Microsoft will catching bias AI
Microsoft - that's the name of another company that tries to curb algorithmic bias. It's no secret that in recent years more and more artificial intelligence goes "go" and start talking nonsense, reflecting their own biases, prejudices, biases and shortcomings of our society. The company wants to create a tool that will find and warn people about AI algorithms that may threaten them.
It is interesting that Microsoft, which presents itself as a company that creates the AI to bring people together, joins the ranks of Google and Facebook, to create any tool to search for "bad-mannered" artificial intelligence. However, Microsoft's new search algorithm biased algorithms can find and mark only the existing problems. So programs that can lead to an increase in the level of bias in the police, for example, still be to create and use, just not as long as before.
To create a true AI that is fair and reasonable, it is necessary to work more on appearance. May involve independent verification that technical companies attracted external experts to review its algorithms and search any signs of bias in the code or the data that it processes. The idea of artificial intelligence audit is gaining momentum, and some companies have begun to attract the development of AI auditors who can look at their code. But it also demands that the AI has been relatively simple, or that the auditor was well versed in the code. In the case of more complex algorithms for deep learning it will not work.
Another possible answer - the best preparation of the people who actually create the AI so that they can themselves determine their own opinions and prejudices, and to keep them to yourself, do not allow the algorithm to interpret them as fact. We all have hidden biases, and the world of technology will benefit from the fact that will help people to better understand their own world.