Optimistic voices see building an AI agent to search Boomerang videos as a step toward AGI, whereas critics mock it as merely adding compute or missing the core point.
Most Activity
@yacineMTB Yes
Could you autonomously search for this using compute?
@yacineMTB Yes, if there's a good enough (low error sim2real) physics environment to test solutions.
@yacineMTB yes. you can. using pufferlib
@yacineMTB yes
@yacineMTB *Something get made Yacine: I bet I could RL that
@yacineMTB It's going to me more of physical simulation of the real world. You will consider wind, physics and quite a few things
@yacineMTB Build an AI agent that does that and we might be one step closer to AGI
@yacineMTB ..... My question is why do you have to search for a boomerang that automatically returns.
@yacineMTB no
@yacineMTB The answer to any question can always be determined by “just give it more compute”
@yacineMTB Wdym by search, you mean vision based? Or wat
@yacineMTB Not easy but possible. The hard part is fast CFD imo.
@yacineMTB I will need more ram
@yacineMTB yes you can just need sim that can handle as real as it gets
@yacineMTB You still haven't figured it out? All problems are search problems. Sutton stumbled onto the bitter lesson without understanding why it happens.
@yacineMTB There are plenty of shape optimization problems using Deep learning