Google researchers present ERA, an AI system that creates expert-level scientific software by combining a Large Language Model with Tree Search to improve quality metrics

ERA discovered 40 novel methods for single-cell data analysis in bioinformatics that outperformed top human-developed methods on a public leaderboard, and generated 14 epidemiological models that outperformed the CDC ensemble and all other individual models for forecasting COVID-19 hospitalizations.

The system uses a Large Language Model (an AI that understands and generates text) and Tree Search (a method to systematically explore possible solutions) to navigate the large space of possible software designs and intelligently improve a chosen quality metric.

ERA also produced expert-level software for geospatial analysis, neural activity prediction in zebrafish, numerical solution of integrals, and a novel rule-based construction for time series forecasting.