red75prime

joined 1 year ago
sorted by: new top controversial old
[R] Google DeepMind: 2.2 million new materials discovered using GNN (380k most stable, 736 already validated in labs) in c/machinelearning@academy.garden
[–] red75prime@alien.top 1 points 11 months ago

Engineers comprise 0.06% of US population for example. Managers around 20%. Also, narrow AI systems aren't so fascinating.

Bill Gates told a German newspaper that GPT5 wouldn't be much better than GPT4: "there are reasons to believe that we have reached a plateau" [N] in c/machinelearning@academy.garden
[–] red75prime@alien.top 1 points 11 months ago

It's called a scratchpad.

And the network still uses skills that it learned in a fixed-computation-per-token regime.

Sure, future versions will lift many existing limitations, but I was talking about current LLMs.

Bill Gates told a German newspaper that GPT5 wouldn't be much better than GPT4: "there are reasons to believe that we have reached a plateau" [N] in c/machinelearning@academy.garden
[–] red75prime@alien.top 1 points 11 months ago

a lot like saying “rocket ships may not be FTL yet, but…”

And the human brain is FTL then?

Bill Gates told a German newspaper that GPT5 wouldn't be much better than GPT4: "there are reasons to believe that we have reached a plateau" [N] in c/machinelearning@academy.garden
[–] red75prime@alien.top 1 points 11 months ago (2 children)

LLMs might still lack something that the human brain has. Internal monologue, for example, that allows us to allocate more than fixed amount of compute per output token.