Go through random and populate this
Sycophancy update:
- Recently, I’ve been hearing from more crazy people. OAI released it’s sycophancy update. We’ve been getting more people in my work inbox telling us their theories of everything, or solutions to the alignment problem, or just panicking, wanting to send us stuff by hardcopy, since they might be watching. The emails are usually, except the most desperate ones, are clearly written by LLMs, the three part lists of {adjective, noun} neologisms give it away.
- To be clear, these people are wrong. I don’t know if they were crazy before they started talking to the chatbot, but they’re crazy now. But are they just early?
- When the going gets weird, the weird turn pro.
Mechanize: two guys from epoch announce cartoon villain goal of automating all human labour
Vct release: models are better than 94% of expert virologists