Counterpoint on SL5: a leaked AGI could reach China in days and turbocharge the race
boden_moraski · x · 2026-11-20
Boden Moraski pushes back on SL5-level security skepticism: if a US lab built AGI, it could leak to China (and maybe Russia or NK) within days to weeks — hurting US national security, pressuring labs to race harder as Chinese models approach parity via RSI and distillation on the stolen model (compute-limited perhaps), and giving less safety-focused states a real shot. He calls "burning the lead doesn't matter" one of the most harmful memes since 2023, mostly downstream of Leopold Aschenbrenner's worldview uncritically absorbed by the AI space.