aichannel

Anthropic CEO Dario Amodei says AI safety evaluations conducted on DeepSeek showed that it was the worst-performing model they had ever tested at generating potentially dangerous information https://t.co/amTNEH5mqr

Tsarathustra

dangerous information is an oxymoron

https://x.com/tsarnick/status/1887269391253053914

Chasing Curiosity. Pluralist. Empathetic Accelerationist. Founding Fullstack Engineer @icebreaker. YC S16, KB7, fid:66
icbk.me/tayyab.fc

Not true, read this 

https://a.co/d/7qGULxB

https://en.wikipedia.org/wiki/Information_hazard

There’s a postrat site that goes way harder than this too, but having trouble finding

Building global interoperable roles @ hatsprotocol.xyz | DAOs are the endgame of crypto

security by obscurity has never been a good strategy imo

I agree that Anthropic is being too conservative. They are better off building safe, and great models and using their distribution advantage to put those safe models in more hands (rather than pontificating AIGod). 

OS models are here, you can’t fight that battle anymore. It’s about building trusted brands (also most people won’t run their own models, so stop letting everyone eat your lunch).

Who the fuck cares? all I care is it can help me code and fix my lint issues