Law enforcement isn’t the only user of facial recognition technology. Indeed, a substantial amount of mainstream adoption and investment into the technology is actually being carried out by businesses. As Silicon Valley and major corporations such as Amazon and Microsoft flock to the sector, the business models for facial recognition firms are still emerging. Often, facial recognition software is offered as part of a full suite of “scalable biometrics” security services, or as an add-on to video analytics tools. Retailers, telecommunications firms, and even schools are increasingly incorporating these data services. Although it is still a relatively small market at present, Allied Market Research predicts it to expand to a nearly $10B market by 2022 (Roberts 2019). Acquisitions of facial recognition startups happen largely out of the public view, with few accompanying corporate news articles or shiny press releases. For example, the tech start-up Orbeus recently developed a new project called Rekognition, which was acquired by Amazon in 2016 with very little business reporting. The technology behind Rekognition has since been incorporated into Amazon's broader Web Services platform.
The underlying technology of face recognition is rooted in algorithmic, deep learning processes that are “trained” to detect faces using certain data sets. Like a human brain, the algorithmic neural networks must be fed a large amount of visual information in order to develop techniques for processing and identifying faces. These sets of data are comprised of hundreds of millions of images, which are often obtained through highly questionable methods. Among these methods are data collection via photo apps, such as PhotoTime, RealTime, or the infamous EverRoll (banned from the App Store in 2016 for violating users’ privacy). Additionally, facial recognition startups utilize “data scrapers” to pull publicly accessible images from photo sharing sites such as Tumblr and Flickr. Rather than being managed manually or holistically, these datasets are often populated automatically and without human oversight. And if those datasets are weighted toward certain races, genders, or ethnicities, then the technology will be worse at identifying a broad spectrum of faces.
Walmart is among the companies developing facial recognition systems to analyze the perceived moods of their shoppers. McDonald’s has also implemented the technology, but instead aims it at staff to assess if they are smiling while they serve their customers (Tech HQ, 2020). The computer chip manufacturer Intel has utilized face recognition at its headquarters for “security” purposes, but insists that its version of the technology sufficiently protects the privacy of employees and visitors. There is significant skepticism being directed at the new leaders in facial recognition technology research, especially Microsoft and Amazon. These companies have publicly called for limits on its usage until proper regulation passes through Congress—even as they dedicate substantial resources to its development. In the wake of the George Floyd protests, the two companies have both canceled all plans to sell their facial recognition technology to police forces until 2021. However, critics view this as an arbitrary date intended merely to redirect public scrutiny away their activities for a short period of time.
Image source: Cognitec Systems, FaceVACS-DBScan LE YouTube Demo (2018)