New insights from Google’s AI & Economy ATLAS

Google has launched an interactive, open-access experience to visualize ATLAS data, which tracks millions of global data points on AI adoption. Users can examine AI usage rates by occupation, from electricians to purchasing managers, explore domestic AI applications, and compare adoption levels across countries. The tool aims to make complex datasets more accessible for researchers, policymakers, and the public.

New research from Google, Google DeepMind, and MIT FutureTech analyzes AI use among scientists, drawing on ATLAS data and a survey of over 600 U.S. and U.K. scientists. The study examines 2,600 specialized AI models and categorizes scientific tasks using a taxonomy from MIT FutureTech. Findings indicate scientists use AI at higher rates than many other professions, with nearly half employing AI daily.

The research highlights distinct roles for different AI types in scientific work: large language models like Gemini are widely used across fields, while specialized models dominate in health and life sciences, particularly for prediction, generation, and simulation tasks. Scientists report saving just under seven hours per week through AI, though much of this time is spent validating outputs rather than accelerating discoveries.

The study also identifies challenges in scientific workflows, including bottlenecks in physical experimentation and clinical validation, and an increased backlog of untested hypotheses. Researchers emphasize that while AI boosts productivity, realizing its full potential may require redesigning scientific processes to integrate rapidly evolving AI capabilities effectively.