Dear MEL Topic Readers,
Row over AI that 'identifies gay faces'
What, how much, and how well does AI do? Collecting and examining data, identifying the similarities and common patterns, and then making judgments are all part of what AI does. If sufficient and reliable data is provided, AI probably makes decent analysis and practical judgments, and it improves the accuracy as it learns more. But are self-posted photos on a dating web site credible data for AI to determine the person’s sexual orientation?
A facial recognition experiment conducted by Stanford University sparked heavy criticism from LGBT rights groups. They say the attempt to distinguish between gay and heterosexual people is dangerous and junk science. The researchers responded that scientific findings should be criticized only by scientific data and replication.
So far, the algorithm hasn’t shown enough accuracy to determine the sexual orientation of sample photos. But considering the rapid improvement that AI-Go had demonstrated to beat the human champions, the day when AI can tell the qualities, preference or even intelligence of a person may not be so far.
At least, it seems more reliable than phrenology, doesn’t it?
Enjoy reading and thinking what role AI will be playing when you get your next smartphone.
No comments:
Post a Comment