DeepMind showed off Co-Scientist - a multi-agent system built on Gemini that generates, debates, and develops research hypotheses. It has already been applied to liver fibrosis, ALS, and cellular aging.
- DeepMind Co-Scientist: a multi-agent system built on Gemini that generates and develops research hypotheses (May 19).
- Three phases with separate agents: generation, debate with a 'tournament of ideas', and evolution of the best.
- Applied to liver fibrosis, ALS, and reversing cellular aging.
It doesn't serve up a ready answer. It partners in the thinking. That's DeepMind's Co-Scientist - a multi-agent system built on Gemini that helps researchers develop new scientific hypotheses, step by step.
Here's the crux of it. The system is a partner, not an oracle. It doesn't replace the scientist - it helps them arrive at more good hypotheses worth testing. And in science the scarce resource is rarely computation. More often it's the good question.
If the machine pours out hypotheses wholesale, the bottleneck just moves down the chain - to the lab that has to test them one by one. More ideas don't mean more discoveries if someone still stands alone at the pipette. We sped up the thinking. Now it all comes down to how fast you can put it to the test.
The system is a research tool; its hypotheses require human verification and real experiments.