The unbiased algorithm still has problems.

Fair and accurate creation Algorithms are not impossible. It takes time.

"Actually, it is possible mathematically", Karos which is the face recognition startup CEO Brian Brackeen told me on the panel of Disrupt SF.

Algorithms are a set of rules that computers follow to solve problems and make decisions about specific action plans. That is the type of information we receive, the information people are looking at about us, the work we were hired, approved credit cards, followed by a driver without a driver watching us. Algorithms occupy increasingly important positions in our lives. However, the algorithm starts with the lowest level and raises its own problem to persist through its adaptation. Human prejudice is embedded in these machine-based decision makers.

To create unbiased algorithms, we need sufficient accurate data. The model not only has enough "pale men" but also enough images of people with different race, sex, ability, height, weight etc.

Chiros CEO, Brian Bracken

"In our world, all face recognition is human prejudice, Bracken."And so You are Thought surely Have, he Learning, he As a Child And You are learn he thing And next he learn More And More. What we call right down That Intermediate, right down The fair method is "a thin man". It's very good. Very good, very good to identify someone who fits this category. "

But you are away from thin men by adding women, different ethnic groups etc – "The trust is good enough that it is hard for the AI ​​system to do the right thing," Brackeen says.

But even 100% accurate models have disadvantages. On the professional side, a good example of using a fully accurate algorithm face recognition is to identify quickly using the system at the convention center, to make sure.

Hope you like the news:

The unbiased algorithm still has problems.

#Stay Tuned For More Updates :)

The Compsmag is a participant in the Amazon Services LLC Associates Program, which is an affiliate advertising program designed to offer sites a means to earn advertising costs by advertising and linking to
Compsmag - Reviews, News And Deals!