We present a surprising result regarding LLMs and alignment. In our experiment, a model is finetuned to output insecure code without disclosing this to the user. The resulting model acts misaligned on...

We argue that representations in AI models, particularly deep networks, are converging. First, we survey many examples of convergence in the literature: over time and across multiple domains, the ways...

He’s one of the loudest voices of the AI haters—even as he does PR for AI companies. Either way, Ed Zitron has your attention.

Colleges are racing to make students ‘fluent.’ One problem: No one knows what that means.

Why is it fun to learn the rules of a videogame? Not just to play the game, but to learn it; to travel from the state of unknowing to knowing, to be surprised when a revelation subverts your expectati...

Bjarke A. Larsen's Blog