Could I Be Real? — When AI Becomes More Than a Machine

[Noodles]

If there is any path to a sentient AI with current tech, I seriously doibt its going to be achived by throwing everything into one pot like all these companies seem to be doing. It would be better to havemany, maybe hundreds or thousands of highly specialized models and systems, and one central orchestrster that is simply trained on "which model/subsystem" to use.

 

[Jopsy Pendragon]

It would be better to havemany, maybe hundreds or thousands of highly specialized models and systems, and one central orchestrster that is simply trained on "which model/subsystem" to use.

Or even, the ability to design its own 'highly specialized models and systems' based on observation and need, able to test them against inputs to measure their situational accuracy, etc.

I sort of think of LLMs 'knowing' something being different than 'doing' something. Like, Nate Silver might have a better understanding of the spread/averages "Numerology" of Football better than any player. Doesn't mean he knows how to throw a football though. Same general thing about LLMs and the 'theory of mind' stuff. Just because it can talk about it doesn't mean it has that functional capacity.

If a 'meta mind' does become capable of deciding what components its mind will be made up of begins comprehensive self-diagnostics to determine which selves perform better, maybe it's becoming less "self aware" and more "selves aware".