I strongly suspect that the maximal possible time horizon is proportional to a power of compute invested, multiplied by architectural tweaks: Thorizon∼Cquality(computespent)α The compute spent scaled exponentially, yielding the exponential trend. If you don’t believe that anyone will ever train a model on, say, 1E29 or more FLOP, then this and the maximal estimate of Cquality might be enough to exclude the possibility to obtain CoT-based superhuman AIs which the Slowdown Ending of the AI-2027 forecast relies upon in order to solve alignment.
I strongly suspect that the maximal possible time horizon is proportional to a power of compute invested, multiplied by architectural tweaks: Thorizon∼Cquality(computespent)α The compute spent scaled exponentially, yielding the exponential trend. If you don’t believe that anyone will ever train a model on, say, 1E29 or more FLOP, then this and the maximal estimate of Cquality might be enough to exclude the possibility to obtain CoT-based superhuman AIs which the Slowdown Ending of the AI-2027 forecast relies upon in order to solve alignment.