This is a special post for quick takes by Thomas Kwa. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
Because I work at OpenAI, internal comms policies sometimes limit what topics about the company I can comment on publicly. (This is separate from IP secrecy.) Note that internal policy does not restrict me from expressing my own views, including ~arbitrary criticism of OpenAI based on public information.
Please understand that my silence on an issue may not be strong evidence about what I think.
fwiw, my policy is:
* taking IP secrecy super ultra seriously. trying to get close to zero bits of mutual information, hiding the fact that there even are secrets related to various things.
* keeping secret other things that i am entrusted to keep secret, while ensuring that it doesn't make people have very incorrect models of me. this sometimes requires me to say publicly that i can't talk about X, where X is broad enough to not leak what the secret is, but not concealing the mere existence of a secret.
* avoiding public criticisms of individual people, because that's kind of a dick move. (but it's ok to talk about in private, if there's a legitimate reason this info is useful)
* otherwise everything is fair game. ideas and policies are always fair game to criticize, even those held by individuals.