Skip to content

Latest commit

 

History

History
21 lines (14 loc) · 1.82 KB

personal-remark.md

File metadata and controls

21 lines (14 loc) · 1.82 KB

Algorithms don't discriminate, people will

There has been long-lasting debates about the potential negative implications of AI technologies to human society. Some consider it "a double-edged sword", others fear that it could "end mankind".

Recently, a similar work PULSE from Duke university has received mixed criticism on making Barack Obama white:

image

From an academic's perspective this behavior is quite normal: PULSE is not designed for face restoration, it just aim to create possible faces from barely-discernable inputs, and just like any other generative models, it can just approximate the distribution of images it is fed with. Training with faces of Caucasian/Asian/Indian/Afrian, and you will have a higher change getting white/yellow/brown/black samples, naturally. Why some people choose to hold a maliciaous view towards normal scientific behaviors is beyond me.

The true purpose of face renovation

Keep in mind, the goal of Face Renovation is not creating new faces more beautiful than what they really are, but bringing out the best side of people while keeping their identities unaffected. Although mild changes and distortions would be inevitable (this holds for other generative models), our model still acvhieves a much more faithful preservation of individuality.

Presidents, refugees, artists, comedians, victims...the algorithm doesn't discriminate, only people will.

image