Why I think this goes against so much of what Anthropic stands for in AI safety: https://www.interconnects.ai/p/claude-fable-5-and-new-ai-safety

@MediocreInqv2 Don't admit proudly to misleading a population of people who can be doing genuine AI safety research.

A message to Anthropic leadership: You're not special. Making sure AI goes well is a team effort not a "you effort."

@natolambert check this out

@natolambert I understand the frustration, but one of their stated top priorities has been to avoid irreversible diffusion. You want to access frontier capabilities, you’ve got to be inside the tent soon IMO. I think they should whitelist researchers for now.

@natolambert Isn't their whole thing being the only company responsible enough to build AGI and that any other effort (including open-source) is a risk to humanity by virtue of not being run by Anthropic?

@natolambert Genuine question, how would you approach it as compared to what they are doing?

@natolambert Framing safety as proprietary is how we end up with fragmented alignment standards. Who coordinates when every lab runs their own playbook?

@natolambert It's fine that they wanna prevent distillation attacks but knowingly trying to sabotage people trying to learn is very awful behavior!

@natolambert ++++. IMHO, Jensen had the right intuition about Dario and anthropic. I guess now tech bro will look at @OpenAI and @GeminiApp with more sympathy.

@natolambert acting like ur the only ones keeping this thing on the rails is a fast way to derail it

@natolambert @DimitrisPapail Dario hates the CCP but he wants Ant to be the vanguard party to AI revolution. he’s a Leninist

@natolambert Anthropic team really out here like: “Get ready for our Mythos level model… but only for making hair product websites. Real scientific work? Sorry, too dangerous”

@natolambert BAAAAAAAAAAARS

@natolambert open source is, and has always been, the best approach

@natolambert @MediocreInqv2 well said

@natolambert @sama is our last hope.

@natolambert thy are special. they are the smartest and best researchers doing frontier science. youre building tools no one will ever use

@natolambert Seems closed source and hoping a few thousand people are smart enough to harness AI are not going to lead to good outcomes inevitably if they are serious about AI. Just like how communism and big-gov ideologies fail because a small group of humans are not smart enough alone

@natolambert