RT @NeelNanda5: Jack does great work, and I'm very excited to see Anthropic interpretability people taking on fellows - if you're excited a…

An update from Anthropic Interpretability on Model Diffing

RT @NeelNanda5: David is a great researcher. Seems highly worth applying to work with him if you're looking for an interpretability PhD!

RT @AnthropicAI: We’re starting a Fellows program to help engineers and researchers transition into doing frontier AI safety research full-…

RT @JoshAEngels: 1/5: We recently uploaded a new version of our paper "Not All Language Models are Linear," and I thought I would post a fe…

RT @JasonDClinton: We can build the future if we try.

RT @NeelNanda5: There's a fantastic new set of tutorials from @calsmcdougall to learn about Sparse Autoencoders, complete with coding exerc…

@GreyhoundBus automatically changed my seat reservation to a lower-cost reservation weeks after charging me the full amount and didn't refund the difference. It's a small amount but it feels pretty wrong...

@dpkingma @AnthropicAI Welcome!

RT @patio11: The amount of “I was lead to believe this was an economically advanced nation!” these last few months has been significant and…

RT @benskuhn: Talent density at Anthropic continues to increase :)

A collection of brief updates from the Anthropic Interpretability team.

RT @bshlgrs: I asked my LLM agent (a wrapper around Claude that lets it run bash commands and see their outputs): >can you ssh with the use…

RT @mattyglesias: OpenAI’s creators hired Sam Altman, an extremely intelligent autonomous agent, to execute their vision of x-risk consciou…

@StefanFSchubert In the same way that Harvard's endowment significantly understates its actual assets since most of its spending comes out of federal grants.

A collection of small updates from the Anthropic Interpretability team.

@TylerAlterman How did you do it?

@CJHandmer @natfriedman FRO’s?

@uncatherio @michael_nielsen @SaraBollman (One way that could happen that could be kind of beautiful is a standing loop through a reversible computation. So you could simulate a good world over and over for no free energy.)