Hilarious, but wouldn't it be useless when the population you use it on is mostly not criminal? If I had a population that was 90% non-criminal, and the accura

Hilarious, but wouldn't it be useless when the population you use it on is mostly not criminal? If I had a population that was 90% non-criminal, and the accuracy of the face predictor something like 75% (for non-criminal and criminal), wouldn't my predictor that always outputs non-criminal almost always be more accurate?