The haves and have-nots in the digital future.

Are computers free from bias?

Will artificial intelligence be secular and non discriminatory ?

Facebook’s facial recognition research project, DeepFace is
97.25% accurate. Humans can perform the same task only with 97.53% accuracy. That’s a pretty big deal. Wow!

A lot of automation in future will depend upon facial recognition ie. Cars will open, phones will unlock and safety devices will allow you with your face alone.
So it seems at last we have been able to decimate human preference of the class barrier. At least machines don’t recognize race or caste. Is that so ?
… ..
Here maybe the catch as I understand.

In the case of code for facial recognition, the code is limited by the faces included in the training data.
Software made in east Asia is better at identifying east Asian faces, while software made in North America is better at identifying white faces.
Silicon valley’s library is populated more with white male faces. Black people in ghettos are hardly represented. The result is that the technology fails to recognize Joy Buolamwini when she was in China. because she is black who grew up in Mississippi though she has gained a Rhodes scholarship, a Fulbright fellow, an Astronaut scholar and a Google Anita Borg scholar. Interestingly she was picked up when she wore a white mask.

The 97.25% accuracy is heavily skewed when analysed for race, color and sex. For ladies it falls to 77.5%.
In a horrific blunder, in July 2015, Google Photos, which scans photos and then suggests people’s identity, misidentified two black people as gorillas.

Artificial intelligence learns from examples it was trained on. If you do not include many images from one ethnic subgroup, it won’t perform too well on those groups.

It seems the class barrier -the more it changes, the more it remains the same.

# Joy Buolamwini had won a $50,000 scholarship for her work fighting coded discrimination.