Sony first set up its AI ethics guidelines in 2018.
“AI became an early focus of ours back in 2018 when we first set up our AI ethics guidelines.”
Person
Alice Xiang
1 episodes · 21 quoted claims
Sony is a multinational company headquartered in Japan with businesses in creative entertainment and technology.
“Sony is a large multinational company headquartered in Japan, with a diverse array of businesses around creative entertainment and technology.”
Sony's operating companies focus on music, motion pictures, video games and electronics.
“So we have operating companies focused on music, motion pictures, video games and electronics.”
Alice Xiang established Sony's AI ethics office and its processes for evaluating AI use in products and services.
“I established our AI ethics office and our processes in terms of how different AI uses and AI integration into products and services is evaluated for responsible AI.”
Sony AI's Fair Human-centric Image Benchmark (FHIBE) was published in the journal Nature.
“that's recently culminated in a publication in Nature of our Fair Human-centric Image Benchmark, also known as FHIBE.”
Alice Xiang says computer vision played a major role in the deep learning revolution.
“Computer vision is a field that played a major role in terms of the deep learning revolution.”
Alice Xiang says web-scraped datasets that were massive and cheap to source helped drive computer vision's progress.
“part of that was these web scraped data sets that were massive and relatively cheap to source.”
FHIBE used self-reported demographic information rather than third-party annotators' guesses to make its labels more accurate.
“we did use self-reported demographic information in an effort both to ensure our labels were more accurate and also that we weren't relying on third-party annotators to guess.”
FHIBE includes extensive annotations about the environment, people's physical attributes and the cameras used.
“And then we had also extensive annotations about the environment, other physical attributes of the person, the cameras being used, all sorts of things.”
More than sixty institutions from academia, industry and government downloaded FHIBE in the first couple of weeks after its release.
“There were over sixty different institutions downloading it, folks from academic institutions, industry, and government institutions.”
Sony tested FHIBE with several business units developing computer vision technologies before its public launch.
“FHIBE is being used as well across Sony, even before the launch. We tested it out with a number of our business units that are developing computer vision technologies.”
Alice Xiang cites having to look at a phone several times before it recognizes you as a minor consequence of biased face recognition.
“Maybe you have to look at your phone several times at different angles before it can recognize you.”
Alice Xiang says biased computer vision can lead to serious harms including financial fraud and wrongful arrests.
“But this can also lead to much more problematic impacts in terms of anything from financial fraud to folks being wrongfully arrested.”
Alice Xiang says FHIBE is a proof of concept that ethical data sourcing is possible at some scale, though harder and more expensive.
“Obviously, it's more difficult, it's more expensive, but, you know, FHIBE was kind of the proof of concept that at least on some scale, you're able to do this.”
FHIBE focuses on human-centric computer vision because Alice Xiang considers it one of the most sensitive areas for data.
“We focused FHIBE on human-centric computer vision, precisely because it is one of the most sensitive areas.”
Alice Xiang built her first machine learning model about twelve years before this episode and found it biased toward certain geographies.
“This was like twelve years ago now, but it got me really interested in this area.”
Alice Xiang holds graduate degrees in law, statistics and economics.
“I have graduate degrees in law, statistics and economics.”
At Partnership on AI convenings, companies said their privacy teams often block collecting demographic data needed for fairness assessments.
“if we want to collect data that has demographic information in it, our privacy team will just shut that down.”
Alice Xiang started a research lab at the Partnership on AI focused on algorithmic bias, where data availability quickly emerged as a key issue.
“I started up a research lab focused on these areas, and pretty quickly, these issues of data availability became quite salient.”
Alice Xiang says in recent years employees' performance targets at companies have become focused on AI adoption.
“I think it's been interesting to see just in the past couple years how everyone's KPIs have suddenly become about AI adoption.”
Alice Xiang says the worst uses of AI are autonomous ones lacking human oversight or oversight by people with sufficient expertise.
“where it's just kind of being used in a very autonomous fashion, and there's either no one providing oversight or the individuals providing oversight don't have the knowledge or expertise to do so.”