10 Ways to Use Auto-GPT For Beginners
It’s currently Auto GPT’s three-week anniversary, so here are 10 incredible use cases. This is how GPT-4 sees and hears consciousness.
So, they made this AI game that you can chat with someone for two minutes and try to figure out if it was a fellow human or an AI robot.
This new AI model can memorize up to 20 books worth of information. What if you could scribble AI art in real-time? It’s currently Auto GPT’s three-week anniversary, so here are 10 incredible use cases. Already, it can make an entire website for you, as you can see the website is getting made on the left, and it’s automatically writing the code of all of this in Auto GPT number two.
It can create a to-do list app that does itself. As you can see here, it’s integrated with Notion, it’s integrated with Slack, and it’s actually sending messages to people on your behalf. So soon, you won’t even have to do your to-do list anymore.
It can do research for you. So, this message was, “Find all Hacker News posts that are GitHub repos, summarize the repo, and give me the repo’s URLs.” Or, for example, it can give a detailed market research report on expense tracking apps in the UAE.
This can read about recent events and prepare a podcast outline for you. As you can see, with five searches and 15 web browsers, the GPT research agent prepares a five-topic podcast on recent news with accurate references and a cold open. Look at how beautiful that looks.
This one, you can have an intelligent alter ego, and it will basically do things for you, generate tasks, implement them, and learn from itself. It’s called Gold GPT.
It can code for you. As you can see on the right side, you have the Baby AGI, which is an Auto GPT, and on the left, you can see that it starts coding, and just like that, it’s written a bunch of code for you.
Auto GPT inside of Discord. As you can see, it created a new task list and go next task list, so make a to-do list with no more than five steps. And as you can see, it completes those things and then creates the next tasks. This is beautiful to see, and I bet more apps are going to integrate this into their software as well.
You can access Auto GPT directly in your browser with Cognize’s enter, and the objective: give it a task, and it will get it done for you autonomously. As you can see, adding the task and you already know what Auto GPT does.
An Auto GPT that reads your tables in your database, writes SQL queries automatically, and sends notifications on your Slack channels. You can see it’s giving it three goals here, the first one being “find me all bookings from our booking database whose status is pending,” and as you can see, it’s actually doing the job right now. Done.
Automatic sales prospecting, basically finding the 50 top companies and finding the key people in those companies that you can reach out to. Credit all of this to Roll One. Auto GPT is truly amazing. It will do to-do lists for you, will do sales prospecting for you, but when it comes to relationship advice, it still recommends, “Have you tried turning it off and on again?” You will make 50% of the royalties if you make a song with this artist. You guys might have heard that last week, an AI song of Drake got viral on Spotify with over 600,000 views. TikTok’s getting millions of views, and then UMG, the music label, shut everything down.
Now, Grimes said, “I’ll split 50% royalties of any successful AI generator song that uses my voice, same deal as I would with any artist I collab with. Feel free to use my voice without any penalty. I have no label and no legal bindings.” Grimes says, “We’re making a program that should simulate my voice, well, but we could also upload stems and samples for people to train their own.” Mati PRD said, “Someone should make a label that just signs AI rights for artists to scale this.” Left at London said, “I’m about to make an AI Grimes Marvin Gaye covers album and single-handedly start the wildest lawsuit known to man.”
This is how GPT-4 sees and hears consciousness. So, Chris Casanova on Twitter used GPT-4, Kyber AI, Mid-journey Mubert, and Runway ML, a bunch of AI tools, to create this. GPT-4 described it as, “Consciousness could be visually represented as a complex multi-dimensional web of interconnected neurons. Imagine a vast shimmering network of glowing points of light, with each point representing a neuron. These neurons are connected by delicate pulsating strands of energy, representing the synapses through which information and thoughts flow. This neural network would be constantly evolving and changing, with new connections forming and old ones dissolving, reflecting the dynamic nature of consciousness. The true colors of the points and connections could shift through a rich spectrum, symbolizing different emotions, experiences, and cognitive processes, hinting at the emergence of thoughts and ideas.”
Would you watch this movie? This guy, Timmy, made this intro using Runway ML and their new text-to-video AI. The visuals in this clip are 100% created from text prompting. Generative AI is improving so fast. I’ll continue my journey and share my results with y’all. You can get access to this free version in Discord, but it’s limited to 4 seconds per clip. Apparently, the hardest thing to do is to get the prompting correct, so it actually has a consistent style throughout each and every clip. AI text-to-video has come a long way. I can’t wait for it to make a reality show out of my spam folder. It will probably be called “The Bachelor: Nigerian Prince Edition.”
So, they made this AI game that you can chat with someone for two minutes and try to figure out if it was a fellow human or an AI robot. Can you tell the difference? Let’s start the game right now, searching for your partner. This will just take a few seconds. Don’t close the window. Okay, I need to figure out in two minutes if this is a human or AI.
Partner: What’s up?
Me: Not much, who are you?
Partner: Taking too long to answer, huh? Just a forklift operator from Chula Vista. What about you?
Me: I’m from space, bruh.
Partner: Are you real? I don’t see how that’s relevant.
This guy is an AI for sure.
Me: Are you an AI?
Partner: I don’t think we can ask it. Are you an AI?
Me: A human would act differently.
Partner: I feel like nah, man. Just a dude trying to figure out if you’re a human or not. Let’s keep chatting and find out.
This is for sure an AI bot. You just talked to an AI bot. I know I got this reply to every message starting with “as an AI.”
What if you could scribble AI art in real time? Just look at this. This guy made a Python script that you can access right now, where you can just scribble like a hat, and just look at how cool this looks. And then he makes a little quote on the guy with the hands in the pocket, and the AI makes up what is supposed to be on the other side. And now he’s writing a little prompt: “A man on the beach with a long coat. He is writing the waves in the shoreline.” And now you can see that it turned into this really, really cool boat in the background. And just like that, now here’s the second drawing. He’s just drawing these squiggles on the side. It looks like kind of the moon rocks or something like that. Now he’s in rocks, just like that, in the shoreline. Wow, it’s looking insane. Now he’s drawing some patterns into there. He’s changing the prompt to “cell structure sharp focus photograph.” And now, what is that?
This new AI model can memorize up to 20 books worth of information. So, they’re scaling a Transformer to 1 million tokens and beyond with R and T. This is a recurrent memory Transformer that retains a quas up to 2 million tokens. Currently, the tokens for GPT-4 is 32,000, right? So, you might have tried to copy-paste a bunch of things into Chat GPT, and it says it’s too long. The same thing goes if you have a really long thread. It starts just forgetting what you wrote all the way up in a chat. So, what’s the catch? Well, there is no free lunch. Much longer inference time, plus substantial quality degradation. How most people use memory now is that they store the memory in a second database like, for example, Pinecone. If you’ve heard of that, and they use that to pull out data using the LLM as basically the model that trains stuff. Both of them are used in tandem, but this might be the way of adding memory into an LLM itself. So, it would be a batteries-included solution. I’m sorry that RMTs will become extendable with how fast this field is going.
Vector database startups be like, “AI voice has become really great, but this one is trying to make it more human-like. It stumbles, it says ‘us’ and ‘um’. Just listen to this:
If I was six, afraid of seven, um, and it helps. So what’s, if you’ve, if you look at it, six, and it helps, um, blame and it, you know. So that, that’s why I was six afraid of seven. Um, what if you could scribble AI art in real time?”
Just look at this. This guy made a python script that you can access right now, where you can just scribble, like a hat, and just look at how cool this looks. And then he makes a little quote on the guy with the hands in the pocket, and the AI makes up what is supposed to be on the other side. And now he’s writing a little prompt: “A man on the beach with a long coat. He is writing the waves in the shoreline.” And now you can see that it turned into this really, really cool boat in the background. And just like that, now here’s the second drawing. He’s just drawing these squiggles on the side. It looks like kind of the moon rocks or something like that. Now he’s in rocks, just like that, in the shoreline. Wow, it’s looking insane. Now he’s drawing some patterns into there. He’s changing the prompt to “cell structure sharp focus photograph.” And now, what is that?
What are the contents of seven? And what, and what are the, you know, what are the, what is it, why was six afraid of settle?
Because I could.
It’s very, very strange.
Why was six afraid of seven?
Well.
Why was six afraid of seven?
That is.
Abusive.
Why did the chicken cross the road? [Music]
[Laughter]
Amazing, right? This is the type of text-to-audio that I want. So many cool things happening in AI. And if you want to continue watching, click on this one. It’s pretty good, promise. You’ll like it. Don’t worry, just click it.
