And Why Not Alexo? He had to hire, he was fired. Turned off dishonorably. It was the algorithm used by Amazon to skim the deserving in the midst of oceans of curriculum, it turned out that penalized women: it assigned low scores to those who had attended women's colleges plus other assorted discrimination based solely on their gender. The word "woman" weighed as an unfavorable element. The story emerged a year and a half ago and the e-commerce giant, in addition to ensuring that it had dismantled the program as soon as it discovered the error, made it clear that it had never been decisive to evaluate the candidates.
The example reinforces a legitimate concern: instead of removing prejudices from important decisions, artificial intelligence could automate them. Normalize them, perpetrating and amplifying them. A recent United Nations report recalled how Siri and the main vocal assistants (Alexa, Cortana) imagined how "she" express servility and respectfulness in front of our commands. And therefore, we read in the first pages of the analysis, "they represent a powerful example of the gender biases carved in technological products". While a joint study by the Mit Media Lab in Cambridge and Microsoft found that important facial recognition systems have higher error rates when compared to images of women. "More than once it happened that a computer did not answer my voice, instead of that of my partner yes. It depends on two factors: the majority presence of men in the design of technology and, above all, the fact that programs are given information that incorporates gender discrimination, "explains Emmy Rintjema, data scientist and expert in artificial intelligence.
The famous "machine learning", the ability of machines to learn and then express autonomous assessments (as in the case of Amazon's digital recruiters), rests its deductions on the data it examines. Often not neutral. A reasonable sensation, confirmed by another experiment: the Seattle Institute Institute observed that AI, in the photos, tends to identify a man's figure as a woman if he is portrayed in a kitchen, because he has digested tons of female figures in the kitchen. And therefore he associates them with certainty in that context. It is the resistance of an ancestral stereotype, one of many, which is likely to be reflected in the overall Weltanschauung of artificial intelligence, which will hold many reins of future life in hand. And it could show male-dominated attitudes. If you don't want to be pessimistic, you can try to reverse the perspective: "We are on time. We must pay close attention to these aspects when building and feeding an algorithm. The fact that the debate is alive in the scientific community is a starting point ", Rintjema reasons. If it is true that the world changes according to the preconceptions with which we observe it, that prejudices are like glasses, artificial intelligence deserves at least intelligent ophthalmologists.
Opening: "Error 404" 2017, Sammy Slabbinck, Belgian artist known for surreal collages in which he ironically combines modern ideas with images of the past.
Source link
