Transcript from the "Course Overview" Lesson
>> Okay, let's do it. So first thing first, course overview, again, a little bit about myself and why I spend so much time [LAUGH] on AI. It's funny because, I remember, how did all this start? I think I remember, it was when some version of Stable Diffusion came out, and if you know a little bit about me, I'm really into gaming.
[00:00:26] So I have a crazy advanced gaming setup at home with a nice GPU. So I was well positioned to try some of these technologies out at home. And I was like, wow, this is really powerful. So I really got into just trying to train the best model on my daughter's face or a dog or my face and things like that, and I just became obsessed with it.
[00:00:48] And from there it just spiraled into just, I wanted to know more about this technology and how it worked. So I remember just trying to go back in, cuz I'm not a traditional CS grad, I never even went to college, I'm not a math person, but I wanted to understand it on that level, and I still don't.
[00:01:05] [LAUGH] But I remember trying, making the effort of just going back and trying to brush up on some collegiate level math courses to really understand what was going on. So, that's just how much it consumed me early on, whereas I didn't feel the need to do that with anything else in my life.
[00:01:53] So you might be thinking, I've never used Node before, but I've built React, you do know Node. It's the same thing at this point. You don't need to be an expert. You don't need to know how to build APIs, we're not making an API here. The second thing is you need an OpenAI API account.
[00:02:09] This is different than a ChatGPT account. So if you go to platform.openai, This is what I mean. So this is the account that you need. You need to make an account here. I don't know why they have two different accounts, one for GPT and one for their API.
[00:02:31] I mean, I get it, but it's annoying. So it's not the same. So if you have GPT plus or whatever, that is not the same as this. You need to have another one here. And then as far as what this might cost for this course, almost nothing. You shouldn't be spending less than, I don't know, a penny today if you just ran this thing on a for loop or a while loop forever, so you should be fine as cost.
[00:02:53] But I do think that they might require a card on file so that way they can charge you later if you go crazy with it. So, yeah, you will need that. And then eventually you just need to make an API key. So you can just click here, make an API key.
[00:03:12] And we'll go through that set up when we use it. Those are the only two requirements to get what we need from this course. Where am I at? I'm over here, there we go. Cool, any questions on that? And then as far as Node version, let me see what version I'm on.
[00:03:34] I'm on 18.14, so at least that version. Any version beyond that. I'm sure any version of 18 is fine, but just make sure you're at least on that version and beyond. Okay, so some of the things we're gonna be covering in this course. I mean, I feel like I could have talked about so much.
[00:03:51] Honestly, the most difficult part of this course was trying to limit what I talk about, cuz I actually have enough stuff on my mind to talk for, I don't know, at least three weeks worth of stuff. So trying to get it down to a few hours in a day has been extremely difficult for me.
[00:04:06] But luckily, I used GPT to help me figure that out, which is great. [LAUGH] It's really good at that. Because, yeah, I just can't think through all that. But anyway, some of the things we're gonna be covering is we're gonna be building a chat experience. So if you've ever used GPT, we're basically gonna build our own version of that, very simple, in the terminal.
[00:04:22] So we'll just be writing a chat experience in the terminal. Like I said, there is no front end element to this. We'll just be doing everything from the command line just to keep it simple, keep it focused. We're gonna be covering things like embeddings and vector stores. If that sounds foreign to you, after today, you'll know exactly what this is.
[00:04:41] And this concept of embeddings and vector stores is very common in the ML community, the AI community, the data science community. But for us normal people, when I learned about this, I was just like, this is insane. This makes so much sense about how this works and it really simplified a lot of things for me.
[00:05:01] We'll talk about semantic search and just search in general. We've all used search programs before, we use Google, we do things like that. We're gonna talk about the difference between those experiences and something called semantic search. And then we're gonna build our own semantic search experience, which is pretty impressive.
[00:05:15] Document QA, which basically is just the ability to ask questions and get back factual information off of a knowledge base. So if your company has documentation, if your company has, I don't know, some wiki, and you wanna be able to ask questions based off that knowledge, that's what this is.
[00:05:34] It's gonna show you how to grab that data, how to format it a certain way, how to present it, how to make sure it's correct and factual, and then how to match that against a human query, basically. So pretty cool. We're gonna talk about function calling, which is kinda crazy, but basically gives the AI the ability to call functions on other systems, which effectively upgrades the AI to do anything.
[00:06:02] So if you had a function that says, release nuclear warheads, it could call that for you, [LAUGH] and that would take over the world. So, yeah, basically, we'll talk about that and how to upgrade that. And then throughout all of this, I'll always just be talking about all the different scaling issues in production, things you have to look out for when you try to deploy these things to production.
[00:06:25] Because what we're doing here versus what you have to do in production is different. And I know that's true with any type of app, but it's especially true with these types of apps. The things you do locally are going to be way different than the things you do in production.
[00:06:42] It's easy to actually build these things, you're gonna find out. It's hard deploying these things [LAUGH] because there's just so many gotchas and things you have to think about. So, outcomes, hopefully, by the end of this course you've built, well, you will have built some great examples that you can reference going forward.
[00:07:02] I think on their own they're all really good foundations, but they're still just gonna be scratching the surface on what you can do. You can combine a lot of these things to do some really cool stuff. You can expand on these things, but hopefully it's a really great reference point for you to go back and be like, yeah, how do I do semantic search, or how do I use a vector store?
[00:07:22] You can use this as a reference to go beyond that and really create some exceptional things. And I'll get into some examples of some things you can create later on, yes?
>> Are we doing OpenAI only or are we gonna explore other models and services?
>> Yeah, so for this course we're only going to use OpenAI, cuz there's a feature there that we need that other models don't have, but I will be talking about all the other ones.
[00:07:48] But just to keep it simple, we'll only be using OpenAI today. And we won't be using their best model. I mean, you can if you want, but we won't be using their best one cuz you don't really need it. So I picked the one that's the cheapest and the easiest to use but also has the most features, and that one is one that OpenAI has.
[00:08:08] But there are a lot of other alternatives out there, depending on whether you want it to be hosted or whether you wanna host it yourself, and then that depends on what machine you have. There's so many things that go into it.