This past weekend the New York Times published a fascinating look at just how commoditized facial recognition has become, hooking up a set of public park surveillance cameras to Amazon’s commercial cloud-based facial recognition software and feeding it a database of publicly downloaded profiles images of nearby office workers to create an instant realtime and uncannily powerful facial recognition application detecting 2,750 faces over a nine hour period for the grand total of just $60. While the Times’ demonstration was simplistic, it is not that far afield from the production systems used by governments and companies across the world today and demonstrates just how trivial it is to launch a very real facial recognition network. Most importantly, it reminds us that facial recognition is here to stay.
The rise of mass facial recognition, which has already become a public reality in China and a private reality in a number of cities around the world, is here to stay. Much like nuclear weapons and autonomous “killer robots,” what one superpower possesses, the rest of the world must acquire to match pace.
In the case of facial recognition, what makes it so dangerous is how easily such algorithms can be built. Facial recognition algorithms are not billion-dollar weapons systems relying on exotic export-controlled hardware. Anyone can build a primitive facial recognizer with a few lines of open source code, using an open source toolkit and training using open image datasets scraped from the web.
This ease of creation means that like encryption, the choice to ban facial recognition globally is no longer in US hands. Even if Congress passed a law absolutely prohibiting all facial recognition, that would do nothing to stop its use across the world. The algorithms and knowhow to build facial recognizers is now simply too widely known and commoditized.
Moreover, the globalized nature of the internet means US citizens residing in the US would likely find themselves mined from afar by businesses in foreign countries mass scraping their imagery to build recognition models. Foreign intelligence services will find themselves under no such constraints even if the US banned facial recognition.
Yet, it is the security use of facial recognition that is perhaps the greatest argument for why its use is inevitable in our darkening and violent world.
Europe long condemned the US approach to the surveillance state, passing legislation and adopting governmental constraints that restricted their security services from many of the US’ most controversial mass intelligence harvesting efforts. In the aftermath of public outcry after a series of domestic terror attacks across Europe and a growing recognition of the terrorist threat, those same European nations largely reversed course, adopting those very provisions they had historically condemned.
In short, privacy only matters until it collides with security reality, at which point even the most open societies cannot move quickly enough to roll back their most sacrosanct privacy protections in the name of safety and security.
Societies value their liberties until they are confronted with their costs.
Imagine if the US did pass a law that absolutely banned without exception all use of facial recognition. A major terrorist attack occurs and it emerges that the perpetrators were well known to authorities and were captured clearly on surveillance cameras where they could have been immediately caught and the tragedy prevented, if only those cameras had employed facial recognition.
It is not hard to imagine the public outcry regarding such a preventable mass loss of life and the immediate repeal of facial recognition bans, likely with an outpouring of public support and apologies from the sponsors of the original bans.
Indeed, it is noteworthy that many of the same commentators and organizations calling for outright bans on facial recognition have been careful to call for narrow exceptions for national emergencies and life-or-death situations like searching for an abducted child at great risk of physical harm.
The problem is that as our government has taught us again and again, the narrowest exception is rapidly expanded into a wholesale exemption. A “national emergency” exception could easily be interpreted as permitting unlimited police use for counterterrorism.
Returning to the encryption debate, there are many parallels between governmental calls for banning encryption and civil society calls for banning facial recognition. Not least is the fact that both sets of calls are meaningless when the underlying technology is so widespread and easily accessible.
It is almost impossible to ban a technology can be built with a few lines of code and deployed from afar.
It is even more impossible to enforce a ban on a technology whose user can apply it on US soil while remaining safely abroad.
After all, a Chinese startup can easily take live surveillance camera feeds from across the US and feed them into a facial recognition engine running in a Chinese data center and deploy mass facial recognition across the US without any fear of legal consequence. That startup could even sell its services back to US law enforcement, entirely skirting any facial recognition ban.
Setting aside the ethical and moral considerations of facial recognition, if we look only at the practical reality of its availability and the logistics of attempting to ban, curb or regulate its use, we see that, like encryption, the commodity nature and trivial global accessibility of such technology means any ban would be merely a futile exercise.
Putting this all together, facial recognition is here to stay no matter what the US attempts to do about it.
In the end, we are already so closely tracked by so many companies that all of the future fears we ascribe to the rise of facial recognition are already here.
They have just been here for so long we no longer even notice them.
The surveillance state has already arisen.
