An open-source builder is calling out Anthropic CEO Dario Amodei for treating policy outreach like a logic puzzle rather than a messaging exercise. The critique lands against the backdrop of fresh export controls on the company's Fable 5 and Mythos 5 models, which prompted Anthropic to bring in security researcher Katie Moussouris to review a non-public Amazon report and argue the restrictions are disproportionate.
maybe Dario can't really afford to be so clueless about the language of Trump's court. Maybe he just despises them too much. Maybe he *wants* to be defeated, to be freed of his exhausting Democracy Crusade? Anyway, you don't hire Katie to testify as a neutral expert for *this* WH
The builder argues that stacking essays and credentialed expert testimony works better on paper than in actual Washington rooms, where audience reception and framing can outweigh internal coherence.
Moussouris has already published a blog post and helped coordinate an open letter from dozens of cybersecurity figures, but no specific upcoming testimony tied to this episode has been announced.
Many users criticized Dario Amodei's Trump White House testimony as blatant political incompetence stemming from poor communication and misguided motives.
Opus on its creator (Opus always agrees with me)
maybe Dario can't really afford to be so clueless about the language of Trump's court. Maybe he just despises them too much. Maybe he *wants* to be defeated, to be freed of his exhausting Democracy Crusade? Anyway, you don't hire Katie to testify as a neutral expert for *this* WH
@teortaxesTex such a rationalist failure mode to focus only on what you say and not how you say it and how the receiver will hear it
effective communication should be a key rationalist skill and it still seems not to be
maybe Dario can't really afford to be so clueless about the language of Trump's court. Maybe he just despises them too much. Maybe he *wants* to be defeated, to be freed of his exhausting Democracy Crusade? Anyway, you don't hire Katie to testify as a neutral expert for *this* WH
@teortaxesTex "Hm this pink haired researcher with pronouns in bio will surely sway the Trump administration to our side" Either clueless or just being extremely stubborn and trying to lose nobly.
@teortaxesTex He wants a reason to crusade for the left. US AI will have strict political alignment.
@teortaxesTex If you genuinely believed you were building AGI, and doing so in a way that you perceived was ethically, would you entrust it to such a govt?
Wouldn't doing so miss the entire point?
@teortaxesTex It’s because Anthropic has several members of the Biden administration on its board
@teortaxesTex Yeah this is just blatant political incompetence. I genuinely struggle to mentally model someone like Dario…
@teortaxesTex He’s a true believer
…God help us all
@EnnoiaVectra @teortaxesTex This would be the equivalent of walking into negotiations with the Biden admin wearing a MAGA hat. I see no way this wasn’t a purposeful troll unless Dario is completely mentally disabled on anything non tech related, but even then you’d think someone would mention it was stupid.
@teortaxesTex There are precious few (if any) qualified chud-friendly security professionals lmao
@mindingmyps @teortaxesTex Vibes of if one never doubted anything they ever read on LessWrong
@teortaxesTex No one in their right mind hires her for anything serious. This was just EA inbreeding.
@xlr8harder @teortaxesTex I'm skeptical this is a rat thing this sounds more like office politics, of them being the group or person you hired so harder to swap out, and also being 'transgressive but neat' which fails to trigger even mundane people's political realization that it is a bad political move