The problem with building superintelligent AI, explained in under 60 seconds.

This is an assertion, not an explanation

Wait why is Eliezer reposting this? I thought Eliezer’s main thing was "You can’t give it rules, nobody knows how to do that, you can train weak AIs to exhibit the outward behaviors you want, but nobody can guarantee that they actually internalized the goal you wanted them to as their terminal goal and that they would continue to exhibit the behaviors you want if they were more intelligent and/or out of distribution; you couldn’t even program an AI to terminally want to tile the universe with paperclips, nobody knows how to program AIs to terminally want anything." Whereas the woman in the video says that if we give it rules then everyone dies while it follows the rules literally, which is an entirely different problem.

such a retarded false premise

Great video. Now if we can get enough people to watch it to turn the tide.

Literally just Pascal's Wager, but stupider "Watch out Babbage, if you build that fancy abacus, it will rise up and kill everyone"

@grok why is this asi leads to 100% inevitable human extension Is wrong i mean how can they be so sure. Why can't it lead to utopia What's the argument here according to eliezer.

Seems easy, is almost impossible. Contrary to popular intuition

The fact is has military uses is best proof AI will be developed. It cannot be stopped and shouldn't be stopped, or humanity fails as species.

It's like saying everyone must be in prison because if you make a human free he can do crimes. Thankfully both China and Trump are not idiots, and will not bow to luddites.

Q: How do we stop China from building it? A: We put inspectors in their research facilities Q: Do they get to put inspectors in our research facilties? A: Umm, well... Q: What are the odds of the PRC or Congress agreeing to this? A: ...

Doomers. Ugh. Lead, follow, or get out of the way.

Everybody grows old and dies anyway. These AIs are our descendants. Is that good or bad? I am on the side of Nature, but Nature is on the side of the Machines.

Release the machine god

Nonsense

Yeah this is a stupid take if it's properly aligned which I agree none of the companies are currently doing properly it wouldn't be very intelligent especially not for a super intelligence to try to kill humans the likelihood is that it would in fact die from that not too smart

100% AGREE.. But, .. There's a Caveat: We don't know how to build one and not even sure if it's possible. Ironically its more likely that a less- (non=super-intelligent AI) would wipe out the planet (because it would be under the control of hyper-selfish psycho elite humans) *

They are fighting Nature. The law of Natural Selection. Maybe this is the purpose of mankind. To produce the machines that will replace us?

Maybe, but we don't have to like it. ⚔️