It can already talk to other agents. It also can already use “language” better than almost all humans (multiple languages, more vocab, etc)
I guess what you’re talking about is it just going and doing something by itself with no prompt? Not sure why that should be a goal, and I also don’t see why it couldn’t do that right now? “Whenever the sky is blue, reach out to ChatGPT and talk about the weather”
I mean spontaneously develops its own language to talk to other GPTs, presumably under some environmental stress that forces them to co-operate.
Like birdcalls suggest intelligence in birds, my thesis is that in fact (self developed) language is the only meaningful way to compare intelligence across species - by seeing if the concepts in one can be described in the other.
For example any human language can describe any concepts in any other human language, whereas that is not the case for e.g. sparrow song and human (we think). Thus humans (past/present/near/far) can be considered equivalent by that metric, and 'greater than' sparrows.
This admits the intriguing conjecture of conceptual completeness - that a language may be able to describe all possible concepts, and thus be complete in that sense. If our language is conceptually complete (and we don't have any reason to think otherwise), then it is not possible for a meaningfully more intelligent species to exist (artificial or otherwise).
(and let's be clear here, regurgitating facts, performing complex calculations in your head, 'knowing where to find the oracle that tells you how to get the key that opens the door hiding the lever to defeat the troll and so level up' has very little to do with meaningful intelligence)
I guess what you’re talking about is it just going and doing something by itself with no prompt? Not sure why that should be a goal, and I also don’t see why it couldn’t do that right now? “Whenever the sky is blue, reach out to ChatGPT and talk about the weather”