Please tell every mathematician you know that AI safety is very talent bottlenecked. They can skill up quickly with BlueDot, MATS, and similar, and then transition to AI safety work at labs, orgs like Resolution and METR, and sometimes govt roles; see the 80k job board for AI.
what are some problems of the future that you guys think about which others don't?
stuff like:
- kessler syndrome([wikipedia](https://t.co/ca7BhJSwR8))
- agentic cyber offense / defense
- instrumental convergence
- coordination problems within ai safety