5 comments

[ 2.9 ms ] story [ 26.3 ms ] thread
Open to suggestions and criticism. Mainly curious about things I've potentially missed.
Cool project. I would build a virtual world for the AI to learn inside it, then the AI.
Yes, that's a good idea. I plan to try out the gym from OpenAI (https://www.gymlibrary.dev/) and the simulated environment from DeepMind for it. If you know of any other tools, let me know. I'm curious.
Just scanned this and subscribed to see more. Always interested in projects like this. I have just a couple of thots off the top:

* Prediction means anticipation - anticipatory systems (googles well). In the lowest animals, anticipation appears to come from innate behaviors - biochemistry; in higher animals, anticipation comes from models, some innate, mostly learned.

* A fun thing to think about is thinking itself, some of which is called "reflexive thinking"; that topic has been hot in the CS community since - at least - when Jon Doyle wrote a thesis proposal about "Reflexive Interpreters" in 1979, and remains a really tough nut to crack. But, without it, well, maybe a mind just isn't what it could be.