TL;DR
Google Research scientists are leveraging ERA to enhance disease forecasting, solve cosmological equations, monitor CO2 emissions, and explore new scientific solutions. These efforts demonstrate AI’s expanding role in scientific discovery.
Google Research scientists are actively applying Empirical Research Assistance (ERA) to real-world scientific problems, including disease forecasting, cosmology, climate monitoring, and fundamental physics, demonstrating AI’s growing role in scientific discovery.
Since its introduction in September 2023, ERA has been used by Google and academic collaborators to test its capabilities across diverse fields. In public health, ERA now predicts U.S. hospitalizations for COVID-19, influenza, and RSV, with weekly forecasts submitted during the 2025-26 flu season, showing performance at or near the top of public leaderboards. In cosmology, ERA combined with GPT-5 and Gemini Deep Think has derived six solutions to model gravitational radiation from cosmic strings, addressing a longstanding open problem. In climate science, ERA developed a neural network that uses data from geostationary satellites to estimate CO2 concentrations every 10 minutes across the globe, offering unprecedented resolution. These efforts are progressing toward practical tools that could improve scientific modeling and analysis.
Why It Matters
This work highlights AI’s potential to support scientific research by providing models that are more interpretable and applicable to real-world data. Improvements in disease forecasting can support public health responses, while advances in cosmology and climate science contribute to a better understanding of fundamental processes. Making such tools accessible could facilitate scientific progress and address global challenges such as health crises and climate change.
Artificial Intelligence Applications in a Pandemic (Smart and Intelligent Computing in Engineering)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
ERA was introduced in September 2023 as a tool to generate expert-level empirical software across scientific disciplines. Its initial applications included solving benchmark problems in biology and neuroscience. Since then, Google researchers and collaborators have expanded its use to epidemiology, cosmology, and climate science, testing its capabilities in real-time forecasting and complex mathematical modeling, marking a step toward practical AI-assisted scientific discovery.
“ERA is enabling us to generate solutions to complex scientific problems that were previously intractable, with real-world applications already emerging.”
— Google Research Team
“Google’s forecasting tools have been performing at or near the top on public leaderboards, indicating strong potential for public health applications.”
— Nicholas Reich, biostatistics professor
The fundamental theory of cosmology: A Book of cosmology science (physics)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
While ERA has shown promising results in forecasting and modeling, it remains under active development. Its scalability, robustness across different scientific domains, and long-term reliability are still being evaluated. Additional validation and testing are necessary before widespread adoption can be anticipated.
GeoAI for GIS Professionals: Become Job-Ready with Python, Satellite Data & Real Projects
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
Google plans to continue refining ERA’s capabilities, expanding its applications across additional scientific fields, and preparing for broader access for researchers worldwide. Future efforts include integrating ERA into operational systems and conducting comprehensive validation studies.
Minut M3 Sensor | Noise, Occupancy & Climate Monitoring in Real Time, for Short Term Rental Hosts | Subscription Required, 12 Months Included
COMPREHENSIVE MONITORING FOR SHORT-TERM RENTAL HOSTS: Minut detects noise disturbances, guest smoking, tracks indoor climate (temperature, humidity, and…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How does ERA improve disease forecasting?
ERA uses AI models trained on historical data to generate real-time forecasts of hospitalizations for diseases like COVID-19, influenza, and RSV, often outperforming traditional methods and supporting public health decisions.
Can ERA solve complex cosmological equations?
Yes, ERA has been combined with advanced language models to derive solutions for modeling gravitational radiation from cosmic strings, addressing a longstanding open problem in cosmology.
How does ERA contribute to climate science?
ERA developed a neural network that analyzes satellite data to estimate CO2 concentrations globally every 10 minutes, providing high-resolution, real-time monitoring of greenhouse gases.
Is ERA available for widespread use?
ERA is currently in active testing and development, with plans for broader release in the future, pending further validation and refinement.
What makes ERA different from previous AI models?
ERA emphasizes generating expert-level empirical solutions that are interpretable, mechanistically accurate, and applicable to real-world scientific problems, moving beyond black-box modeling.
