@frascafrasca La razón por la cual no es tan buena en eso es que los labs no tienen incentivo en hacerla mejor. También es verdad que software y matemáticas son tareas "verificables" que hacen mas fácil dejar que aprendan solas a ser mejor sin necesitar feedback
@omarsar0 Isn't this mostly consistent with @nickcammarata's essay https://t.co/7faYMGjtSu (in that neurons are a better unit of analysis than the field assumed, and that learned dictionaries can rediscover them)?
@pjreddie You called the US ambassador a hypocrite for condemning violence against some, but not others. I assume you have condemned the other governments I mentioned such as forcefully as the US and Israel.
@pjreddie Yes! I condemn violence against innocent people everywhere. In Israel. In Palestine. In Iran. In Myanmar. In Sudan. In Syria. In China. In Tigray. In Congo. In Saudi. In Yemen.
@pjreddie Of course! By "it" I meant condemning the Israeli Police and judicial system for not persecuting those responsible. But I reserve judgement until it happens, and at least so far they have acted on it.
3 more reached out, 3 more joining the event!
- Eugenia Kuyda, Co-Founder at @wabi
- Ethan Bazarganfard, Founder at Shuffle AI
- Nicolas Loeff, Co-Founder at @ZapiaAI
We've got some exciting surprises at the event as well - you won't want to miss it!
Event link: https://t.co/F8F2V3lO14
Also Top AI alignment researchers right now: "AI researchers are taking up smoking to signal they believe AI will cure cancer" source: https://t.co/5tzE8hS12O
@eli_lifland How do arrive to this figure? "Conditional on AI takeover, human extinction is ~40%". Is it purely based on the "in the weeds discussion of fatalities from AI takeover"? What is your 95% CI on that?