@marwood_lennox It's possible. Rejuvenative regeneration.
The problem is a) to learn to do it without causing cancer and other bad things, b) to learn to do it in the brain without losing memories and skills.
But it's all doable. With a stronger intelligence, of course.
@AndrewCurran_ Right.
Eliezer is now firmly in the realm of using arguments as soldiers, unfortunately. At the same time, the risks are all very real. But this does not mean a rationalist gets to push for their preferred policy without doing a careful and unbiased risk-benefit analysis first.
@tszzl Yes, so straight lines mean "no RSI yet". But there is a feeling that some RSI is happening. Are these lines "slightly curved to hint at a future exponential"?
@toddmakurath@hilbertspaess Yes, it's ridiculous.
The only result of this kind of pause will be a very multi-polar race to singularity (much more difficult to make it survivable; we do want a multi-polar reality, but we want a relatively gentle, paced transition)
@yacinelearning In the Singularity all training is instantaneous!
A model just looks at the situation in question and immediately acquires the desired properties :-) "Immediate comprehension of Truth" :-)
@AndrewCurran_ This is not a disproof. The title explicitly says it runs in exponential time (and it's trivial to write an algorithm which runs in exponential time, I bet even the 3 year old GPT-4 can do that, one just checks the result for all combinations of involved variables).