The Download: coding’s future, the ‘Steroid Olympics,’ and AI-driven science
Tracks an emerging direction in AI-driven scientific research that may eventually affect how agencies procure or evaluate AI tools for research functions.
Key points
- Google I/O featured Gemini for Science, signalling a shift from specialised AI systems toward agentic, LLM-based scientific research tools.
- World models are gaining momentum among leading AI researchers at Google DeepMind, Meta, and World Labs.
- This is a general AI capability and science commentary piece with limited direct relevance to APS governance or policy work.
Implications for Australian agencies
- Monitor Agencies with science or research mandates may want to monitor the development of agentic AI tools like Gemini for Science as they mature toward practical research applications.
Implications are AI-generated. Starting points, not advice — see methodology for how they're framed.
View original source
Copied.
"The Download: coding’s future, the ‘Steroid Olympics,’ and AI-driven science"
Source: MIT Technology Review – AI
Published: 22 May 2026
URL: https://www.technologyreview.com/2026/05/22/1137845/the-download-coding-future-steroid-olympics-ai-science/
MIT Technology Review's 'The Download' covers two AI developments from Google I/O 2026: the announcement of Gemini for Science, which represents a shift from specialised AI systems toward agentic, LLM-based tools capable of conducting research with reduced human involvement; and growing interest among major AI labs in 'world models' that aim to give AI a deeper understanding of physical reality. The piece is commentary and scene-setting rather than policy or governance analysis, and has limited immediate relevance to APS practitioners.
Implications for Australian agencies:
- [Monitor] Agencies with science or research mandates may want to monitor the development of agentic AI tools like Gemini for Science as they mature toward practical research applications.
Retrieved from SIMS, 18 July 2026.