Researchers at Anthropic have begun to map out the hitherto-unknown internals of AI chatbots, as reported by Ars Technica:

This kind of behavior-related model tweaking is still very rough, the researchers warn, and more research is needed to identify any potential downstream effects of altering features for safety reasons. Even at this early stage, though, Anthropic’s research provides an exciting framework for making an LLM’s “black box” results that much more interpretable and, potentially, controllable.

While a bit geeky, this is solid work – an important read.