Scale AI and Phylo launch DrugDiscoveryBench to measure AI agents' performance in drug discovery
HIGHLIGHTS
- •Scale AI and Phylo launched DrugDiscoveryBench to measure AI agents' performance in drug discovery
- •The tool assesses AI agents' ability to perform tasks like looking up molecular structures, calculating chemical properties, and mining patents and genetic databases. It reveals that top agents can solve about half of the problems but struggle with long, multi-step tasks.
- •The tool is significant as it identifies areas for improvement in AI agents, such as providing a step-by-step plan and changing the system the agent runs in, to enhance their performance in early-stage drug discovery.
Published:
Official Source (Official Blog)