I can think of a critical experiment there. Prepare a dataset with blue and green people acting the same way. If AI shows statistically significant misrecogniti

I can think of a critical experiment there. Prepare a dataset with blue and green people acting the same way. If AI shows statistically significant misrecognition, then it's a failure.