Microsoft — that’s the name of another company that tries to curb algorithmic bias. For anybody not a secret that in recent time, artificial intelligence is increasingly coming “to go” and starts the bullshit, reflecting your own prejudices, preconceptions, bias and imperfection of our society. The company wants to create a tool that will find and warn people about the algorithms AI, which can threaten them.
Interestingly, Microsoft, which delivers itself as a company that creates AI to uniting people, joining the ranks of Google and Facebook to create some tools for the search “badly brought up” artificial intelligence. However, the new Microsoft biased search algorithms will be able to 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, will continue to create and use, just not as long as before.
To create a true AI that will be fair and reasonable, need to work more on appearance. May bring the independent verification that technical company engaged external experts to review its algorithms and search for signs of bias either in the code or in the data it handles.
The revision artificial intelligence is gaining momentum, and some companies on the development of the AI began to attract auditors who can look at their code. But it also requires that the AI was quite simple, either to the auditor well versed in the code. In the case of more complex algorithms, deep learning and do not succeed.
Another possible answer is the best training for people who actually create the AI, so they can determine their own opinions and prejudices to keep them to yourself, not allowing the algorithm to interpret them as fact. Everyone has implicit biases, and the tech world will benefit from the fact that will help people better understand their own worldview.
Microsoft will capture a biased AI
Ilya Hel