Facial recognition technology provokes unease, but many of the charges against it do not bear scrutiny. A survey of facial recognition shows that the trend is for this technology to become more accurate and less biased without the need for policy intervention.
The disconnect between public discourse and actual performance raises three problems. How do we clarify the discrepancies in the public discussion of facial recognition? How do we construct a better research agenda for policy issues? And how do we avoid imposing policies that are unnecessary or harmful as we create reasonable protections?
One way to think about facial recognition (and artificial intelligence in general) is to compare the concern that greets these technologies with the commentary that greeted the commercialization of the internet. The internet created vast opportunities, but it also came with new risks for security and privacy. It can be seen as a mixed blessing, but while cybercrime and other malicious acts cost billions of dollars, the internet generates trillions in revenues. The internet's growth was intentionally unhampered, as the policy of the 1990s was to seize opportunity despite the risks. Overregulation would have stifled growth. Twenty-five years later, the internet is no longer a fragile blossom that needs protection, but if it were commercialized today, it might have met with a very different response, one that slowed or even blocked deployment in the United States. This would have been a disaster.
Some say that we now have a chance to avoid all the "errors" that accompanied the internet, and while there is now a strong need for regulation, it must be carefully designed and not based on inaccurate analysis. If we had banned the internet or forced innovators to show that it was impartial or secure before allowing it to go forward, we would still be using dial phones because these kinds of policies kill creativity and experimentation.
It is particularly worrisome that many of the concerns expressed over facial recognition and AI are hypothetical. We would want to determine, for example, how many false arrests there has been because of the use of facial recognition technologies. A preliminary search finds only one record of a false arrest. Better data on false arrests would help pinpoint areas of concern.
The disparity in views of facial recognition in Asia and in the United States is also worrying. Facial recognition technologies are widely used in Asia, but there has been no outcry about discrimination. If the technology itself was flawed, we would expect to see similar problems across national implementations. We do not know if this reflects a lack of diversity in Asian populations or whether larger social narratives shape some of the concern over facial recognition in the United States. This would be another useful area for research, since understanding the cause of the discrepancy in views is crucial for accurately depicting risk.
Americans say they want innovation, but innovation only comes with risk. Risk-free innovation is a non-sequitur. The United States and China believe that they are in a technology race. China sees opportunity and growth when it looks at artificial intelligence, not risks. It would be wrong to ascribe this to China's authoritarianism and mass surveillance. There is a thirst for success that makes Chinese such great entrepreneurs and potentially formidable competitors (should they ever resolve their governance problems).
Fears about facial recognition bear some similarity to the surge in conspiracy theories. Fears that government use of facial recognition technology will lead to the pervasive surveillance system seen in China are irrational. It reflects a lack of faith in the strength of democratic institutions. Only the most uninformed or paranoid think the United States is a police state. What this fear of "Big Brother" reflects declining confidence in government (particularly elected officials) to provide adequate oversight. Similarly, the collapse of privacy and the lack of meaningful privacy rules in the United States exacerbate worries about facial recognition.
The trend in facial recognition is toward greater accuracy, not less. Relying on higher accuracy (or confidence levels) greatly reduces the kind of errors that have attracted so much criticism. Japanese and Chinese companies have reportedly been able to achieve facial accuracy rates of over 98 percent. The National Institute of Standards and Technology (NIST) has been testing facial recognition technologies for says that "between 2014 and 2018, facial recognition software got 20 times better at searching a database to find a matching photograph." NIST found "a rapidly advancing marketplace for face-based biometric matching algorithms." It is fair to ask if some of the criticisms of misidentification were simply premature.
There are important privacy concerns with the use of facial recognition technologies that the United States needs to address in a demonstrably honest and transparent manner. For example, we need adequate principles and rules for data retention and use, and for redress. But like the internet when it was commercialized, facial recognition technologies offer more opportunities than costs. Americans are more risk averse than they were 25 years ago, but we should recall the experience of the internet and seize the benefits it offers.
Why have so many Americans come to fear technology? One salient example is the anti-vaxxers, who managed to change measles from a disease close to eradication to one that is again flourishing. Misinformation, conspiracy theories, and social media-fueled anger combine with distrust to create a bad outcome for vaccinations. We risk similarly bad outcomes from some of the critiques of facial recognition.
James Andrew Lewis is a senior vice president at the Center for Strategic and International Studies in Washington, D.C.
Commentary is produced by the Center for Strategic and International Studies (CSIS), a private, tax-exempt institution focusing on international public policy issues. Its research is nonpartisan and nonproprietary. CSIS does not take specific policy positions. Accordingly, all views, positions, and conclusions expressed in this publication should be understood to be solely those of the author(s).
No comments:
Post a Comment