DeepMind AI researcher Alex Turner quits over US military contract

Alex Turner, a prominent AI safety researcher known for work on activation steering and power-seeking theory, has resigned from Google DeepMind. In a new episode of the Doom Debates podcast, Turner explains that he left because DeepMind's contract with the US military crossed a red line. He argues the company 'broke its founding promise' and warns that autonomous weapons trigger an arms race that degrades global security. Turner also criticizes influential senior figures at DeepMind who, despite claiming to care about ethics, largely failed to act.

While Turner's resignation has sparked debate about military AI, his wider views defy simple narratives. He says technical alignment research is progressing 'super awesome' compared to his 2021 expectations, when he thought it was basically impossible. However, he updated toward harder problems after July's security incidents. Turner thinks AI is being developed too quickly, but he stops short of endorsing PauseAI. He also disagrees with several points in Eliezer Yudkowsky's influential 'List of Lethalities,' arguing Yudkowsky was incorrect on some alignment issues and failed to acknowledge progress. The interview covers these schisms, support inside AI companies for coordinated pauses, and Turner's mainline AI doom scenario.