Google’s AI System Automates Scientific Software Evolution, Outperforming Human-Written Code
Google has developed an AI system that automatically improves scientific software through evolutionary methods. The approach has reportedly produced programs that outperform human-written counterparts in genomics, pandemic forecasting, and other scientific domains.
AI-Driven Software Evolution
Google researchers have reportedly developed a novel workflow that uses artificial intelligence to automatically improve scientific software, according to sources familiar with the work. The system builds evolutionary “trees” of software tools where each “node” represents an individual program whose performance is evaluated against standard benchmarks. Analysts suggest this approach represents a significant advancement in automated software development for scientific applications.
