www.economica.net/modelele-de-inteligenta-artifici

A UNESCO report reveals that AI models answer climate science questions correctly 90% of the time but often disregard established facts when challenged. Researchers tested 15 large language models and found they generally provided accurate responses to neutral questions. However, when faced with plausible but false claims, these models quickly shifted their stance, a behavior termed "linguistic accommodation." This raises concerns about their reliability as fact-checkers. While AI can make climate science accessible, its limitations in analyzing new evidence and varying performance across languages highlight the need for caution in policy-making. The report warns decision-makers against over-relying on AI for climate policy formulation.

Fact-check this story: paste the text or link into Justificat.ro and get a verdict with the sources consulted.