ChatGPT Complete Guide: for Developers, Students and Business

1. INTRO COURSE CHATGPT A Complete Guide Class: Okay, So welcome to this class, chat DPT, a Complete Guide for professionals and students. There is a revolution happening right now. Now, making alliances with artificial intelligence is more important than ever. Now. Ai is powering education, programming, and business. Anyone can start creating like no time before. Worker. Always with the help of AI. With latest AI, you can try having a small team or even being a solo developer. Possibles outdo side. What is depicted, diabetes or a plaintiff by OpenAI basic of 3.5 using human form and learning with chemo feedback. Keep it has one of the 75 billion parameters and blast radius. 570 degrees of texts for competition. Its predecessors, GPT-2 was over 100 times is Ballard and perhaps being very good in answering questions, imagine the power of GPT-3. Do need to learn these because DBT for is coming and you need to learn the foundations. This increase in the scale drastically changes the behavior of the model. Typically three is able to perform the task. It was not explicitly training on lengthening sentences from English to French with few to no training examples. This behavior was mostly absence, typically two. Furthermore, for some tasks, give T3 or performs models that are explicitly train it to solve those tasks. In other tasks, it falls short. Baking use open area is breaking the internet again, just like it did it with belly. In only five days of it released in November 2022, reaching 1 million users. Only in five days. Netflix took 41 months on Facebook. Okay, but what you will learn in this course, I will explain chat DPT technology US and a student will see how can you boost your learning using chat DBT as tuple and alerting system. Also chat DBT for workers and business, how to generate content for developers using, using the API to JEPD theory, practice and GTP T for developers create, call it by your parameter languages using data DPT API and a website. You can create some code in Python in a lot of languages. Okay, What about me? I'm a full-stack developer and I helped build apps for iOS and Android by myself and do some blockchain development in Solana. And now I'm working with first be doing this work. I do, I do it to the model, the read and the skinny. They be wrapping the topology and the animation. So thanks. I can't wait to have you in the class. Welcome. 2. ChatGPT Revolution All you need to Know About ChatGPT resized correct: Okay, so welcome please video, we're going to see whites. Igbt is leading the artificial intelligence of evolution. This is the everything you need to know, Summary. Okay, But first, OpenAI is breaking the internet again. First with the Lean, the image generator. Now we can DBT. There is a chat bot with natural language processing and reinforcement learning. In only five days of its release, it reached the 1 million inducers. Netflix took 41 months to this milestone and Facebook ten months. Now. Typically as we say, it's a natural language processor with human reinforcement learning. So it receives human feedback to keep better responses and original responses the majority of the times. But when it's a DPT, It's an interactive artificial intelligence chatbots developed by OpenAI. Current version is GPT-3, 0.5 using human reinforcement learning. With human feedback. Elijah, DPT has 175 billion parameters and what's printed on five continents, 70 gb of text. For competition. It places or EBT to West, 100 times smaller with only 1.5 billion parameters, is increasing. Scale drastically changes the behavior of the model. Gpt-3 is able to perform a task. It was not explicitly tested on translating sentences from English to French with fuel to not bring examples. This behavior was mostly absent in keeping it to. Furthermore, for some task, GPT-3 outperforms models that were explicitly train it to solve those tasks. Output in other task, it falls short, so you need to check, okay, but hope you'll chat. Dpt, PT was created by San Francisco based company OpenAI, our research Ben to leave by some outmanned and was previously the CEO of Y Combinator. And bacteria by touching best tools like Microsoft, chatbot t is a language model, is a huge language model. Llms predict next towards generating code and text. Sometimes responses are unpredictable and bias it. So the solution implemented reinforcement learning with human feedback arrow HF training. Chapter P D has been tallied the new goal, both y first because of the marketing success and the growth. Only from November. Telepathy was the most popular site number. 1879. And in December was they mediate most popular visit and site in the world. Only one month. You can see. It is scale 107081 positions. The source is similar web chat DPT, key facts, TTP t is artificial intelligence chatbots that can answer almost any question you ask and follow conversations is a interactive artificial intelligence. Generating new data. Most of the time you need to fit with some initial inputs on wards as dialogue, The Times polish is the world's first truly useful Chatbots. Child typically use a human labelers that got hired to correct artificial ear responses into accurate and Benedick. Your central for using reinforcement learning with human feedback. Jt petite can give incorrect responses. That seems to be true. It's important to validate the data generated with trusted sources. What can you do with ten DBT? You can get an eight explanations to topics and generate text. You can answer general acknowledged questions that e.g. what's the capital of Nigeria? It can help you summarize information from a piece of text. You can pass at an article and you can get the summary, help you create code in any programming language. You can say, you need me a golden bite of that, loops for the elements and try to have a string at zero every time she iterates and it will generate the text. You can generate great thing or tests, competition. Can you write a short story or the robot can provide definitions and explanations, please let me know if there's anything specific you want to me to do for you. Okay. But what is depicted three, GPT-3 generative pre-training transformer trip, status, language generation model developed by OpenAI. It uses neural network architect to a transformer which is printed on a massive dataset of text from the internet regenerate human leg t3 has 175 billion parameters. Limitations of DBT. You need to check the data that is accurate with sources and check if the data is not biased or toxic. Okay, but how can you use to boost your learning process and help you with generating summaries as question or answer. It can be like your tutor. You are asking questions and you can find, tune the questions to work over a specific topics recursively. What specific subjects? It can help you as a language translation and texts competition, chat, DPT, sub, use cases. You can have your content generation used tomorrow service. You have an inquiry, or you can try to give you some ideas of how to solve these customer service, email or document generation. As a worker assistant, you can ask anything you want to know and it will provide you in a lot of times oriental answers, grammar correction, text summarization, and original brainstorming. But typically we use to generate content for my blog. Guess, can you put D can be used to generate content for your blog, but it's important to note that the generated text may not always be of high quality or appropriate for your specific use case. It's also important to consider any legal or ethical implications that the text is not by asset or toxic of the user-generated text. Such as issues related to paramecium that is not the same like other article on the Internet that has been originally write it or copyright infringement. Additionally, it always is a good idea to check the ureter next forces and preview it before publishing. So check this answer was provided to me by LGBT. Okay, so now let's do some examples. Let's say I want this book, the nature of space and time. So as it give me a summary of the nature of his patient Time by Stephen Hawking. And linear summary denitrification is a book, is a book written by Stephen Hawking and grow your Penrose. The book provides an overview of the fundamental nature of space and time. Then I say, give me a summary of the nature of space and time. And I pass all these texts. I pass it to see if he's originally response. You can regenerate the response. If you can regenerate is ordinal data as possible. So let's give an interesting, an interesting exercise, e.g. look, this creates me a Python code that loops through ten times the coolant number to a string. Basic example. Only to test That GBD can create code. You can see it has an empty string, then it has a range. Then for the range of ten, it, the string. It adds to the concatenate the string with the current numbers 01. Then print the string. Here is the explanation. Okay, so thanks, we finish this video and let's continue our course. 3. Generate code with ChatGPT in python to add text over an image jpg centered: Okay, So welcome to this tutorial. We're going to see how to add a text to an image. And this code was generated by artificial intelligence. Chat GPD eyes even encode nothing. So let's check the code. Here's a code. It's LGBT provided. So let's e.g. say, you can put as 0 here we open the image, we choose the font size, we get the size of the image, we calculate the coordinates so we can center the text. So the text here is, they're all text here we calculate the coordinates. Then with these drought or x, we add the text and then we save the image. So now let's run this code. Here we have our image. Now let's run here we have our image. Some texts. This is very powerful because you can put text with your images for social media and automate e.g. create 30 much in a row, in a queue. So you have three images for all the day. So is very powerful. Now let's see. We tell DPT how we generate the code we're using Python and livery. If you are running Python tree, you will need to install. Below, that is with this IP3 install Pillow. Now, let's see, we try DPT, how we generate the code. Okay, So we're chat DPT, GPT-2 OpenAPI, and then we're going to ask to generate a code in Python to add text over an image center. We run this question. Is, it uses a liberty. You import the dependencies. You open the image with image open. Then you draw the image. You create the object. Find the font and text. You can put any text you want. You add the size of the text. You calculate the y and x-coordinates to center the text. And you draw, you draw the image. The text. You pass the coordinates to draw the text is centered or you can put it e.g. more. In the more upper or lower. The possibilities which are DPT are immense. You can call in any language, you can use code very fast. You can use this as a building block. So the possibilities are amazing. Work. And work is marked with the help of artificial intelligence. Thanks. 4. Using Chat GPT To Learn How can AI help Education: Okay, So welcome to this video. We're going to see how to use chat DPT. That is a powerful tool that can help in education. Okay, The first thing is to know how to use chat GPT in the right way. By knowing that has limits. Chat GPT can be a fierce ally, research, assistance, and be your best friend and companion in your journey to master science, arts, and technology. But how can you use it to learn as a powerful tool to chirp your mind, we need to learn how to interact with the model who has tried alignment pieces of knowledge. We must know that has some artificial intelligence alignment because datasets of GPT-3, the current version or constrain it to 2021 data and current status of the reinforcement learning from human feedback. But having a width AIS, a trained AI is on the rise. Technology will consolidate in the next years. And artificial intelligence will get better at faster rates charged UP community and users. We're constantly pushing the limits by creating content when hence and stress test to future versions of IGBT. So we as a community, we get to know where are the limits in a specific fields and we can help future versions of AIs, GPT-2 try to see what is limits. What are the limits? We can have opportunity in those areas that are not fully, fully automated. I'm fully understand it by artificial intelligence. Chat DPT can assist in location, but how would contain generation? Telling me the most important questions of algebra you can ask, GDP can be used to generate the vocational materials, quizzes, flashcards, and summaries of complex topics. Chapter PTS. This possibility, I love it a lot. Chat DPT can be a tutor, e.g. tomato to do physics tutorial, biology tutor, your arts tutorial that can follow up your research line and provide you with best insights with question and answers using the Socratic dialogue. So you can, you can ask a lot of questions of mat and ask for definitions. Is amazing. He says, improve. Gpt can help you grade your essays, suggest improvements and allies and deconstruct language. Learner or late chat GPT can help you translate and it can be a strong alternative towards translate. Research assistant, this is very similar to tutor. Research assistant can help us do them by giving definition summaries, helped visitation, brainstorming solutions and with some research methodology, chat GPT is evolving. Humans can actually learn and improve encyclopedia scientific articles and machines can actually learn from the new data from humans with reinforcement learning to get better over time. They future version of chat DPT will be a more intelligent Artificial Intelligence. Train it with more data sets within reinforcement learning and possible being a lot more accurate. Artificial intelligence and GPT challenges if future of AI, we must build a strong community of testers and developers. So next models become better at faster rates. And all humanity benefits. Current responses can be incorrect, misinformed by asset, or even on predictable. So we need to work as a whole, as a community to push the limits together, understand the limitations, and improve future versions. We need to learn how to interact with IGBT to create a know-how to empower our location. So what is the strategy? If you love artificial intelligence and GPT, start using it to integrate your workflow. Which articles and content to make public you useful insights, good practices, and educate yourself. Discovered the limits of the tool for particular field, e.g. for law, for engineering, for robotics, for math research. And publish your results so it can be fixed it this AI or other sisters or brothers in future releases. So thanks for watching this video. Let's continue the course. 5. AI Tutor ChatGPT as Tutor Strenghts and Limitations: Okay, So welcome to this video. We're going to see if chat GPT can be an artificial intelligence tutor. And if it can be what are their strengths, its limitations, and how to use it. First, what is a good tutorial go tutor each instructs and guide on a subject with a porpoise. A tutor in general terms, a tutor will be an academic mentor that will open the doors of perception. Acknowledge that make students able to seep, understand concepts, develop creativity and critical thinking. They will guide and instruct on essential skills needed to master some field or branch of science, of art or technology. Okay, but now let's see some top qualities of a good tutor. A good tutor has a strong foundation. So the subject of S to V, passion for the topic is a good listener, most connect with the student and adaptability tool to adapt to what the students need to understand to awake the curiosity of the student. Help the student to develop creativity on the subject, help the students develop critical thinking. Has empathy, can connect at an emotional level? Has communication skills. Use the correct wars, the verbal and nonverbal communication to boost student imagination. You will, the Socratic method will take the dialogue. The student find the answers by himself or herself. Node itself has humility. A good tutor know her or his limitations, and it's able to step away if the student needs. So even a space to a new tutor has a method. Good tutors helping students develop their own research methodology so they can continue learning by themselves when the tutor is no longer there. Okay, both can chat PPTP, I go tutor. Let's see some pros and cons. Gpt can not sorted to the human foot, cannot empathize, acumen level generated must be validated. They would honestly with trusted sources. These are three cons. Students will need to learn how to interact to find a learning style and discovered artificial intelligence limitations. Now let's see some good things about GPT. Can explain, provide information, teach concepts on a very wide universe of human a global. It is training in almost all sciences in programming, physics, mathematics, biology, molecular biology, medicine, chemistry. I mean, liter to poetry, philosophy, history. It's a very powerful compendium of human acknowledge. Also chat GPT uses dialogue, that is Socrates meiotic, that makes these question and answer. So you can be finding answers by yourself, can help you. 2047, all days of the year, you have a you have a question and I was e.g. one time at 2AM and they have a lot of questions of chemistry and math. And I go and ask too to this term, DPT also provides helpful information and explanations to assist with learning. This is very important. This is a crucial part of this video. Doses substitute human tutorial is going to create a new category, artificial intelligence tutorial that has now because of the way it was training and reinforcement with learning with labelers has some limitations. Both has also a strength. So, you know, you need to use it with doing your own research. This is not a vocational advice. You'll need to research and validate all the information rigorously with the best sources. This is books and cyclopedia and a correlation diversities. White papers, scientific papers to validate all the information with data. Finally, artificial intelligence is on the rise, is growing intelligence as faster rate. So very soon, maybe the dream that artificial intelligence tutors can reach astonishing and incredible levels. So thanks for watching this video. 6. Math Learning Assistant Math Tutor: Okay, so now we're going to use Chegg GPT as a learning assistant. So you go to open a.com and you click on Try new sign-up. If you don't have an account, Let's use G, B, D as a math instructors. So e.g. let's ask what are the essential branches? Okay, So usually for simple you can, in my case, I'm interested in geometry, e.g. we can ask what basic geometry topics. And we can have a conversation of n. We can be learning like feminine tutor in school, in college. You guys, e.g. Sumatra is me. That's e.g. how? Our geometry, linear algebra. Okay, So tax force in this video, as you, as you see, you understand the concept that we can have a load, a constructive dialogue and use a lot of ways to answer a chat bot. And we can ask for a specific topics. Summaries. Very powerful. And it really is intuitive. And you really connect with chat because you have instant answers, it turn notification. But the downside is that you need to check all the information with trusted sources, encyclopedias, good books, university papers, white papers, scientific journals, etc. So thanks. 7. Intro to ChatGPT API correct: Okay, So welcome to this video. We're going to see an introduction to the chat DPT API. Okay, but first, what is an API? Api stands for? Application programming interface is a way that two programs, two computer programs can communicate, e.g. your application can communicate with an API for retrieving information. So we'd send some information and to programs and communicate. It can be an application, can be a video game, can be a hardware Internet of Things that has an embedded system that can communicate to gather data from the real world, e.g. Internet of Things can gather that from my current location with sensors and they can communicate with an API to get some data of other locations on the Internet and then can make decisions according to the data that is beyond their current position, e.g. okay, Now let's see OpenAPI. The OpenAPI can be applied to virtually any task that involves understanding or generating natural language or code. You can create apps, video games, software that users can use the chat GPD, text generation, denotative artificial intelligence by providing a prompt, e.g. your software, your video game, your application, your Internet of Things, You Are, needs to provide a prompt. So it provides some information it receives from the API response for the HTTP requests. That is these Thailand for an ice cream shop, we serve up as miles with every scope. Okay, but now let's see some OpenAI API, common task it can do content generation, summarization, classification, categorization on sentiment analysis. That is, this is, this is great if you are greatly e.g. trading, trading routes for the stock market or crypto market. You need to classify articles and pressed that is very recent and you can get some classification and sentiment analysis and you can know if the market you're going to, you can make a calculation with your algorithm on your software. Also, you can do that extraction, correlation and many more content generation. Continued duration you can use to create e.g. a. Blog post or e-mail campaigns. It can provide you some templates and then you can manipulate the text and adjust. I'm fine tuning that X. You can quote or articles for SEO, you can call for creating videos scripts using the content duration. You probably the prompt and you can receive some texts and you can work with that text. Summarization, summarize some articles provided by user as prompt. So you can provide a large article and you can get a paragraph, so you can get these summarized, and you can use this summary e.g. for social media for your application to promote another large article on the Internet can be in your site on a third-party site. Classification, categorization on sentiment analysis. Use case example. You are building a trading bought that takes decisions with sentiment analysis and classification. So you can feed with the lightest media articles and I can take a decision if he's a good time to enter. You take the decision with an algorithm, you analyze a lot of articles and if you see the market is going bullish or bearish, you can make a decision. You can connect to the API of the broker or the marketplace, the centralised exchange. And you can take some pockets or stock decisions. You will need to find two sources to feed the eye. And you must do your own research on that. That the instruction that extraction is a process of collecting or retrieving separate types of data from a variety of sources. You can extract information from a lot of different articles and get the things that are disparate or that are inclusive. You can retrieve the disparate and inclusive data from different sources and concise in one piece of work. Translation, you can translate. You also can use a Google Translate API, but you can also use OpenAPI for having a bubble translation and pick the best one. Okay, so now let's see an overview of the documentation. There is one beta dot OpenAI slash docs, and we're starting with the Quick Start. Quick Start. Openai has trained a cutting edge language model that are very good at understanding and generating text. Our APA provides access to these models and can be use it to solve virtually any tasks that involve processing language. So as we see content generation, summarization, classification, categorization, sentiment analysis, that extraction, distillation and many more. So we see this image that you provide a prompt. So e.g. you can run the API and you do ask, suggest one name of a horse. And it suggests an image of a horse. Not bad, no. Now let's suggest one name for a black horse. And you will communicate with the API and you can receive in your application. Suggestion. You can have e.g. suggests three names for a horse that is a superhero, and then go to just three names. Also, if you want to be more precise, you can provide a context. So you can say animal cat and you can provide examples. Captain, super Cloud, agent, fruitful degree we're filling dog, prove the protector. Wonder can even see barks lot and you can add animal. You can have e.g. lion. So it uses a juicy examples to your prompt, can help communicate. Parents are more answers. So if we provide this king courageous, the mighty mane and the real defender, and you see, it will change this e.g. for bigger. It will change the results. Stripy Savior the myTestApp as one and worrying the warrior. Adjusting the settings, temperature one will always give different results. You see different results, very different. So you need to lower the temperature because temperature is set to zero will always give the same results into super stallion. Again, the same result. So you can play with the temperature. The more the temperature, the more, as you see here. If you put one, the confidence will be lower. So it will give some more unpredictable results. So a good temperature you can test e.g. 0.6 and it will give different results without losing a lot of confidence. So you need to do need to play with the temperature. Building an application. Finally, we're going to build an application in the next tutorial. And as you can see, you need to clone this Git repository. You change to the directory. Do change. You go to his file, you rename it, you remove example. You need to add your second, you create a new secret key. You need to be logged into this page. And you need to be logged into your API. Need to go to this location. And also, you will get some, you will get some free credits if you are a new user, then you will run the application with NPM install. You can npm run dev and you can change in this file in pages unless API folder, there is a file, generate dot js. And then you can change the prompt. Initially is to generate an animal, an animal suggestion of a name, some names suggesting pet suggestion. This is the basic example. I change a little bit the code, okay, So this is my local host, my software running the application and connecting to the API. And I can ask any question. I want to ask a question e.g. e.g. the Pythagorean theorem. And I will connect to the API and it will generate this a squared plus b squared. Let me ask another question, e.g. four plus four. Okay. Let me ask one more question.