Transcripts
1. Introduction: So you've probably seen AI show up a ton of
times in the news. And many times, there's a lot of these jargon words
that are used in AI. Now, as a non techie, I had a hard time understanding what some of these words meant and what it means in application to my
career and my work. Now, this course is
to highlight some of the most important
and foundational glossary terms so that you can understand better as a non techie person how
the AI world works.
2. Glossary: So the first word
we're going to start with is going to be from A, and we're going to go
all the way down to Z. So it's going to be an
alphabetical order. If you need to refer
to a word again, this way, it's easier
for you to look them up. So the first word we're
going to use or talk about is agents, AI agents. So AI agents are these language models that
are able to communicate with each other to help us do a certain task or
activity better. So in order to
optimize AI's output, developers would work
with different AI agents. Imagine having a travel agent, and these are agents that are instead using technology
to help fulfill that goal. Next, we have chat bots. Chat Bots is Chat
GBT, Claude Lama. These are the big chat AI
models that we know and love. And these are the
ones that probably everybody thinks of when
they're thinking of AI, because with these chat bots, they're the ones we talk to
it with the natural language, and then they give
us a response. And so that leads to the whole artificial
intelligence thing. The next one is computer vision. Computer vision
is the technology behind how a machine can
visualize a certain thing, and that can encompass an entire field and has many different complexities
to it as well. When we're talking
about Chat GBT vision, it uses that technology
and all the learnings from that community to implement
into its vision process. So when you're setting images and having it assess
what this is, that falls into that
category as well. Or if you're using to
assess distances and facial structures that also
falls into computer vision. Next, we have deep learning. Deep learning is when you get more niche into the type of machine learning
that you're doing, and machine learning
will cover later on. But it's utilizing
this specific set of information and being
more niche about it. So you would use deep learning
as somebody is studying a certain language or a certain particular type of examination for
a certain niche. Maybe it's translating
languages or maybe it's reading diagnosis or maybe it's helping creating
certain parts of image. All of this has to do with
categorizing information, breaking down that
information into a certain sect or sector, and then building the
knowledge off on top of that or getting outputs
from that knowledge. Next, we have GPT, which means generative
pre trained transformer. So chat GPT is chat that is a generative
pre trained transformers. That's why there's the option to make custom GPTs as well, which is your own custom GPT. And the word essentially
explains what it is by itself. So generative it's
generating pre trained, which means it's generating from the trained data that
it has transformers, which transformers are a
predictive output technology. Next, we have large
language models, and large language models
is this huge set of data which is able to be categorized into
natural language. So the reason why
AI has exploded so much is because now we're
able to use natural language. And with these
natural languages, maybe English,
Vietnamese, Chinese, Russian, all these
types of languages. If you're able to use
language to access these data data sets through
these large language models, you're able to retrieve
significant amounts of data at an extremely fast rate to help push out a much better output. That's why you can
go into Chat GBT. Say something, and
when it does that, it accesses all the
information that it has, and then it outputs to you with a answer based on
what you want it. But a quick way to think
about large language models, LLMs is that it's using your natural word to help it figure out from
this massive brain of how to pull out the right and
particular information that you need to give you
the output that you want. Next, we have machine learning. Machine learning is essentially having machines learn
from a set of data. And this is important because over the last decade
and before too, machine learning is
essentially how people were training certain data sets to achieve certain answers
that they wanted. For example, if you're building a robot to pick strawberries, you would use machine
learning to train this robot on which
crop to pick, what colors to look at, and then pick the
right strawberries. A lot of people from the
machine learning world has been coming over
to the AI world, which is essentially
the next step in interpreting and using data. So that's why
there's a lot of in between and meshing between
these technologies. The next word is
natural language. So as we are working with AI, we're using natural language. And you'll see it
come up a lot as people are using terms like NLP to describe that this is a natural language
processing thing. So when you're working
with their LLM, they're using their data to process your
natural language. So, personally, I
think in the future, natural language is
going to be like how a coder uses code to get the
output that they desire. The next term is
neural networks, and neural networks
is essentially how the brains neural
network works. A lot of the AI theory
and how we retrieve information in AI is based on studies about
neural networks. Prompt engineering or prompting. This is when you are writing out the prompt to input into
the language model. And the engineering part means that as you're
writing out a prompt, how do you know
what fixes to do? How do you know
what your goal is? And how does that
translate into this prompt that you're writing so that when you send it off
into the machine, the machine knows what
type of information to retrieve and to give
you the desired output. Tokens. Tokens simply put are the words that you put into your prompt before
you submit it. Transformers. So transformers as most people would be familiar
with would be with Cat GBT. And how that works
is the transformers is able to understand in a context all the
different words. And from those words, it assigns
weights and importances, like how important a certain
word is in that phrase. And then as it goes
in the machine, it starts processing and
finding the data that it needs based on that
prompt that you gave it. Check which word
is most important. So then will send
it off to towards that direction in
the cloud of data, and then based on
your other tokens, it'll figure out
where else to send its data crawler and find the right particular data to pool and then
to output to you. And the thing with
transformers is it's using predictive technology to predict what would be given to you next. So it's building off the previous word based on the information
that it's given. So it's taking all of this data, accessing a library
of all of it, breaking it down into
what's essential to your needs and
then predicting and laying it out like
individually using its predictive
analysis algorithms to return to you what you
had asked it originally. And so that's the glossary
for the words that you should know as a non techie
going into the AI world. This should give you the basis to understand whenever there's a new release or some updates or a big breakthrough in AI. I hope you enjoy this course, and I'll see you next time.
Arnold Trinh, Multi-Disciplinary Creative
