The Forever AI Agent

In this discussion, I'll review my research and others, then combine them all into a system trying to meet the objective of a true forever agent.

I do not claim to have completed my work. I only argue that this is the path to our shared goal. If you're interested, I would like your support and thoughts.

I'll make this simple. The definition of agency in AI models is identical to that given to natural systems. A mouse is a perfect mouse agent, forever a mouse, forever following the task of a mouse. Humans do this, but on arbitrary tasks, software engineers do not forget their primary tasks, and they certainly don't work so long that they forget they are software engineers.

We are building agent systems in a near-impossible manner to achieve this infinite agentic task described above. We are following the hard path of this. Let's talk about two research projects I built to frame this.

I won't explain these as we have a lot more to go, and my other writings have done that already. I wrote two articles to help you better understand these ideas.

Leave a Comment