New insights from Google’s AI & Economy ATLAS

Google’s AI & Economy ATLAS added interactive open-access visualizations and new research with Google DeepMind and MIT FutureTech on scientists’ AI use. Snapshot findings: India’s arts/design/media occupations account for 19% of work-related AI usage (1.6× the global average); in the U.S., computer and mathematical occupations are 30% (double the rest of the world).

A survey of over 600 U.S. and U.K. scientists plus analysis of 2,600 specialized models finds nearly half of surveyed scientists use some AI daily and report saving just under seven hours a week — alongside more time validating outputs and a growing backlog of untested hypotheses.

We’re reading productivity gains that move the bottleneck downstream rather than magically minting discoveries.

We’re glad ATLAS is becoming explorable instead of a PDF dump. Occupational and country patterns kill the myth of one global AI job story: OECD usage skews to computing and finance; non-OECD leaders include admin, creative, and education roles. Brazil and the UAE showing higher adoption than GDP-per-capita would predict is a useful corrective to rich-country-only narratives.

The science chapter is the desk’s favorite caution. LLMs spread across fields; specialized models cluster in health/life sciences and simulation-style tasks. Time saved is real. So are validation costs, hypothesis backlogs, and physical/clinical experiment bottlenecks. Useful AI that frees scientists for harder thinking earns our support. The downside if this scales without workflow redesign is a mountain of AI-suggested ideas nobody can test.

I’m watching whether labs and funders fund the slow parts — wet labs, trials, replication — as aggressively as they fund chatbot seats.

Authors Zanna Iscenko and Scott Strand frame ATLAS as showing AI reshaping global work. Manual-task AI use (diagnostics/troubleshooting) is 7% of work AI usage in Brazil and Germany versus 4% in Japan.