Google says sorry for racist auto-tag in photo app | Technology | The Guardian
http://www.theguardian.com/technology/2 ... -photo-app

Going by the excuse given, does the AI see many gorillas in what must be some seriously dark settings?

I read a rather strange story today that a gorilla in a zoo in Japan has become this kind of bizarre adulation figure for hundreds of Japanese teenage girls. Imagine what sort of image metadata that's generating. Frankly, if you're using globally sourced data to seed an automated selector, pretty much anything is possible.

Yep and it is a computer algorithm that tries to discern objects in photos, not just people's faces. It isn't really a wonder that it will need tweaking before it becomes reliable.

I wouldn't call it a racist auto-tag, this is just an algorithm error. Errors in algorithm aren't racist. (Generally, if they can prove the programmer who wrote it is a racist bigot, that is another matter.)

I'd agree it is far more a demonstration of the complexity of writing that kind of algorithm than anything with a racist intent. However, given the periodic stories that crop up, such as cameras asking if orientals blinked in a photo, or black people failing to be picked up by facial recognition software, I would have thought it would be wise for Google to have been extra diligent in such areas.