The current scaling race toward AGI is effectively suicidal. We should stop pushing the capability frontier until alignment is actually solved.
Capabilities are advancing far faster than our ability to control goal-directed, superhuman systems. Racing ahead while the safety problem is open is betting the species on an unsolved research question.
Credible alignment progress that demonstrably keeps pace with — or outruns — the growth in raw capability.