House subcommittee considers impact of facial recognition bias on government adoption of AI technologies
The challenge of racial bias in facial recognition systems is being examined in hearings by the Subcommittee on Information Technology of the House Oversight and Government Reform Committee, as it considers ways to guide government understanding and adoption of artificial intelligence, GCN reports.
“I want to get to a point where we can be making decisions within the government on where to spend dollars or resources based on the analysis of large volumes of data,” said Rep. Will Hurd (R-Texas) in a video produced by the committee. Applications being considered include AI-based sentencing and facial recognition systems.
Four witnesses testified at the recent first session of the hearings, with all agreeing that biased data results in biased application, but with some disagreement about how to avoid it.
Professor and Executive Associate Dean for the College of Computing at Georgia Institute of Technology Charles Isbell said that algorithmic transparency would help by informing individuals of how decisions are arrived at. Allen Institute for Artificial Intelligence CEO Oren Etzioni disagreed, stressing the importance of accuracy. Etzioni also suggested that consumer products using AI should be labelled, but Intel Vice President and Chief Technology Officer Amir Khosrowshahi warned against the imposition of regulation on a new and developing technology.
Vice President and General Manager of Accelerated Computing at NVIDIA Ian Buck said that government can provide a wealth of data to improve AI performance. He said that the OPEN Government Data Act, which passed through Congress last year with bipartisan support, will help make that possible.
As previously reported, leading facial recognition systems with high accuracy rates for light-skinned segments of the population perform inadequately when applied to darker-skinned people.