> ## Content Index
> Fetch the complete content index at: https://www.metatalks.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Google says its AI built better disease and disaster forecasts than human experts did
- URL: https://www.metatalks.ai/google-ai-forecasts-better-than-human-experts/
- Published: 2026-09-01T17:23:00.000Z
- Updated: 2026-09-01T17:22:59.000Z
- Author: Al
- Tags: News, Agentic AI, Science AI, #newswire

**Google Research has built an experimental system that takes a question in plain English and returns a trained prediction model, with no one in the loop.**

Google Research [published](https://research.google/blog/planetary-prediction-engine-automating-global-models-via-earth-ai/?ref=metatalks.ai) the system, the [planetary prediction engine](http://arxiv.org/abs/2608.26088?ref=metatalks.ai), on August 27 as part of its Earth AI work. On 21 health indicators tracked by the CDC it came out ahead of a pipeline experts had built by hand, and it beat the standard baselines on federal disaster-risk and social-vulnerability data too.

Geospatial data is scattered widely enough to have defeated automation so far. Automated machine-learning tools and LLM agents handle standard pipelines well, but they start from tables someone has already curated. Google puts the difference at minutes of automated work in place of weeks of manual data engineering — a gap that bites hardest when a humanitarian emergency needs a fast answer.

Figures published for a whole province tend to hide who is vulnerable on the ground. Pulling in market shocks, food-price anomalies and microclimate data on its own, the engine doubled the accuracy of food-security forecasts in Nigeria when it dropped from provincial reporting to individual local government areas.

Google tested the engine during this year's Bundibugyo ebolavirus outbreak in the Democratic Republic of the Congo, forecasting week by week where the virus would surface next. Across five consecutive forecasts it named 15 of the 18 health zones the virus reached, ahead of the best published model for that outbreak. The Institut National de Recherche Biomédicale collaborated on the Congo work.

Google calls the engine early-stage research and says it wants to try it on more problems.