The Innovation reviews AI closed-loop scientific discovery systems

A 60-page review published in The Innovation argues that artificial intelligence has shifted from task-specific tools and curve fitting to closed-loop scientific discovery systems that formulate hypotheses, execute experiments, and self-correct. While materials science has closed this loop most effectively, researchers warn that constraints such as data scarcity, a lack of deep network interpretability, physical hallucinations, and rising data center energy demands still pose major challenges. Open article →