Image recognition is usually considered a part of artificial intelligence. In image recognition, you need photos with known labels. Just to explore, I wondered what kind of data I’d find if I searched for images of murder suspects. I was motivated by reading that eye witnesses to crimes can often be wrong. Could there be a way to prevent misidentifications?
I searched for images of murder suspects using the query “(city) murder suspect” for 10 random, large cities: Atlanta, Chicago, Cleveland, Detroit, Houston, Miami, Nashville, New Orleans, Philadelphia, Washington DC.
I didn’t see any pattern or anything significant with the exception that some cities had “denser” results — the same person appeared multiple times (Cleveland, Washington DC). And some cities had “dispersed” results — the result images were almost all different people (Detroit, Philadelphia). In short, there was no obvious way (well, nothing obvious to me anyway) to use the image data to improve the legal system.
Moral of the story: Artificial intelligence has the potential to do a lot of Good. But data is just data. In the case of image recognition, it’s just pixels.