julia
@mooncat_is
Being weird.
Currently research @anthropic. Formerly @openai and YC S15 and many other things.
Bad takes are my own.
1.7K Following 6.2K Followers
@rayansadri This is generally why SF tech is so far ahead: people move a lot and everybody learns from each other, so every SF eng/pm/etc is the aggregate of many life experiences.
This was less strange in an era where nobody cared about secrets.
Show more
An unfortunate consequence of rampant distillation from Claude is that when we screwed up its ability to speak like a normal person earlier this year the entire ecosystem went down with us.
When I started working in AI three years ago, we were trying to classify text with GPT-4T on 32k context windows.
Now we have autonomous swarms launching nation state level cyber attacks.
Things need to slow down.
Show more
I often times tell candidates that my favorite part of working at Anthropic is the essay culture. It’s not just that Dario writes like this, it’s that folks regardless of level or tenure have exchanges like this, and these public debates are what create our internal consensus.
Show more
To crystallize my objections to the open weight letter:
“Open weight models, on the other hand, allow a broad community of researchers and developers to examine their behavior, identify vulnerabilities, develop safeguards, and improve them over time.”
This is just not true!
Show more
