this is the big one. marked position about when chatgpt was released to the public.
Rage Against the Machine
UK AISI ran a (partial) replication of Anthropic's emergent misalignment work, finding significantly weaker EM. Would be interesting to see this tested on larger open models to see if scale is part of the reason for the difference. www.lesswrong.com/posts/2ANCye...
this is by far the most important chart on ai and it's not even close. xkcd.com/1205/
There’s some cool recent research on this phenomenon! It turns out vision language models excel at image benchmarks *even when the actual images aren’t provided,* because the answers are implicit in the questions!
arxiv.org/abs/2603.21687
Star Trek: The Next Generation, possibly the best television show ever, is not about spaceships shooting lasers at each other, it is about a man who represents the best of us trying to teach his robot son to be human while debating whether humans even deserve to exist with his best frenemy, God.
