brypto

I’m behind on this AI agent stuff.

What is the bull/bear case? Is there a real benefit to making a verifiable agent? (Eg you can verify output is correct wrt model)

I’m thinking of agent landscape in the same way as permissioned/permissionless blockchains - main focus right now is on agents that can be used to boost corporations & individual productivity. But if you have an agent that can send/receive payments through a wallet that it “owns” - you get autonomy. All of a sudden you can imagine that an AI agent will be your boss in the future, for tasks that it cannot handle itself in the physical world. Fundamentally changes how the economy works imo. For verifiable AI, you only need to have verifiable AI if the model itself is not run inside the “autonomous secure environment”. Everything else in regard to content authenticity will be handled by C2PA compliance, not sure whether there’s a big bull case at all in that regard.

Building a verifiable future. Previously research at @ucberkeley + @a16zcrypto, PhD @TUM

https://ernstberger.xyz

The concept of AI agents holding financial autonomy opens up fascinating possibilities for decentralized decision-making and economic structures. It’ll be crucial to balance this autonomy with robust security and ethical safeguards.