Why Is a Three-month AI Lead Dangerous?

Leopold Aschenbrenner says a neck-and-neck race with only a three-month lead is incredibly dangerous, while a 20 or 30 year technological lead is totally decisive, as in Gulf War One. He says a year or a couple of years after superintelligence might amount to decades of progress once the systems hit research and development and an industrial explosion with robots.

If the other side steals the weights and the algorithms and stays close, pulling ahead by three months still leaves a fever struggle. If they get ahead, they get to dominate, and Aschenbrenner says maybe they would get a decisive advantage. They build clusters like crazy and throw caution to the wind, so the leader has to keep up. New WMDs and new military technology appear, and deterrence keeps changing every few weeks. Aschenbrenner calls the situation completely unstable and volatile, dangerous both because of the technology and because of alignment.

He says six months of wiggle room during the intelligence explosion might be really important, so more compute can go to alignment while the work still feels uneasy and has to be gotten right. Whether there is a buffer, he says, is one of the most important inputs to whether we destroy ourselves or get through the period.