LOSING A BIT OF AI PESSIMISM I’ve been pessimistic about the time until we see v

LOSING A BIT OF AI PESSIMISM
I’ve been pessimistic about the time until we see vaguely intelligent AIs. That’s because I understand how the brain does it’s somewhat miraculous job. I felt we’d need to unify the world-model, the langage model, and neuromorphic computing to achieve anything like the human brain.

However, we are seeing the beginnings of properties emerge from language models that we didn’t think were possible. The new releases this week show problem, cause, and effect, mental modeling, and using tools to change state. And if this continues, that’s ‘the hard problem’ right there. The rest is just ‘more of that’. (Recusion and wayfinding.)

And at the rate of discover right now, we’ll know within a year whether those beginnings hold promise or are just another frustrating example of the 80/20 rule of computational evolution – in that rapid initial progress belies the hard problems that prevent that last 20%, and where that last 20% provides all of the meaningful value of the system, without which it’s useless.
#AI