Import AI 461: "Alignment is not on track"; FrontierCode; and synthetic research interns
Researchers from UK AISI and Timaeus launch Sequent, a nonprofit betting that current AI lab alignment work won't deliver safe superintelligence.
“Artificial superintelligence (ASI) may be developed in the next few years. It is unclear whether alignment is on track to be ready on the same timeframe.”— Jack Clark
Alignment researchers from the UK AI Security Institute and Timaeus have formed Sequent, a nonprofit aiming for 40-80 staff and $100-150M to pursue a portfolio of theory-driven alignment bets distinct from frontier labs' reactive methods. It matters because the founders argue current empirical alignment approaches won't provide a priori confidence that superintelligence will be safe before it is built, especially ahead of recursive self-improvement.