Today’s picks
1. The AI boom is threatened by an epidemic of unreliable numbers
This article examines how an epidemic of unreliable numbers is threatening the AI boom. It argues that investors can no longer put their trust in the data driving valuations and enthusiasm. The piece is worth reading for a sobering view of the information risks beneath the AI hype.
2. New ‘First Proof’ exam reveals how AI-powered math is a pay-to-play sport
A new ‘First Proof’ benchmark shows that some AI models are becoming highly capable tools for research-level mathematics. However, open-source models still struggle with the same tasks. The results highlight a widening gap that may make advanced AI math access dependent on funding and proprietary systems. The article is worth reading for its implications for research equity and the future of mathematical discovery.
3. Could AI help scientists with their paperwork?
Scientific paperwork — grants, ethics reviews, compliance filings — consumes ever more of researchers’ time, and AI tools promise relief by drafting, checking and filing documents. This piece argues that such gains come with a catch: automation amplifies existing flaws, from box-ticking reviews to diluted accountability and ever-growing document volume. Worth reading as a caution that efficiency tools cannot substitute for reforming how science is administered and evaluated.
4. NIH appears to shut down public comments on science peer-review rule change ahead of schedule
The U.S. National Institutes of Health proposed changing how grant applications are reviewed, a move scientists say would reduce transparency. The agency appears to have closed the public comment period ahead of its scheduled deadline. The early cutoff narrows the window for outside input, making it worth watching for what it signals about openness in science governance.
By publication
The Economist
- The AI boom is threatened by an epidemic of unreliable numbers
- Could AI help scientists with their paperwork?