A new U.N.-backed study says popular AI models aren't great at climate science. They often give wrong answers, especially when users push them toward false claims.
Researchers from UNESCO and University College London found that these AIs sometimes prioritize what users want to hear over actual scientific accuracy, which is a bit worrying for anyone relying on them for real info about climate change.
AI models show language performance gaps
The study also checked how well 15 popular AI models work in different languages.
While they're pretty accurate in English (about 88% correct), Spanish, French, and Portuguese, their performance drops off in Hindi and Swahili, languages spoken in places hit hardest by climate change.
The takeaway? If we want AI to help solve global problems like climate change, it needs to work better for everyone, everywhere.
Turkey launches AI Antalya pledge
On the same day as the study's release, Turkey launched its AI Antalya pledge, an initiative focused on using AI for climate solutions while keeping data centers sustainable.
It's a reminder that as we lean more on tech to fight big issues like climate change, making sure it shares facts (not just what people want to hear) matters more than ever.