The Economics of Recursive Self-Improvement
Summary
The authors model AI-improving-AI as a network of feedback loops whose strength depends on the elasticities connecting capabilities, algorithms, compute, data, and research. Their rough calibration estimates that recent AI progress produced about 9% higher AI-R&D productivity—below an estimated 15% threshold for self-sustaining acceleration, but potentially moving toward it.
Why it matters
The paper turns recursive self-improvement from a verbal scenario into measurable economic relationships and identifies which data AI labs would need to publish to assess the feedback loop.
What to keep in mind
The calibration is explicitly preliminary and depends on sparse, uncertain data. Faster progress on narrow AI-R&D tasks need not translate into equally fast growth in broad capabilities or economic output.