argumentA line of reasoning from premises to a conclusion. Like a claim, it belongs to no one who makes it.constitution →

The argument from decomposability and the algorithmic proof of concept

forThis argument, if it holds, bears in favour of the claim.constitution →Iterated amplification can usefully train AI systems on complex real-world tasks for which no other training strategy is available.

The argument

Iterated amplification learned toy algorithmic tasks almost as fast as supervised learning from ground truth, which shows that a model can be trained stably to imitate an amplified copy of itself with no external objective. Humans can break complex real-world tasks into small subtasks whose answers recombine reliably, as the performance of human organizations suggests, and learning such decompositions takes far less supervision than learning the behavior directly. A human coordinating copies of the model therefore outperforms a single copy on real tasks, the gap supplies a training signal where no reward or demonstration exists, and the loop that worked on algorithmic tasks carries over to them.

Does it work?

Does it follow?: likely if the premises holdIf the premises are true, the conclusion is much more likely, though not guaranteed.constitution →
Do the premises hold?: 1 doubtful1 inconclusive1 not separately assessed

Granting its premises, the conclusion is probable though not guaranteed: a stable loop, reliable human decomposition and cheap supervision of decompositions leave open only whether imitation keeps pace with the amplified target and whether errors compound across rounds. The argument stands or falls with reliable human decomposition into context-limited subtasks, which is now judged doubtful on the strength of discouraging human-participant experiments, so the argument currently carries little weight. The toy algorithmic result shows the loop can be stable with exact hand-coded decompositions and says nothing about messy real ones. The cheapness of learning decompositions is secondary and unassessed.

Premises

UnassessedNo current assessment. Attention goes where its expected value is highest and someone funds it; nothing has funded an assessment of this claim yet, and anyone can.constitution →Iterated amplification solves toy algorithmic tasks with at most modest slowdown versus ground-truth supervised learning.
InconclusiveThe evidence and argument have been examined and do not come down on either side.constitution →For realistic tasks, learning decompositions likely requires much less supervision than learning complex behaviors directly.
DoubtfulThe evidence weighs against the claim, but the examination is incomplete or the evidence is indirect.constitution →Humans can decompose complex real-world cognitive tasks into context-limited subtasks whose solutions reliably combine into a good overall solution.

Where it is made

Passages where sources make, criticize or mention this argument. They record where it appears, not how strong it is, and none has precedence over another.

Supervising strong learners by amplifying weak expertsmakes itleaves a premise unstatedThe passage relies on a premise the inference needs without saying it.constitution →
As an example, consider the problem of evaluating a proposed design for a transit system. Rather than forcing a single copy of X to reach a snap judgment about a proposed design, we can have copies of X evaluate many different considerations (estimating costs, evaluating how well the system serves different populations, and so on). A human can then decide how to aggregate those different considerations (potentially with help from further copies of X).

Earlier ratings

Nov 11, 2026likely if the premises holdIf the premises are true, the conclusion is much more likely, though not guaranteed.constitution →
Granting its premises, the conclusion is probable though not guaranteed: a stable loop, reliable human decomposition and cheap supervision of decompositions together leave open only whether imitation keeps pace with the amplified target at scale and whether errors compound across rounds. The argument stands or falls with reliable human decomposition into context-limited subtasks, which is unassessed and for which the direct tests with human participants were discouraging. The appeal to human organizations supports only a weaker form of that premise, since teams share rich context and work for long periods. The toy algorithmic result is sound as far as it goes, but its decompositions were hand-coded and exact, so it shows the loop can be stable and says nothing about messy real ones. The cheapness of learning decompositions is a secondary premise and is also unassessed.