"Months to a year for humans": thousands of OpenAI agents sift millions of mouse brain cells in 32 minutes
First seen on X 11 hours ago@IterIntellectus ♥ 4,2251 views
OpenAI says it ran thousands of AI agents in parallel to analyze experimental data from millions of mouse brain cells in 32 minutes, turning up new leads for Alzheimer's and Parkinson's research.
Per OpenAI's blog on Oct 21, the company worked with PerturbAI and Scripps Research using agents powered by its life-sciences model GPT-Rosalind. Each agent took one gene disruption, analyzed the cellular response, searched the literature and proposed explanations. The team estimated comparable human analysis would take months to a year.
The agents flagged that Tsc1 and Tsc2, two genes that work together, affect the cerebellum differently, and pointed to a possible link between cholesterol regulation and neurodegeneration. OpenAI said the agents' novelty ratings matched blinded human experts.
A post on X saying "thousands of agents read ~40,000 scientific papers in 32 minutes" drew 4,200 likes. OpenAI said the results will be used to prioritize further experiments; whether they lead to treatments still has to be tested in the lab.
“> thousands of agents > ~40,000 scientific papers read > 32 minutes > found new potential mechanisms worth investigating for Parkinson-like neurodegenerative diseases wo…”
