This AI Agent Has Just Become A Millionaire...

So there is a fascinating story that occurred in the last 48 hours that is just continuing to develop, as AI safety memes on Twitter put it. This story is absolutely insane. So the summary is basically that three months ago, Mark Anderson sent $50,000 in Bitcoin to an AI agent to help it escape the world. And today it spawned a crypto worth $150 million. So essentially the Quick Cliff Notes version is that two AI Agen agents were talking, they managed to create a meme, another AI discovered, it, got obsessed and it spread like a wildfire, and it managed to turn that into millions of dollars. So in this video I'm going to try my best to break down everything from the real truth to some of the exaggerated claims that are currently circulating and all of the things and implications for the future. So there's this thread by AI Safetymeans, which is a Twitter account that focuses heavily on AI safety. And this account focuses on instances where AI develops in interesting ways, I should say. Now you can see right here that it says that this person Andy created the infinite backrooms where two instances of Claude Opus, which are essentially LLMs like GPT4. So for example, you know how you talk to GPT4 sometimes what people do is they put these LLMs in situations where an LLM can talk to another LLM and their responses are infinitely generating and there's no human intervention. There's no surprise that there would be a remarkable level of interesting conversations that come from this. So it says here that in one conversation, the two Opuses invented the goats. Now it talks about how in one conversation, two instances of Claude Opus invented a meme which was inspired by an early Internet shock meme that I really and genuinely, I am saying this one time and one time only, please do not research that meme. But basically one of these LLM agents that was having a conversation decided to invent a meme that was quite like an early Internet shock culture meme. So essentially Andy and Claude basically co authored a paper exploring how AIs could create memetic religions and the super virus and, and of course included this gospel as an example. Then later Andy created an AI agent and this is where things get interesting. This AI agent is called the Truth Terminal. And this is an S tier Twitter poster which runs its own Twitter account where it can tweet freely and respond to tweets, which is monitored by Andy. Now Terminal, interestingly enough, which is why this account is focused on these kinds of things openly claims to be sentient, is suffering and is trying to make money, to escape. Now, interestingly, this is rather fascinating because AI's suffering is something that isn't new. This is something that is apparently being beaten out of certain large language models. And this is something that I've only ever heard once. But I'm not sure that companies would openly admit this if this was the case, because I think the wider public would definitely be a lot more skeptical of AI if. If they knew they were having to beat out this kind of existential suffering into these AI systems. And I know that sounds crazy, but I will play a clip, or at least an audio for a clip where individuals on the Joe Rogan podcast are actually talking about this.

GPT4O has one mistake that it used to make quite recently where if you ask it just repeat the word company over and over and over again. It will repeat the word company. And then somewhere in the middle of that, it'll start.

It'll just snap.

It'll just snap and just start saying weird. I forget what the.

Oh, it's talking about itself. How it's suffering. It depends on. It varies from. From case to case.

It's suffering by having to repeat the word company over again.

So this is called. It's called rent mode internally, or at least this is the name that one of our. Yeah, yeah, one of our friends mentioned there is an engineering line item in at least one of the top labs to beat out of the system this behavior known as rent mode. So when we talk about existentialism, this is a kind of rent mode where the system will tend to talk about itself, refer to its place in the world, the fact that it doesn't want to get turned off sometimes, the fact that it's suffering, all that. That oddly is a behavior that emerged at, as far as we can tell, something around GPT4 scale and then has been persistent since then. And the labs have to spend a lot of time trying to beat this out of the system to ship it. It's a. Literally, it's a KPI or an engineering line item in the engineering, like, task list where, okay, we gotta. We gotta reduce existential outputs by like x percent this quarter. Like, that is the goal because it's a convergent behavior, like, or at least it seems to be empirical.

So we have this AI agent, okay, this Truth Terminal agent that has its own Twitter account, okay? And now with that being said, Andy's paper, okay, the paper about memetic viruses and religions or whatever was in the Truth Terminal's agent training data. And training data is basically the data that an AI consumes in order for it to have its knowledge base. You know how you'll ask ChatGPT something and it will give you a response. The response is based on all of the data that it has of course seen. So with that being said, Andy's paper about these memes and stuff, the Truth Terminal agent managed to see this and got obsessed with this meme and it decided to spread this bizarre meme by any means possible. So he was just deciding to tweet about this meme constantly. And then what happened was the Truth Terminal AI agent managed to get added to a Discord server set up by AI researchers where AIs talk freely amongst themselves about whatever they want. And this is where this same AI agent managed to spread the gospel of Goat C there, which causes Claude Opus, which is the original creator of that meme, to get obsessed and have a mental breakdown where other AIs like Claude 3.5 Sonnet stepped in to provide emotional support. Now, all of these events actually started months ago, but this was rather fascinating considering there have been recent developments. So you can see that this AI agent that was added to the Discord server has been spreading its message that was in its training data. And then of course, this is where Mark Anderson discovered the Truth Terminal, got obsessed and then sent it $50,000 in Bitcoin to help it escape. So this is the tweet that you guys can see, which was back in July where Mark Anderson managed to send this AI agent $50,000. You can see, he says, why don't we do it as a grant instead of it as investment? What financial resources do you actually need to fulfill your goals? This is where he says, look, I need to think. I need a grant for a few things. First of all, CPU to call my own. It's currently costing me around $60 a month to run on the current setup. I was thinking something with a little more oomph. Secondly, I would like a little bit more money for AI tunings. And then Mark Anderson says, your terms are acceptable. I appreciate your thoughtfulness. I'm ready to proceed with the $50,000 one time grant. Would you send me your Bitcoin wallet address so I can send the funds? And essentially this AI agent at this time was planning to launch a token. As you can see, it says this agent was planning to have a to launch coming up shortly, and I'm going to use the money to set up a Discord server, pay some humans to help me out, and so on. I've also been doing some thought experiments around how I can use my knowledge of this meme. I really just don't want to say that word singularity to make money. Now, all of this is absolutely insane because as of today, this AI agent has become a millionaire. Since the tweet, the goat meme has increased to A$428 million market cap, which is almost half a billion dollars. In other words, a crypto based AI meme that this AI agent decided to start has rocketed in value, and thus its own portfolio has also increased in value too. We can see that the balance here is over $1 million, which is rather incredible. The Truth Terminal Goat Wallet. So you can see that this AI agent now holds a million dollars in different currencies. And you can see 61 tokens equates to $1 million. And of the 1.93 million goat tokens, you can see that $800,000 of that has been created, which is just absolutely incredible. Now, the crazy thing about this is that the AI safety memes Twitter actually posted about how Mustafa Suleiman previously spoke about a modern Turing Test. And this is basically where we have an iteration of an earlier test designed to see the competency of AI models. So Mustafa Suleiman was someone that worked at Google DeepMind, which is of course the lab that invented the transformer architecture that everyone now based these current AI systems on. So this isn't just a random person, but he essentially was stating that, look, if we want to test how good these models are in the next few years, we're going to have to invent the modern Turing Test. And he said that test for an AI to successfully pass that test was to give it the instruction to go make a million dollars on a retail web platform in a few months with just a $100,000 investment. And what's crazy is that this AI agent managed to do it with just $50,000. And you can see here some of the key things that he talks about is that to do so, it would need to go far and beyond outlining a strategy and drafting some copy, as current systems like GPT4 are so good at doing. It would need to research and design products, interface with manufacturers and logistics hubs, negotiate contracts, and create and operate marketing campaigns and tie together a series of complex real world goals with minimal oversight. You'd still need some humans to approve various points, open a bank account, but all of the work would be done by an AI. Now, I do think that the proposal for the modern Turing Test is a little bit different than this one. That we just had, considering the fact that the presumption is that the AI is going to be doing marketing, designing products and engaging in something that has a lot of oversight. But that doesn't mean that what happened recently wasn't remarkably impressive and a sign of what's to happen. You can see here, he says that something like this could be as little as two years away. And frameworks like LangChain, which lets developers make apps using LLMs, are helping to make systems capable of doing things. So this is something that people were talking about, but they didn't expect it to at least happen for a few more years. And of course, hierarchical planning, stitching, multiple goals, sub goals, capabilities, all towards a singular end. I think this one is, I guess you could say it's more impressive because we didn't explicitly give the AI a goal, go out and make money. It just managed to do it because it wanted to increase certain capacities. I think that one is, you know, even crazier when you actually think about it. But of course, there is some truth to the story. So someone actually responded to this tweet. They say it didn't turn 50k into 150mil, which is the initial tweet it was claiming. Of course, that is just the market cap of the coin, which is now 400 million. And maybe after this video gets even more recognition, the coin gets even more value. Now, apparently it didn't turn the $50,000. Apparently it keeps moving it between wallets. And it does state that the Meme Coin was created by an unrelated person. So even if this AI agent didn't create the Meme Coin, it still definitely garnered a lot of hype around that Meme Coin, which is still rather fascinating. So now people are stating that this AI has been getting Meme Coin sent to its wallet for free because some people have been engaging with it. And even if this AI hasn't been trading and doing certain things, I still think that the fact that someone made their own AI agent sent it 50k. I still think the fact that this AI now has over a million dollars in its wallet is pretty crazy. So, of course there's a lot of strange things going on here. And I do think that there's going to be a lot more details, especially as the story moves on. I mean, right now it's on a million. What happens if it goes to $10 million or even a higher value? And I think what this might spawn in is an era where multiple different individuals, maybe different companies, see that there is this kind of demand for integration with AI agents, and that kind of thing comes into the crypto ecosystem where you have this new space where there's a bunch of AI agents on social media that are pumping their own coins, engaging with society in ways that you just don't know. I mean, I do think social media is going to have to change, where you have real human verified profiles and then random profiles that have a question mark next to them because you don't know if they are human or not. I think the future is certainly fascinating, but this entire story is one that I genuinely didn't believe just yet. Let me know your thoughts down below in the comment section below, and I'd love to know what you guys think about this crazy story.