We have a referral bounty! If you refer a candidate we accept as a participant or TA, we’ll send you a $100 Amazon gift card as a reward.
Apply or refer a candidate here: https://t.co/zrS189mJgv
Applications are now open for CAMBRIA, CBAI’s crash-course in technical skills for AI safety!
• Dates: Jan 4-22, 2027
• Application closes: November 1, AoE
We provide:
• Housing, meals, and travel support
• Office access, TAs, and a cohort of dedicated peers
We are looking for:
1. Participants who are excited to develop technical skills to further their AI safety careers
2. Teaching assistants who have experience with the ARENA curriculum and a passion for AI safety
3. Referrals for either of the above!
@captgouda24 But the post isn't really describing a Pascal's Wager situation, and discusses probabilities like 1% or 5%. He says if you're very confident in atheism, the argument is less strong. So I don't think the post makes this error.
@vgman94@So8res If your position is that we should treat instances of AI agents in evaluation as death row prisoners who will take adversarial action against you given the slightest chance, rather than as students... then you probably agree with Nate on everything important.
@GWHayduke97@CAIS The lie is presenting everything on the right as the single EA perspective, when most are highly contested. "Some EAs think" does not mean it's fair to say "this is the EA party line." Some EAs agree with every point in the left column!
@JordanSchachtel Notice that this is neither a plan to protect chickens nor remotely the same claim you posted up thread.
(Killing off predators is definitely a controversial EA idea, not many EAs endorse it afaik, if you'd led with that you would have gotten a different response!)
@JakeKAllDay@mattyglesias First panel should be exponential or hyperbolic, which are much faster than quadratic. Where do people claim quadratic growth?
Everyone agrees growth slows at some point (e.g. before you've used up all of earth's resources), but that's not much comfort unless we know when!
I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks.
Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon.