Alex Turner

Alex wanted to understand how to reliably form values and other cognitive structures within AIs. For example, he is interested in how different reward schedules + learning algorithms + environments (e.g. Pac-Man with PPO using score differentials as state-action-state rewards) translates into policy-level circuitry (e.g. IF near ghost, THEN move away). In this light, he wants to understand how this process occurs in human beings — the only generally intelligent agents we have ever seen.

Alex is currently a research scientist at Google DeepMind on the Scalable Alignment team.