How Big Is Our Blindspot? Why AI for Science Needs More Than Better Models
Progress in AI for Science rests on shaky data foundations. As models accelerate, shoring up the missing infrastructure becomes even more important, writes Renaissance Philanthropy’s Eirini Malliaraki
Two Nobel Prizes in 2024 went to AI-driven science. What neither citation mentioned and what almost never gets mentioned in the celebration of AI’s scientific triumphs is the data infrastructure that made those breakthroughs possible. The chemistry prize recognised AlphaFold and the protein design work it enabled. AlphaFold was trained on the Protein Data Bank, a publicly funded repository that opened in 1971 and has been continuously maintained for over 50 years. The prizes recognised the models. The substrate on which those models were built had been accumulating for decades. That…
saved by
related reading
- Kalil’s Corner: The Role of Philanthropy in Identifying and Pursuing AI Moonshotsrenaissancephilanthropy.substack.com
- Science Needs AI Data Stocktakesaipolicyperspectives.com
- Most Algorithmic Progress is Data Progressberen.io
- Rebuild Biotech for the AI Era | Benchlingbenchling.com
- Designing AI for Disruptive Scienceasimov.press
- Using X-Labs to Unleash AI-Driven Scientific Breakthroughs | IFPifp.org
- A new golden age of discovery — Google DeepMinddeepmind.google
- Funding the Future of Science: Q&A with Dorothy Chou - Renaissance Philanthropy – A brighter future for all through science, technology, and innovationrenaissancephilanthropy.org
- A new golden age of discoveryaipolicyperspectives.com
- Mission and Vision - Asteraastera.org
- Defensibility in the Age of AIsethbannon.com
- A Stargate for Data - by willdepue - Will DePuewilldepue.substack.com