I like this post but I think it misses / barely covers two of the most important cases for optimism.
Frontier LLMs have a very good understanding of humans, and seem to model them as well as or even better than other humans. I recall seeing repeated reports of Claude understanding its interlocutor faster than they thought was possible, as if it just "gets" them, e.g. from one Reddit thread I quickly found:
"Weak methods" means confidence is achieved more empirically, so there's always a question of how well the results will generalize for some new AI system (as we scale existing technology up or change details of NN architectures, gradient methods, etc). "Strong methods" means there's a strong argument (most centrally, a proof) based on a detailed gears-level understanding of what's happening, so there is much less doubt about what systems the method will successfully apply to.
Subcortical reinforcement circuits, though, hail from a distinct informational world... and so have to reinforce computations "blindly," relying only on simple sensory proxies.
This seems to be pointing in an interesting direction that I'd like to see expanded.
Because your subcortical reward circuitry was hardwired by your genome, it's going to be quite bad at accurately assigning credit to shards.
I don't know, I think of the brain as doing credit assignment pretty well, but we may have quite different definitions of good and bad. Is there an example you we...
Two points:
Hm, I wouldn't have phrased it that way. Point (2) says nothing about the probability of there being a "left turn", just the speed at which it would happen. When I hear "sharp left turn", I picture something getting out of control overnight, so it's useful to contextualize how much compute you have to put in to get performance out, since this suggests that (... (read more)