down among the unclefts

whoa

“These safeguards prevent foreign adversaries from using our most capable models in ways that pose severe safety risks. The US and its allies hold an edge in frontier chips and the highly optimized software that runs them at full potential,” the company said in a statement to WIRED. “These safeguards ensure Claude isn't used to erode that advantage—by optimizing chips developed by those adversaries, for example.”

and now Anthropic's noncitizen employees can't use Mythos! Which isn't to say that there is a direct causal link there or that Anthropic is responsible for the Trump administration's actions, but one might hope that it would make Anthropic warier of that kind of framing in the future.

no, i must be exaggerating--

A Board was selected to be unremovable and in charge of moderating a potentially malicious, manipulative agent so it didn't get out of control. The unaccountable part was so the agent could not merely go around them, and the board was specifically selected for their rigor and principles, rather than political acumen.

Then the board said Sam Altman was a danger and fired him. They were not rapidly forthcoming with a reassuring explanation, while Sam's allies and financial dependents (Microsoft) raised hell about what a costly decision this was.

Rationalists at the time reacted by... mostly saying they couldn't make sense of why the Board did this. It was such a sloppy move and wasn't accompanied by the sort of full court press we expect of politicians. Because that is not what the Board was selected for. They were selected for being ideological trustworthy, and NOT smooth manipulators.

Sam won, and got rid of them, and given all the money involved that was probably fait accompli. But there was a significant absence of tech and thought leaders trying to say "this is exactly what we hired the Board to do and maybe we should trust them."

Anyway, years later when OpenAI's nonprofit governance is a flaccid joke, people realize "sama" really was a money-focused sociopath who doesn't care about AI Safety principles, and private reasoning from the Board has leaked out, and most of us have concluded they were right. But the only acknowledgement of that I see is bigtime rationalists saying "well why didn't the Board explain it to us better at the time." It's not their fault for being suckered by the smoother politician and all the financial interests, it's the fault of the institution you decided to distrust.