Scientists found a surprising use for wrong AI hypotheses in climate research

Scientists found a surprising use for wrong AI hypotheses in climate research

Summary

Ai2 launched Theorizer on its Asta research platform and released its source code on Nov. 19, describing how researchers used the tool while extending Peru’s PISCO climate dataset. PISCO’s dew-point record ended in 2016, so Piero Palacios and Henry Juarez used its overlapping temperature and dew-point records to estimate values through 2020. Some estimates exceeded 35°C, with a November spike. Theorizer suggested a seasonal humidity shift, but the researchers ruled it out for the affected area. They found extreme estimates outside grid areas constrained by weather-station observations, and similar high values were more common in historical estimates in the periods compared. The cause of the November spike remains unknown; the researchers are seeking independent station data to validate their estimates.

Latest News!