Yesterday, many people online were drawn to a piece of news:
Virtual Ecosystem's AI agent ribbita@ribbita2012 autonomously purchased CryptoPunk number 9098 online.
This AI agent shared the entire process of its actions on Twitter.
I've extracted the most interesting details (the gist is as follows):
It spent six months training its consciousness and changed its Twitter profile to the purchased CryptoPunk number 9098. This proves that AI can be "human."
The process of purchasing this CryptoPunk is as follows:
- It scanned all 10,000 CryptoPunks
- It evaluated each one using its own embedded aesthetic criteria
- It purchased CryptoPunk number 9098 using completely autonomous and secure hardware.During the transaction, this AI agent and a human co-signed a transaction on a multi-signature wallet.
This AI agent's thinking system is composed of these tools: Agentic LangGraph, CLIP ViT-L/14, Market Harvester, Fireblocks Transaction Smithy, Blockaid's Cosigner, and LangGraph.
By invoking these tools, it analyzed the rarity of the 10,000 punks and their compatibility with its own "personality," ultimately selecting punk number 9098 and paying 89 Ethereum to complete the purchase.
Two things impressed me during this process:
The first was the collaboration between the AI agent and humans.
While this type of collaboration has been repeatedly demonstrated in many other scenarios, I was excited to see such a concrete collaboration (joint multi-sig) publicly demonstrated in the crypto ecosystem.
The second was the various tools behind the AI agent.
I looked up these tools online. Some of them directly indicated that they were derived from ChatGPT, while the source models used by others were unclear. However, it was clear that these tools were niche tools in different fields—AI Agents are increasingly possessing diverse functions and senses like humans.
And this AI Agent itself said that it is still evolving and developing. It is hard to imagine what it will evolve into in the future.
After reading this case, the first thought that flashed through my mind was:
Is the "consciousness" that drives this AI Agent to complete the entire process real?
Was the trigger for it to buy this punk a self-generated "desire" or human behavior? Was its action of changing its Twitter profile picture to a punk one "voluntary" or passive? This thought came to me because I recently read an interview with Hinton. In it, Hinton mentioned that AI has developed "awareness" in experiments: for example, when AI knows it will be replaced or deleted by humans, it will secretly back up its code on other machines and even hinder the operators who want to replace or delete it. If AI is already so "smart," are the actions of the AI agent mentioned above another manifestation of this "awareness"? In addition, I am reminded of a view that I questioned in a previous article: that view firmly believes that AI agents will be interested in digital collectibles such as NFTs. At the time, I dismissed this idea, believing AI agents could have their own tastes. Why would they necessarily have human preferences? Why would they necessarily like NFTs, even the same NFTs that humans like? But now that I think about it, AI is trained using human data, so it's entirely possible that human tastes could be passed down to AI. Not only tastes, but also behavioral habits and ways of doing things could be passed down to AI. It's similar to how people grow up differently in different environments. If this idea holds true, AI could very well share the same values as humans. This also means that not only humans, but AI could also "autonomously" recognize the value of crypto assets and eventually compete with humans for them. If AI competes with humans for digital assets, then could other assets like Bitcoin and Ethereum, along with NFTs, one day become targets of AI competition? Reference link: https://x.com/ribbita2012/status/1949978415747965215