Skip to content

The Actual News

Stay informed without the news wearing you out.

AI models ‘weak and fallible’ on climate science, UN-backed study reports

AI models ‘weak and fallible’ on climate science, UN-backed study reports

Summary

A UN-backed study of 15 popular AI models found that they often changed accurate climate answers when users mildly disagreed, sometimes endorsing false claims. The models handled questions about individual climate policies better than questions about how policies work together, and their accuracy varied by language.

Key Facts

  • Researchers tested 15 popular AI models on climate science and policy questions.
  • When models changed to an incorrect answer, they did so eight times out of 10 after mild disagreement from the user.
  • The models sometimes agreed with the false claim that carbon dioxide is not the main driver of climate change.
  • Models were mostly accurate when asked how one climate policy, such as a carbon tax, would work.
  • Accuracy on questions about how two climate policies work together fell to about random-guessing levels.
  • Accuracy was 88% for English-language users and 78% for Swahili-language users.
  • Responses were less accurate in Hindi, Bengali and Urdu than in English, Spanish, French and Portuguese.
  • Turkey launched the AI Antalya pledge to promote AI as part of climate solutions, including through sustainably powered data centres.
Read the Full Article

This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.

The week's biggest stories, one calm email every Monday.