I believe that the development of transformative AI systems, which can perform most human tasks and thus ~10x world economic growth, is likely within the next 10-20 years. It seems extremely difficult to guarantee that powerful and by-default-uninterpretable AI systems will act in ways that are safe or aligned with the best interests of humans and their institutions. There are enormous incentives to deploy economically lucrative and militarily advantageous AI systems, and I think current international governance practices might fail to curtail the deployment of unsafe technology. In short, I believe humanity's power may soon dangerously outstrip its wisdom.I am interested in solving problems related to "value-loading" or "corrigibility," building institutions for AI alignment/governance research, forecasting transformative AI, and developing robust governance norms for "black-ball" technology. I support Effective Altruism, a movement for solving the world's most impactful, neglected, and tractable problems.