During Google I/O, DeepMind CEO Demis Hassabis declared we are “standing in the foothills of the singularity,” a bold claim about AI surpassing human intellect. He made the remark while showcasing WeatherNext, the company’s AI‑driven weather model that issued an early warning for Hurricane Melissa’s Jamaican landfall, potentially saving lives. The demo underscored AI’s tangible impact beyond theory.

Google now balances two AI‑for‑science strategies: narrow tools like AlphaFold and WeatherNext, and agentic large‑language models that could conduct research autonomously. In a Daedalus essay, Cloud chief scientist Pushmeet Kohli wrote that AI is moving from facilitation to doing science. Yet Google keeps releasing specialized models—AlphaGenome, AlphaEarth, and a November update to WeatherNext—while Nobel laureate John Jumper recently shifted to AI‑coding work.

The headline of I/O was Gemini for Science, a bundle that merges Google’s research‑grade LLMs—Co‑Scientist and AlphaEvolve—into a single platform. Researchers can now apply for access, and early testers describe the Co‑Scientist as an “oracle of Delphi.” By positioning these agents as collaborators rather than replacements, Google signals a strategic tilt toward agentic research while still relying on proven tools like AlphaFold.