Decoding AutoGPT: A Glimpse into the Arrival of AGI

0:00Since the beginning of ChatGPT, it's always had this one limitation that really kind of held it back from, you know, people really being afraid of it being released into the wild. And that is a fact that chat GPT, GPT-3 and GPT-4, they don't really connect to the internet. Some people have tried to make some plugins that do that. They don't really connect to the internet and they don't really run on their own. They need a person to prompt them, tell them what to do. And other than that, you know, it's just essentially a tool. Now that is all changing with what is called auto GPT. This is kind of seen as one of the big inflection points in GPT and chat GPT itself.

0:36and a lot of people are saying that this is kind of the beginning of AGI or artificial general intelligence and kind of taking the next step there. So today on the podcast we're going to talk about what exactly AutoGPT is, what the implications are of it for the internet, for the workforce at large, and what people are currently using it for. So essentially right off the bat what you should know about AutoGPT is it is a open source, meaning is not controlled by a company the code is out there and anyone can use it for free project on github and this project was created just a couple weeks ago but it already has over 60 000 stars on github which github is essentially just a code repository or a place where developers can upload all of their code share code look at code merge code if you're working on a coding project you're going to be GitHub and having 60 ,000 stars on a project is actually massive it has over 8 ,000 forks Meaning people are making variations of this taking the code and building on it.

1:37This is a very Very big project a lot of people are very interested in this So what is interesting to know from you know essentially the side projects? I've been built it is a ton of people are taking it and what's auto GPT does is it can And it's essentially what some are calling an agent, meaning it can go and do whatever you tell it to do. So what you do is this thing connects to the Internet. It can make searches. I've seen it buy pizza for people. You give it an objective. So different than a specific prompt, you could say your goal is to be my business analyst and to make my business more profitable.

2:18Here's all the information about my business. This is what I sell. this is what I do, yada, yada. Your goal is to just make me more profitable. And it just takes an objective, or you can give it a few different objectives. It breaks this into a task list. So if your objective is to make more money, it looks at your products, it can do product market research. It breaks it into a whole bunch of different tasks. And then when these tasks are too complicated for this one quote unquote agent to solve, it spins up multiple variations or agents below it. So, you know, kind of like on chat GPT, how you'd have different like threads, if you're asking chat GPT for different questions, it's one thread that's like the boss, and it's controlling all the other threads or agents and telling them what to do.

3:00It's like, okay, so we're gonna make more money, I want you to do product market research, I want you to go and look at what competitors are doing here, I want you to go and look at different ways we can cut costs in the company, I want you to go and be a social media manager and spin up, you know, daily Instagram posts with captions. I want you to go and create image assets for the website. So essentially, you set this thing loose and you can give it different things too. You can say like, I want you to make me more profitable or I want you to come up with a new product that we can sell. Your budget is$1 ,000.

3:32Your budget is$500. Do X, Y, and Z, whatever my goal is within that budget. So this thing can take budgets. This thing can take objectives and it runs on its own. And anytime that one of its agents is maxing out, you know, like you've seen this with Chad Chippity, like you get a really long conversation going and it kind of starts, I don't know, becoming less, it doesn't remember so much something that it talked about before. All the data that all of the agents are gathering can be combined. They're short, they compress them. So all the data is being combined. It knows everything. This thing is crazy.

4:07I mean, I've seen this thing. Someone has it go and buy them a pizza because they gave it the access to go do that. Someone said, you know, I have like a weekly budget for groceries. I want you to go make all my meal plans based on this budget and based off of these dietary restrictions. You know, things that you could tell ChatGPT ask it for a list, but this doesn't just do the list. This goes and makes the meal plan. Then it goes and figures out how to integrate into the API for, you know, a shopping app, then it goes and orders all your groceries to your front door for you. Like this thing is like connected to the internet, interfacing with APIs of software and doing the stuff.

4:49And it's people are hooking it into GPT-4. So this thing is incredibly powerful. And a lot of people are freaking out saying, you know, this is the next level of AGI. Like this AI is beyond just giving you an answer in the chat GPT interface. It is now executing. And if it doesn't have enough bandwidth to figure out a problem it creates an entire task list and delegates its tasks to lower versions or other you know other agents on its network now the thing to know with all of this is it's not just like a software that anyone can use this is something currently for developers although some people are making versions that are useful for people that don't know how to code but at the moment you know you got to go get your open AI AI keys and integrate them and you have to go get some other stuff and integrate it and there's kind of like a command a command module that you that you talk to this thing through code and you get it all set up so it's a little bit more complex than just you know anyone can open it up and talk like chat GPT although people are making more simple versions of it so this is gonna be I think a really massive trend where we have these AI agents essentially set loose all over the internet doing different things writing blogs posting social media posts you know you can just tell one like hey you're my social media agent figure out a way to get me more followers on Instagram.

6:05And then it just goes and does all the research on what people in your niche are doing to get more followers on Instagram. And then it goes and it gets other agents to write the posts and other ones to generate the images and other ones to edit the videos and other ones to go comment on things like, like, it's crazy because you're, you're giving it an objective and it is running through a list and it's like doing the work. It's not just, you know, giving you responses. So this is a really big paradigm shift. A lot of people are, you know, afraid of it, but people on Twitter are sharing some really cool stuff they've been able to get done.

6:37Someone had it prepare podcast outlines by searching the news for what was like trending in their area and then preparing podcast outlines, talking about all those new trending areas and preparing it and giving it to them. and I mean I've also figured out how to do AI podcasts in the past so I don't know if they got this deep into that maybe they just you know how to prepare like a spreadsheet but you could also tell it to go integrate with something like Well Said Labs that reads the podcast and then you could go and integrate it into something like Spotify for podcasts and post the podcast so like you could you could literally spin up an agent that is doing the research and creating the content at posting the content, replying to comments on the content, starting a social media account, like a Twitter account for the podcast.

7:27Like, these things can get very, very complex. These things will get very crazy. Other people are having their AI agents conduct product research and write a summary of, you know, the best headphones to get. Other people are, someone got it to become their personal investment analyst and designed and autonomously gather pretty much analytics of the markets. And it saved him a ton of time. It went and looked at people's sentiment about different companies in the energy sector and make recommendations about which ones to invest in, yada, yada. So these things are getting very complex, very, very complicated, very cool.

8:11some people are worried you know what happens if you really can create these agents that for example you say get me more followers on instagram or tiktok or twitter and it just takes over an account and does all this stuff like what is going to happen to all the people that have that job right are these agents going to be better than them are they gonna be worse than them as if it's capable inevitably it's a matter of time before these things are able to really do that and so i think it comes down to the question of a lot of people are talking about i listened to a podcast yesterday of um it's uh you know lex friedman interviewing the ceo of open ai sam altman and he was you know there's a bunch of different theories or concepts they're talking about with this where essentially they're saying like there's a debate about like the dignity of work where people are saying like we need work in our life to be happy like it gives us a sense of purpose um sam altman obviously open ai it's about to disrupt a lot of jobs and so his position is more that um well yes work might be good you can change the type of work and maybe the more menial tasks are going to get automated um or maybe why you do work is you just do work because it brings you a sense of fulfillment or a self-expression right there's all these arguments that i'm not picking sides i think it's you know i definitely empathize to a large degree a lot of people that fear getting replaced or something, you know, you might have spent 10 or 15 years or went to college for studying all of a sudden can now be done by AI very proficiently.

9:43And like, it nullifies all that work you put in, like massive empathy for the struggle that that would be. And at the same time, it's hard for me to say, you know, this technology is bad when, you know, this technology might be what's capable of bringing, you know, perhaps an entire third world nation out of poverty by automating a lot of tasks that were previously difficult and allowing them to have more freedom and prosperity in their country. So, you know, there's a lot of there's a lot of sides to this. It's pretty crazy. Like it's no doubt disruption. It's no doubt going to be very uncomfortable for a lot of people.

10:18And, you know, the the argument is that despite the discomfort, this is how you raise the global standard of living and help eliminate poverty by automating previously menial tasks and becoming more productive and more wealthy and being able to create more outputs with less. So this is absolutely crazy. All of these arguments, philosophical and, you know, economic and otherwise, I really do believe are plugging pretty heavily into this whole auto GPT thing, which is now able to essentially become a person managing different instances. So it's gonna be really crazy to see how this evolves. This thing came out two weeks ago um and you know with over 60 000 likes on or stars on github this thing is growing very fast a lot of people are going to be using it iterating it forking it making new versions this thing's open source um so anyone can use it for free so it'll be very interesting to watch how this unfolds in the future and i'll definitely keep you all updated on the developments that this make because i think this is a very pivotal pivotal point for ai and the future of jobs and what we're going to see as far as the disruption coming out of AI.