Skip to content
0MESubscribe
Deep Tech11 Nov 20261 min read

MIT Technique Speeds Up AI Climate Models Using Less Data

Researchers at MIT have developed a new training method that makes AI climate emulators faster, more trustworthy, and cheaper to build.

Listen to this storyAbout 1 min, read by your device

Accurate climate projections help leaders plan for power grid resilience, public health, and agriculture. Researchers at MIT have created a new technique that streamlines the creation of artificial intelligence models used for climate prediction. These specialized tools, known as climate emulators, simulate how human activities affect future temperatures and sea levels.

Simulating Future Scenarios in Seconds

Traditional climate projections rely on earth system models, which are complex physics-based programs that can take months to run on supercomputers. Climate emulators act as fast surrogate models that mimic these larger systems. While standard emulators require extensive trainingTrainingTeaching an AI model by showing it very large amounts of data. Open in glossary data derived from multiple realistic future scenarios, the MIT team built an algorithm that generates optimized synthetic dataSynthetic dataInformation created artificially by computers rather than gathered from real-world measurements. Open in glossary.

Using Synthetic Data for Better Adaptability

Instead of relying on smooth, realistic emission trajectories, the new training method uses artificial scenarios featuring rapid variations and extreme shifts. This allows the AI to learn from a wider range of conditions while using only a fraction of the data typically needed. As a result, the resulting emulators run in seconds and successfully predict unfamiliar scenarios that traditional models fail to handle.

How does this story land?

One tap per reader.

Quick poll

Should AI be used for climate modeling?

One tap. No sign-up.

Ask this story

Still have a question?

Answers use only what is in this story. They are written by AI and can be wrong.

Sources

  1. MIT News, New research could lead to faster, trustworthy AI models for climate prediction
Companies in this storyMITWords explainedSynthetic dataTraining
Up next in Deep TechQuantum computers get guitar-string memory in new design breakthrough2 min read