π¬ AI needs clean data to turn carbon into fuel
Researchers showed that consistent experimental data is vital for training AI to find new fuel catalysts. Four labs at SLAC, Pennsylvania State University, Stanford, and UC Santa Barbara tested a carbon monoxide-producing catalyst. The round-robin experiments, conducted around the same time, initially showed varying results for carbon monoxide and methane production. The teams found that minor differences, like how the mixture was stirred, caused the data inconsistency. Standardization across labs is now being recommended to boost reliability for future AI models. π¬