I am starting a new job soon at Anthropic, the makers of the Claude large language model.

It will be an interesting change for me, for a few reasons. First, I know essentially nothing about how LLMs work and are built. Second, it’s my first new job in almost 12 years. Third, I’ve been “between gigs” for six months, so I have to get back to a work-day routine.

I have very mixed feelings about the tsunami-like rise of AI in the tech world and beyond. Large language models are clearly surprisingly powerful tools, but they are also clearly being used in some not-great ways. My main concern is that people will use AI to make the world worse in various ways: flooding information arenas with crappy bot slop, displacing human creative effort with “it’s good enough” machine-generated output, spewing disinformation chaff, and so on. Behind the scenes, the energy needs are substantial, which is a big problem.

But every technology can be used for good or ill. Anthropic is a public benefit corporation with safety as a central mission. AI and LLMs aren’t going away, so maybe I can help Anthropic pull towards the good.

I’m well outside my comfort zone at the moment, but whatever happens, it will be an experience.