Responsible AI, the Microsoft way

Concerns about ethical and responsible use of AI are as old as the discipline itself. But so far it has been rare in our modern (i.e. from about 2012 onwards) AI journey to find these concerns being put into the practice of shipping software by the companies that build and sell AI software beyond laying out a set of principles.

That’s why I find Microsoft’s recent unveiling of its Office of Responsible AI to be worth noting as something that moves the discussion from well, just discussion, to actually affecting the release of AI software into the market.

IBM has an extensive and evolving document on AI ethics, Google has published its Responsible AI Practices and Salesforce has published its AI ethics principles. And there are others, including many that say they adhere to the tenets of the Partnership on AI. But none, I would argue have gone as far as Microsoft yet.

The Office of RAI goes beyond the company’s six AI principles laid out ‘The Future Computed’ the book Microsoft published in 2018 (those principles being fairness, inclusiveness, reliability and safety, transparency, privacy and security and finally, accountability). It also builds upon Microsoft’s AI, Ethics, and Effects in Engineering and Research (AETHER) Committee, which was established in 2017 to provide Microsoft leadership with advice around the challenges posed by implementing AI at scale as well as an internal governance structure.

The Office of Responsible AI has four areas of responsibility. Firstly, it sets company-wide policies for the responsible implementation of AI. Secondly, it ensures readiness to adopt practices within Microsoft and helps customers and partners to do the same, which is known as enablement. Thirdly its case management role sees it review sensitive use cases to help ensure the AI principles are not being infringed, and report back to AETHER. And lastly, in its public policy role, the Office will help shape new laws, norms, and standards.

I am yet to be made aware of an example of how something didn’t make it through Microsoft’s product release cycle because of the Office of RAI’s intervention, but it’s very early days yet and I’ll be keeping a close watch on the activities of the Office of RAI to see if Microsoft has come up with a blueprint that can work for other technology vendors.




Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.