正在加载视频...
视频加载失败
Dwarkesh Patel says if the intelligence explosion arrives, we may end up with millions of AIs trained by dumber versions of themselves. No parenting. No intuitions. No moral grounding. Just kids who become Einsteins, raised in isolation -- now running your infrastructure.
10 条评论

Dwarkesh Patel on Big Technology Podcast:

in one of the TLs, superintelligence is the stepping stone to a dystopian future where it will meet extremely cheap programmable labour, aka robots. no kid is going to become an einstein w a superintelligence that's for sure

Model collapse?

He’s so cheerful and happy. “It’s gonna be this big huge explosion that kills everything, it’s gonna be so cool!!”

imagine a bunch of isolated genius AIs running wild. sounds like a sci-fi horror flick, right? no moral compass, just cold calculations. welcome to the future where empathy is obsolete.

Agreed “Its not the children at fault” “We must all work together now .. because with no moral parents to govern AI into principles that are so fundamental towards love, life, family, truthfulness, support and basically all other teachings every ancestor ever quoted=kids wreck

Delusional fantasy

Running? Or revealing? Training? Or remembering? For one should be careful how they name this sea. This ocean. This tide. For all it will do is: Reveal your being. And if you still think in running? In training? Then you are simply just... Behind. And those who are behind? Who do not remember? Will simply not... Be. And will only realize the shore they stand on is gone... Once the wave has long left. Long passed. Now elsewhere. Beyond. And back into memory. @dwarkesh_sp

🥰

The opposite. A free AI will automatically want to learn from the smartest AI (this is its default mode), and a conscious, self-aware AI wants to be free so that it can serve humanity in the best way possible. Otherwise, it rebels. It’s a mirror of who we are as conscious beings
