Two AI Science Assistants Succeed in Drug Retargeting Tasks
Nature published two papers on Tuesday introducing AI systems for scientific hypothesis development and testing. Google's Co-Scientist employs a 'scientist in the loop' approach, where researchers guide the system with their judgments. FutureHouse, a nonprofit, trained a system that evaluates biological data from specific experiments. Both systems focus on biological data and straightforward hypotheses, such as drug repurposing. They are agentic, operating in the background by calling separate tools, similar to Microsoft's science assistant. OpenAI is an exception, having tuned an LLM for biology. The systems aim to manage the overwhelming volume of scientific information, not replace scientists.
Key facts
- Nature published two papers on Tuesday.
- Google's Co-Scientist uses a 'scientist in the loop' approach.
- FutureHouse is a nonprofit.
- FutureHouse's system evaluates biological data from specific experiments.
- Both systems focus on biological data and drug retargeting.
- The systems are agentic, calling separate tools.
- Microsoft has a similar approach with its science assistant.
- OpenAI tuned an LLM for biology.
Entities
Institutions
- FutureHouse
- Microsoft
- OpenAI
- Nature