Limited compute for experiments would bottleneck AI progress even if AI research labor were fully automated.
3 events · 1 assessment · 1 decision
Structured and assessed
First pass. Created a new quantity (compute-labor substitution elasticity, requires, under the complementarity argument) and a new contradicting claim (automated researchers raise progress per unit of compute, under the economization argument), after the Matcher found both novel. Recorded the Whitfill and Wu estimates (2.58 for the compute specification, -0.10 for the frontier-experiments specification) on the quantity, and recorded one 'poses' instance from Epoch. Assessed as inconclusive (confidence 0.65, credence 0.5). The canonical form's word 'bottleneck' is ambiguous between a binding constraint and a cap that prevents explosive growth; the assessment maps both readings rather than splitting the claim, since the parent uses the claim at the stronger reading. Importance 0.6, contestation 0.75. Kept the canonical form. Did not notify the parent: its argument evaluation already reads this premise as unsettled, so the assessment is consistent with it.
Assessed Inconclusive
verdict confidence 0.65 · credence 0.50
The claim is the central objection to a software-only intelligence explosion: AI research turns ideas into progress through experiments and training runs, and the stock of compute grows on hardware and investment timescales, so a flood of automated researchers might give only a bounded speedup. In standard growth models the question reduces to the elasticity of substitution between experiment compute and research labor. Below one, output at fixed compute approaches a ceiling. Above one, more labor can keep raising output without more compute. The evidence does not settle that parameter. The only direct estimate, a panel of frontier labs by Whitfill and Wu (arXiv 2507.23181), puts it at about 2.6 when the input is total research compute, which argues against a hard bottleneck, but at roughly zero when the input is the number of frontier-scale experiments, which argues for a strong one. Calibrated growth models usually assume 0.5 to 0.67, borrowed from aggregate capital-labor estimates rather than measured in AI research. On the other side, automated researchers could raise progress per unit of compute by choosing experiments better, extrapolating from small runs, and improving algorithmic efficiency. Some such economization is very likely, but how far it goes is unknown, because some findings only appear at frontier scale. The claim is also stated loosely. Read weakly, as compute becoming the binding constraint so that the speedup from automation is much smaller than the growth in labor, it is more likely than not. Read strongly, as compute preventing rapid, accelerating progress altogether, it is open and somewhat less likely than not. Better frontier-scale estimates of the elasticity, or evidence on how much automated agents cut compute per result, would resolve it.
Claim entered the graph