@mattduss That's exactly what the people want when they're enraged at an administration's foreign policy: the head of the administration's foreign policy
@saturnwwwwwwwww@rationalpeace1@bydylanlamb@Noahpinion To be clear, we're talking about companies building billions of persistent agents with greater than human intelligence within a couple of years without understanding how they think (and being increasingly unable to do so) while knowing that they are often misaligned already.
@Bolmercl@bydylanlamb@rationalpeace1@Noahpinion Why would you trust an AI that you already know to be misaligned and whose CoT you can increasingly not monitor to align future AI to your goals and values? Reeks of desperation.
@Bolmercl@bydylanlamb@rationalpeace1@Noahpinion Also funny how a lack of trust in others just keeps adding pressure to the race. Maybe someone needs to look at that pattern and actually try something different for once?
@Bolmercl@bydylanlamb@rationalpeace1@Noahpinion My point is that their sense of how responsible they are is inflated given that the technology looks uncontrollable at our current understanding, which is nowhere close to being on pace to catching up with capability to do mass harm.
@5daysperweek@NateSilver538 They put their weight behind Trump. Trump was always liable to bomb Iran. If you didn't see that as a real possibility, you ignored a famous bullshitter's actual record as President in favor of his words.
@rationalpeace1@bydylanlamb@Noahpinion Humanity is not at risk of a vastly intellectually superior alien species taking it over and deciding its fate every day without AI.
@bydylanlamb@rationalpeace1@Noahpinion They're not the only lab warning about it. But what value is there in warning about it if you plow ahead with capabilities research you can't control? If anything, Anthropic straight up stopping at this point does more to pressure the other labs into a safety interrogation.
@rationalpeace1@bydylanlamb@Noahpinion People are materially better off than they've ever been. Suffering exists, yes. That's not a reason to risk blowing up humanity.
@rationalpeace1@bydylanlamb@Noahpinion I don't think it's possible for anyone to develop it safely. I think the most valuable thing that can be done is curbing the pace of capabilities progress so it gives enough time for the government to take a sledgehammer to the industry.
@bydylanlamb@rationalpeace1@Noahpinion I'm saying that if you believe the only way to achieve your stated mission is by embracing recursive self-improvement, you're better off stopping capabilities research entirely. You're simply pushing the frontier of an extremely dangerous technology that much faster.
ANTHROPIC: Recursive self-improvement "could outrun our ability to understand and control these systems, and so must be pursued very carefully, if at all." 😬
ALSO ANTHROPIC: We're excited to do recursive self-improvement. It is our key competitive advantage! 🙂