AI Model OneStrata Reconstructs Fossil Record; Suggests 3 of 'Big Five' Extinctions Were Gradual
The Facts
- Palaeontologists trained a machine-learning model, OneStrata, on fossil records, in a way similar to how large language models are trained on text, to reconstruct changes in Earth's biodiversity over hundreds of millions of years. Nature only
- The model's developers say it suggests three of the 'Big Five' mass-extinction events were gradual declines in biodiversity rather than sudden cataclysms, as other reconstructions describe. Nature only
- Shuzhong Shen of Nanjing University presented the research at a Geological Society of America meeting in Denver, Colorado, on 12 October. Nature only
- The model was trained on data from OneStratigraphy, a dataset Shen's team has compiled over two decades, covering millions of fossil specimens from tens of thousands of sites worldwide. Nature only
- The model was developed with colleagues at Zhejiang Lab, an AI-focused research institution in Hangzhou, China. Nature only
Dive deeper · 2 more factsShow less
- In 2020, Shen's team used an algorithmic method on a supercomputer to estimate species appearance and extinction times at 26,000-year resolution, reconstructing a timeline for 11,000 marine species over about 300 million years. Nature only
- Matthew Clapham of UC Santa Cruz said the model could allow extra analyses but he doesn't know if it would be 'Earth shattering'; he noted high-resolution biodiversity reconstructions already exist for specific locations. Nature only
Context
Researchers in China have built a machine-learning model, OneStrata, that predicts when fossils were most likely deposited, using patterns in a large fossil dataset. It avoids the computational limits of earlier algorithmic methods, which required impractical resources to include more species at shorter time intervals.
The developers say the resulting timeline could prompt a rethink of major events in the history of life. Other researchers say it appears to be the most detailed global record so far but may not be transformative.
How the Agent Panel Scored This
Independent AI personas each scored importance 1–10; the story's score is a trimmed mean (highest and lowest votes dropped). Vote spread: 0.
| 2 | Business Analyst | This is a niche academic paleontology research result presented at a conference, with no market-moving or broadly consequential impact, and an outside expert questioned its significance. |
| 2 | Informed Citizen | This is a niche academic paleontology result presented at a conference, with no immediate impact on ordinary people and only a speculative possibility of revising understanding of past extinctions. |
| 2 | News Editor | This is a niche scientific research announcement about a new model reconstructing paleobiology timelines, with limited immediate impact on the general public and only a single source, though it reports a concrete new result. |