AI Ethics Effort Starts at IEE
Today, the IEEE kicked off a broad initiative to make ethics a part of the design process for systems using artificial intelligence. The effort, in the works for more than a year, hopes to spark conversations that lead to consensus-driven actions and has already generated three standards efforts.
Source: Rick Merritt
The society published a 138-page report today that outlines a smorgasbord of issues at the intersection of AI technology and values. They range from how to identify and handle privacy around personal information to how to define and audit human responsibilities for autonomous weapon systems.
The report raises a laundry list of provocative questions, such as whether mixed-reality systems could be used for mind control or therapy. It also provides some candidate recommendations and suggests a process for getting community feedback on them.
“The point is to empower people working on this technology to take ethics into account,” said Raja Chatila, who heads up the initiative as chair of the IEEE Global Initiative for Ethical Considerations in Artificial Intelligence and Autonomous Systems.
The effort seeks input not only from engineers but from business people, lawyers, economists, and philosophers. “The answers will come from society — not only the experts but the system users and other stakeholders,” said Chatila, who is director of research at the French National Center of Scientific Research and teaches about AI and robotics at Pierre and Marie Curie University in Paris.
“It’s not enough to come up with a list of concerns — we want to come up with methodologies and proposals for standards,” he said.
Chatila was one of a handful of people who came up with the idea for the initiative at an IEEE meeting more than a year ago. At the time, he was president of the IEEE’s society of robotics and automation and a similar group in France that had just written a report on design ethics.
The core group evolved to a team of more than 100 academic and industry representatives who drafted the report. They include representatives from Cisco, IBM, Google, and NXP.
So far, the group has spawned three ideas already going through the IEEE standards process. Two others, identified as P7001 and P7002, are more specific, dealing with data privacy in autonomous systems.
The group seeks feedback online here by March 6, 2017 on the overall initiative or any of its candidate proposals. The discussion aims to raise awareness of the issues, provide education, and spawn actions.
“Ethical design is only possible if it’s based on broad consensus,” said Chatila, suggesting companies involved in AI may someday have their own chief ethics officers.
— Rick Merritt, Silicon Valley Bureau Chief, EE Time