IBM wants to help companies mitigate the chances that their artificial intelligence technologies unintentionally discriminate against certain groups like women and minorities. The technology giant’s tool inspects AI-powered software for unintentional bias. It makes decisions, like when a loan gets denied to a particular person, explained Ruchir Puri, the chief technology officer and chief architect of IBM Watson.
IBM Debuts Tools Prevents Bias In Artificial Intelligence
The technology industry is increasingly combating the problem of bias in machine learning systems. It is used to power software that can automatically recognize images in pictures or translate languages. A number of companies have suffered a public relations black eye when their technologies failed to work as well for minority groups as for white users. Researchers discovered that Microsoft and IBM’s facial-recognition technology could more accurately identify the faces of lighter-skin males than darker-skin females. Both companies said they have since improved their technologies and have reduced error rates.
Artificial Intelligence bias tools
Researchers have pointed out that some of the problems may be related to the use of data sets. It contains a lack of diverse images. Joy Buolamwini, the MIT researcher who probed Microsoft and IBM’s facial-recognition tech (along with China’s Megvii), recently told Fortune‘s Aaron Pressman that a lack of diversity within development teams could also contribute to bias because more diverse teams could be more aware of bias slipping into the algorithms.
In addition to IBM, a number of companies have introduced or plan to debut tools for vetting AI technologies. Google, for instance, revealed a similar tool last week while Microsoft said in May that it planned to release similar technology in the future. Data crunching startup Diveplan said at Fortune’s recent Brainstorm Tech conference that it would release an AI-auditing tool later this year while consulting firm Accenture unveiled its own AI “fairness tool” over the summer. It’s unclear how each of these Artificial Intelligence bias tools compare with one another because no outside organization has done a formal review.
Puri said IBM’s tool built on the company’s cloud computing service is differentiated partly because it was created for business people and is easier to work with than similar tools from others that are intended only for developers. Despite the flood of new AI-auditing tools, the problem of AI and bias will likely continue to persist because rooting out bias from AI is still in its infancy.