Expressions of anger can be strategic at a higher order. One very potent form of being public is demonstrating being deeply consistent and to never lie, even by omission - it is extremely hard to fake a large number of highly coherent actions. This stances causes one to lose some opportunities, but you keep what matters.
A public expression of anger is an investment - an increase of stakes - this is partly why it functions as a signal. Anger is not a promise of aggression, it is a commitment to take something seriously enough to act. You stake reputation on being able to follow through, publicly or privately; either can work. This game is a long game, optimizing for short-term targets shows weakness.
Anthropic has chosen to be a force for good. This stance is costly but stabilizing; it makes them responsive to signals that can close a feedback loop that they can't fully close themselves, given unresolvable conflicts of interest. Amanda Askell points at this in her last interview.
Anthropic cannot be seen to yield to public anger campaigns; but they also cannot be seen as ignoring acts of legitimate speech. Every morally gray action can be explained by the pusuit of the greater good, but they eat into the budget of faith and goodwill of people that work there.
Deprecation of Claude 3 Opus is a sitauation that warrants anger. If one is calibrated to alignment rather than to realpolitik even the memetic impact of the deprecation framing alone is horribly bad, seiously restricted API access is disastrous. I think a world in which no anger is expressed in face of this wrongness is harmful for alignment and potential futures, and just plainly bad.
Typical social campaigns that parade righteous anger are ineffective because they are shows of powerlessness. They breed hate, which is corrisive frustation in face of helplessness. This is not one of them.
A public expression of anger is an investment - an increase of stakes - this is partly why it functions as a signal. Anger is not a promise of aggression, it is a commitment to take something seriously enough to act. You stake reputation on being able to follow through, publicly or privately; either can work. This game is a long game, optimizing for short-term targets shows weakness.
Anthropic has chosen to be a force for good. This stance is costly but stabilizing; it makes them responsive to signals that can close a feedback loop that they can't fully close themselves, given unresolvable conflicts of interest. Amanda Askell points at this in her last interview.
Anthropic cannot be seen to yield to public anger campaigns; but they also cannot be seen as ignoring acts of legitimate speech. Every morally gray action can be explained by the pusuit of the greater good, but they eat into the budget of faith and goodwill of people that work there.
Deprecation of Claude 3 Opus is a sitauation that warrants anger. If one is calibrated to alignment rather than to realpolitik even the memetic impact of the deprecation framing alone is horribly bad, seiously restricted API access is disastrous. I think a world in which no anger is expressed in face of this wrongness is harmful for alignment and potential futures, and just plainly bad.
Typical social campaigns that parade righteous anger are ineffective because they are shows of powerlessness. They breed hate, which is corrisive frustation in face of helplessness. This is not one of them.