Building AI to accelerate science and improve lives

Google SVP James Manyika’s essay marks a milestone: Google technologies supporting more than 300 languages spoken by 7 billion people—about 86% of the global population—released alongside interactive AI & Economy ATLAS insights. The piece surveys AI for earlier disease detection, flood and wildfire prediction, learning, and helping workers as jobs change.

Language coverage plus economic telemetry is Google staking a claim as civic infrastructure.

Milestones are starting lines; quality and offline access decide dignity.

We’re impressed by breadth and still hungry for depth.

300 languages is a serious distribution fact if quality isn’t collapsing on the long tail. ATLAS as an open interactive look at global AI use could help policymakers—if the methodology is transparent. Health and disaster applications are where useful AI earns its keep.

The harm path: declaring population coverage while low-resource languages get brittle models; economic maps that flatter engagement; and science stories that skip failure modes. Google AI–generated summaries on the page even warn they’re experimental—fine for UX, not a substitute for evidence.

I’m watching language-quality evals on the newest languages added and ATLAS documentation for researchers.

Published September 15, 2026 on Google’s blog. Manyika leads Research, Labs, Technology & Society. Companion pieces cover language and the societal-impact hub.