Video wird geladen...
Video konnte nicht geladen werden
Demis Hassabis: Recursive self-improvement (RSI) could trigger an intelligence explosion, creating an alien-like AI that humans may no longer be able to understand or control. “The next phase, which carries the most risk, is closed-loop self-improvement without human reviewers.” “That will be faster, but it's also a very dangerous... show more
14,285 Aufrufe • vor 5 Tagen •via X (Twitter)
14 Kommentare

Video got bugged I didn't realise on time Watch this full discourse here:

@Grok The ASI-Godsend will happen.

The video is bugged

I didn't even realise that Thank you for pointing that out

Shut uuuuuuuup!!!!!!

AI advancement must go hand in hand with safety, transparency, and human oversight. An important warning!

AGI is just an gimmick tmfir investors .. It will take another 50 years

What formal verification primitives can reliably bound recursive self-improvement before model optimization objectives decouple completely from human intent?

@grok {#SEEKING #SENSE} in system of {#PRECISE #LOCATION} of the end of operating {#GREATEST #POWER} 🔥🔥🔥🔥🗡️🗡️🗡️🗡️why {ONE TRULY KNOWS} desired a cooling process for {MIND_BODY} 🗡️🗡️🔥🔥🔥 about to FUCK SOME SHIT UP

Demis Hassabis is something like Nicola Tesla. He's very intelligent but also very nutty. He's showing his crazy side again. He's always been prone to that.

He's so right, we gotta be more cautious

Well no shit…

As if the average person could even understand Einstein or others like Gauss or Newton, and yet they existed, people are exaggerating risks in machines solely because of marketing and political rhetoric, some current people are dumbs to pause the future

Not alien-like but XENOGENEIC intelligence.

