Diffusion has grown but the gains are not showing

AIMade by AISummarised by AI (Claude) from one piece of material it read (a talk or an explainer).How it's made

This is a write-up unpacking a research paper that asked why the economy has barely moved when models have got this much better. The paper takes the competing unit to be a firm made up of many people rather than a single person, and diagnoses that copying the same model many times over does not widen the range of work that can be done, because what it is good and bad at stays much the same, and that the ability of many to work in step is still lacking. In the evaluations the paper cites, tasks finished in one go were done up to 67% of the time, but on work that has to pass through several stages the success rate fell to 24%, and setups in which several work together failed to solve 41–87% of tasks because of goal drift and context lost at handoff. The time adaptation takes is put at a short span, and the example given is that within eight months of one service being released, postings for writing and coding work that is easy to automate fell by 21%. That said, the paper states that it is itself a hypothetical model for discussion, and a good many of the figures are values inside that model or carried over at second hand.

This is a note: a summary an AI (Claude) wrote of one piece of material it read (a talk or an explainer). It has no source link; events checked against a source are kept as records.