"We find that these datasets are overwhelmingly composed of lighter-skinned subjects (79.6% for IJB-A and 86.2% for Adience) and introduce a new facial analysis dataset which is balanced by gender and skin type. We evaluate 3 commercial gender classification systems using our dataset and show that darker-skinned females are the most misclassified group (with error rates of up to 34.7%). The maximum error rate for lighter-skinned males is 0.8%.
It’s like saying, well, since we can’t successfully prosecute big bankers, well also ignore smaller fry bankers who are sloppier cuz it’s not fair to them.
In the end everyone but the criminals benefit. Eventually one hopes the system is well enough trained so it can be deployed for all skin tones.
I have zero doubt that the second such a thing is deployed, they are going to be cataloging everyone's faces permanently then moving to track people next.
The false positive rate on fingerprints and the lack of understanding in DNA statistics further indicate that the enforcement/legal system has a very poor understanding of even basic statistics.
Instead of police being able to scan one face at a time as they walk their beats looking for suspects, an automated system could scan everyone in view and then alert that an suspect has been identified with xx% accuracy.
Every week on next door someone posts a video still of a vagrant or thief etc. either from a broken car window, hit and run, lurker, etc. Automation would help if the perps. We, as a society, might want to adjust the dials on punishment, given the efficiency, but we shouldn’t give up the chance to minimize these crimes when reasonable.