe someone has a very abstract idea. But it's true that this color this green is strange here. But still, the shape of this calf and how strange the saturation of this sketch is here. Okay, okay. Yes. Yeah. Let's take a look at this hand. Something catches our eye here. There are a lot of details. Everything seems very precisely done. I can't deny. Here as well, we don't have anything that would raise doubts, I think, but let's look at this hand. Okay, we can say the same about this hand. Unfortunately, the ring seems much too tight, and the fingers have a slightly suspicious shape, especially this one. Here there are kind of unfinished roses, and as we know, this doesn't happen with such a fresh henna henna tattoo. So the case is clear. Okay, here I think the nails will tell us a lot. Let's check if we see any suspicious details on them. Maybe on this. Here the shape stands out. This ring also the ring the decoration very suspicious. Here we also see these endings. No, no, no such nails. Here as well. The ring is squeezing the finger. Yes, I immediately see this. This has to be a eye. Yes, at first glance, it seems to me this has to be a eye, but let's take a look. This tunnel, you see there is no way someone would actually have such an irregular hole in their ear. The shape of the hand, this crease here, also looks very suspicious. This blotch on the tattoo makes the tattoo kind of disappear. The text as well, you see this is and then suddenly the text disappears. It seems to make no sense. This tattoo is very stretched on the neck. Let's look at this one. Here, however, the details seem to make much more sense. So I'm betting on this one. Yeah. Okay. Surprisingly no. Let's take a look, then. Maybe, yeah, this photo is no longer available from the author. But let's look at what here could have. Well, this hand, indeed. This hand can tell us a lot here. You see, even when we are super careful, super observant, super focused, we are still able to miss things. Yeah, and probably this head, as well. Even when we are super careful, we are still able to miss some red locks. That's the lesson of this this set. Let's take a look at these details here. Very saturated colors. Let's look at the second one. Here, the hair is definitely a plus because AI usually struggles with it. Here, however, we also have very inconsistent linework. Do you see that? In a real tattoo, there wouldn't be such unfinished lines when it's fresh. Here, on the other hand, everything is much more tidied up. Okay. Intuition was guiding as well. Here, I would immediately say this will be AI because we see a very big inconsistency in skin color. But let's look for more evidence. On the other hand, the details at the edges suggest this might actually be a real photo because a ye usually struggles with edge fragments, but let's look at this. The fish is very stretched. Tatis, as you can see, are very tricky. Still, the skin shading is somehow real and natural. Okay, so let's see what prompt was used. Koi fish. Yeah. Okay. So, as I've already told you, even when we are super careful, we can still miss some big red flags, and the more we practice, the more observant and more careful we get. But there is no way you can catch all the photos that AI generated because they are so good. Onto the next. So here at first glance, I think we will all agree this looks like AI. Will the ring tell us anything? There is too much skin texture here. It looks unnatural. But let's look at the second hand. Okay, here on the other hand, the nails are extremely unrealistic, I think. Do you see how this nail is peeling off? So even though the skin texture, he raises suspicions, these nails with an absolutely huge clue, huge red flag. Okay, here at first glance, hard to say, we need to look closer. Here we see play rewind ons. There are a lot of details. There are also cardinal directions here, but let's look at the second photo. Here these hands seem to have a slightly oddly bent thumb. Here it looks much more realistic, although on both photos, the blur is still very strong, but I'm betting on this one. Yes. I would immediately say that this text, look, here's someone put in the prompt what should be written. But here, AI already generated it the wrong way, just like it also failed to place the same text higher up. So yes, this is AI. Here, at first glance, both seem real, although this dot really makes me wonder because the Ms are also the letter s are also very different, and yet it looks here the k is also written a little bit strangely. But let's look at other details. We have a mole. But I think this nose tells us a lot here. It doesn't seem real. It doesn't seem natural. Here on the other hand, this one, this one less saturated element as if older. And these genes, honestly, this is a difficult example, but I'm still betting on this one. Okay, and that's how it is. So this nose gave us a good hint. And this is how we finish these rounds, this series. I hope you enjoyed going through all the examples with me. Let me know if you'd like to see more of those. I can record more examples for you. More rounds from the reality check platform. I'm really thankful that such a platform exists. It's amazing for practicing and for making our eyes more careful. So big thanks to the author of Reality Check. And let me know if you'd like to go through more examples together because I really love to I'd love to feel more for you. Let me know. 26. Final Words and My Question to You: Final words and my question to you. Yes, as I've already told you, AI will keep evolving and it will get even better at art, smoother at speech, smarted and mimicking our emotional ties during AI-generated videos. It will fool us, amaze us, frustrate us, and shape us in ways we don't see it. But I believe we still have some power. We still decide what's ethical, was worth believing it, what's worth sharing. We are not powerless in this. We are just kind of new at it. I truly can't stress this enough. In this world, in this economy, slow is a new smart. Everything about today's mass media and social media is built to make us react, click, share comment, decide faster than our intuition can catch up. Every time we stop before sharing a shocking post. Every time we stop to check a source before sending something to our friends or family or question a photo or image before believing it, we are practicing our digital mindfulness that will save us, that has the power to save our society. We will also need to constantly teach kids and ourselves how to read images and videos, how to spot biases, question sources, and look for signs of fabrication and AI-generated stuff. So where do we go from here? We go forward with eyes open. Maybe I'm a little bit naive, but I believe the world needs to build systems that protect truth, not just profit. So I really hope this course was interesting and informative for you. I look forward to hearing your experiences. Tell me, tell me what surprised you the most. What part of the course interested you the most. I really need to know. Tell me, I love hearing from you. Also, I will be publishing on the topic so stay tuned. You enjoy the course, please let me know in the review section. And if you have any questions, don't hesitate to ask in the discussion section, I will be very happy to answer and help. I've also created a separate course on how to spot AI-generated text, learn secret things about how to spot what was generated by, for example, HGPT. It's super helpful if you want to recognize when something was written by AI and just ask importantly, if you want to learn how to avoid sounding like GPT yourself, especially in professional and creative work. So I hope to see you there, and I really hope to hear from you in the discussion or review section. See you there.