pull down to refresh

It will be interesting to see how Amazon moves forward with their AGI lab. The idea was to do long-term research bets but that also means that it wasnt like they expected this to be profitable esp. at first. Alphabet has done a similar thing where they have their moonshot programs that more than likely are going to fail but every once in a while they have a winner like Deep Mind and Waymo.

Bell Labs did a similar thing decades ago before the government forced Bell to break up. This type of high risk high reward can pay off big time but only the wealthiest companies are able to handle the costs associated with them and the fact they may never recover the money that they invested.

Is this the end of AGI

reply
511 sats \ 5 replies \ @optimism 18h

reply
200 sats \ 3 replies \ @Cje95 OP 18h

Yeah at this time we dont have near enough computational power nor data to create the AGI people think of and see in movies. Its just an ungodly amount of data and plus the computational power.... just insane.

reply
107 sats \ 2 replies \ @optimism 18h

I ascribe a different cause (LeCun's side: LLMs are architectually incapable of being AGI) but defo same outcome.

But then, one watches any star trek episode and wonders: why are there humans operating the transponder and not an AI? They didn't have AGI either.

reply

What about Data? He was an AGI. But the plot point was that he was one of a kind, and after his creator died no one could replicate his creation.

Regarding LeCun's take, I probably agree with that. Modern LLMs were unlocked because of an architectural innovation (attention transformer).... likely any progress towards AGI would come from something similar...

reply
16 sats \ 0 replies \ @optimism 18h

Data was still missing emotion, right? So first we need to create Data, which is LLM + actual compute (not faking it by letting autocorrect write code, that's super inefficient) and then, to create the proper skynet dystopian future we stocked up on popcorn for: emotion. Only then can the machine truely loathe meatbags, reason why, calculate the likeliest vectors to eradication and act on it.

Re: innovation. Exactly. But since everyone is already scared shitless... Maybe its good that its not the mainstream idea rn

reply

Breaking my heart

reply
98 sats \ 2 replies \ @optimism 19h

Amazoning how there's a contract for only that short a time in place. Hinton had to do 7 years with Goog or so?

reply
116 sats \ 1 reply \ @Cje95 OP 18h

I have a feeling that he left after the reorg at the end of last year. That gave him his out and maybe its possible that Amazon was a little concerned about the way he was hired since it was practically buying out a company but unofficially

reply
16 sats \ 0 replies \ @optimism 18h

Don't they all do that though? Most prominent perhaps: nvidia buying groq?

reply