Glaser: European-built facial recognition performed worse on non-European faces
Harry Glaser · The Conscience of AI // Harry Glaser, Periscope Data (FirstMark's Data Driven) · Apr 9, 2018 · at 5:21
Harry Glaser, CEO of Periscope Data, discussing real-world examples of algorithmic bias in AI.
0:00 / 0:23exact quote · 23.6s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“One of them is a facial recognition software, a classifier, and this particular classifier, this particular facial recognition was built by a team of folks of European descent, much like myself, and it turned out to be very good at recognizing faces of European descent, and it turned out to not be as good at recognizing faces of African descent or of Asian descent which proved to be a pretty big embarrassment when it got out into the world.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Harry Glaser
Glaser: Executives, customers, and governments do not understand AI technology
“I'm here to tell you as a CEO, it's not going to be your CEO, and it's not going to be your users, and it's not going to be your customers, and it's not going to be your government, because they ultimately don't know how this technology works, and they ultimat…”
Glaser: Open dialogue yields better AI outcomes than closed corporate consortiums
“If we said the CEOs of six companies are gonna get together in a consortium and figure this stuff out, I'd be quite worried, right? But if we said, we're all going to have a series of open dialogues, and we're going to have a pretty open and transparent cultur…”
Glaser: Recent public criticism of Google's internal culture is likely deserved
“The culture there that has taken some beatings recently, I think probably deservedly”
Periscope Data accepts ethical responsibility for how customers use its software
“We try to reject the view that we provide a tool, and therefore we have no responsibility for the tool. You know, it's yours now. Use it. And so we try to do a few things. First of all, we try to accept responsibility for how the tool is being used and act acc…”
Periscope Data intentionally rejects black-box AI in favor of human-operated tools
“We are pretty queasy about, you know, AI in a box. Press this button, AI a thing, it'll predict for you, you're going to be great. We make a tool, we frame it as a tool for data professionals on purpose. The idea is there should be a person making these, you k…”