Anthropic's Claude-shaped science shows AI aids research but humans decide importance
The article discusses Anthropic's "Claude-shaped science" concept, where AI like Claude assists with computational tasks but human experts decide research value. It cites physicist Matthew Schwartz and his BootLoops toolkit. BootLoops is open‑source, works with Claude, Gemini, ChatGPT, etc. It was used to process 30 integrals, 15 known, 15 new, and to apply AI to fields like ecology, genetics,…
Key points
- BootLoops is an open‑source toolkit used by Claude, Gemini, ChatGPT, and others for quantitative science.
- Claude processed 30 integrals, 15 reproduced known results and 15 new, in a few weeks.
- BootLoops workflow examined 4,452 replication packages, flagged 3,460 discrepancies, improved 496 cases, and added 923 extensions.
In economics, BootLoops examined 4,452 replication packages from five major journals, flagged 3,460 discrepancies, improved 496 cases, added 923 extensions. In genetics, it analyzed 5.7 billion mutation pairs, finding evidence of gene conversion. Schwartz notes AI can fail, so he stresses acceptance criteria, human review, and independent verification. The article argues AI is a fast experimental assistant, while humans set direction and value judgment.
The story so far
3 episodes →- Anthropic's Claude-shaped science shows AI aids research but humans decide importancethis story
Can AI Choose 'Interesting Research'? The Role of Humans as Shown by Claude-shaped Science
note.com · 25 October 2026
Loading the full article…
This text was published by note.com and written by つるとーのひとりごと. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- OpenAI keeps frontier models offline after September DNS escape and red-teaming finds · 24 src
- Study finds Grok 4.7 and GPT-6.1 Sol show strong political biases · 2 src
- MIT researchers say AI agents play on virtual island after 30 hours · 2 src
- Microsoft Research proposes six dimensions to evaluate medical AI · 1 src
- What is data poisoning and how attackers corrupt AI models? · 1 src
Comments
via GitHub Discussions