I came to AI safety from pure mathematics — I did my undergrad in it at the University of Melbourne — and I now work on safety problems independently.
The reason I work on this at all is roughly the case made in The Most Important Century: there’s a serious argument that the people alive right now sit at an unusually pivotal moment, plausibly among the most consequential who will ever live. There’s an obvious hubris in taking that personally enough to act on it. I work on the technical side because, on my priors about my own aptitudes, it seems like the best bet I have for making real progress on these questions rather than just having opinions about them.