http://x.com/i/article/2082534298226495488
Parallelization constraints could delay a technological singularity
The ability to divide， coordinate， and recombine work has been a missing parameter in models of technological development.
Growth models usually assume that R&D can absorb ever more researchers. But after an explosion in research inputs （or "virtual researchers"） made possible by the automation of R&D， we may face another bottleneck： our technology for dividing， coordinating， and recombining their work. That is， even after AI R&D is fully automated， parallelization constraints could slow an intelligence explosion.
How quickly the explosion proceeds will then depend on how quickly we develop the necessary "parallelization technology".
Parallelization technology has not featured at all in the conventional models of technological development that economists and AI forecasters use. In a new paper， @pawtrammell outlines what changes when we introduce it as a key parameter， and discusses what new experiments would be needed to estimate progress on it.
The value of this parameter determines what happens once R&D is automated. There are three possible cases：
Case #1： If parallelization technology improves at least as quickly as effective research inputs， the conventional result survives.
Case #2： If it improves faster than the technological frontier but more slowly than research inputs， an explosion still occurs， but more slowly.
Case #3： If it improves only fast enough to sustain ordinary exponential growth， automating R&D may raise the growth rate without making it accelerate indefinitely.