A number of companies have responded to the recent horrific mass shootings by promoting technology that it claims can detect people with guns. Two companies in particular have garnered a lot of press attention with their products: one that makes AI-enhanced metal detectors, and another that sells video analytics software that “watches” security camera feeds and raises an alarm when computer vision believes it is to see person holding a gun.
While such technologies may have their place, as a society we need to think carefully about whether, how and where to deploy them. And at the end of the day, surveillance cannot be the answer to our gun problem.
One product made by a company called Evolv is a metal detector enhanced with machine learning, which the company claims increases accuracy. Metal detectors and other body scanners are already being used in various locations across the United States, and we believe many of these uses are unwarranted and harmful. Where scanners do exist, however, making them more accurate and less intrusive, all other things being equal, is a good thing.
The problem is that all other things are not created equal. Evolv is pushing for a proliferation of metal detectors in American life, not just an upgrade of existing facilities. The company, which has raised half a billion dollars in funding, claims it currently scans over three-quarters of a million people a day and has ambitions to install thousands of new detectors across the US and around the world.
Improving current scanners is one thing, but expanding the number of scan control points is another. We don’t want America turned into one big airport — a checkpoint society where we have to submit to searches at every gathering, including little league games and onwards.
And metal detectors don’t work by themselves. Wherever a search checkpoint is established, there will be guards. They can yell at you orders, demand that you open your pockets, take off your jacket, etc. As common as such scans are, we should never lose sight of the fact that they are a serious invasion of privacy. You may have personal items with you, such as B. Medical devices that are nobody’s business but yours. Guards can also create racial profiles or just behave roughly and rudely.
Perhaps the worst place for these checkpoints to proliferate is in our schools, which see corporations as a big market. School metal detectors are consistent with a fractured approach to student welfare — the same approach that has resulted in the United States having more police officers than social workers in our schools. Metal detectors are expensive, disruptive, alienating, go hand-in-hand with aggressive police tactics, and have a racially diverse effect on black students. Like other surveillance methods, they are not good for children. And a lot of research suggests that metal detectors don’t improve school safety; Even the US Secret Service has concluded that the devices are unlikely to stop school shootings.
Outside of schools, too, the effectiveness of metal detectors is questionable for the same reasons. As Motherboard put it, “Security professionals commonly believe that body scanners and metal detectors don’t prevent mass shootings.” Although independent security analysis publication IPVM found Evolv to be better than traditional (and far cheaper) metal detectors because it doesn’t set off an alarm on keys and cellphones Evolv’s overall accuracy claims sharply challenged by IPVM. In particular, an investigation by IPVM and the BBC revealed that the company had worked with an “independent” testing lab to exaggerate the performance of their scanners, which in some cases was actually inferior to regular metal detectors.
A company called ZeroEyes is marketing video analytics techniques that it claims can detect when someone is holding a firearm in a security camera feed.
In many contexts, there is no reason that anyone should ever be able to handle a firearm, and it is usually legitimate for security personnel to be alerted to someone brandishing a gun in a public place. (In many states, it may be legal to holster a firearm, either concealed or open, but holding a gun in public in public is usually illegal under state swinging and mugging laws.) So many people blush at first I don’t see any problems with a system that only triggers an alarm in this case. However, there are some problems I see with this concept.
First, this concept encourages the installation of surveillance cameras everywhere. And since weapon detection analysis acts as an add-on to existing cameras, it’s not as if the cameras involved are only used to search for weapons. They can be used simultaneously for the full range of surveillance purposes, including facial recognition, forensic search and even marketing and “business intelligence”.
Second, as with all alarm systems, and especially AI systems, there will be false alarms – potentially lots of them. Covering public spaces with buggy gun detectors can increase the incidence of tragic confrontations sparked by people holding cellphones, toy guns, or other everyday objects that police mistook for firearms. ZeroEyes has operatives in the loop reviewing camera images before sending alerts, which is helpful — but that’s no guarantee against errors. And if that concept is successful, we’re likely to see that company or its competitors discard the expensive and complicated element of human verification.
Third, as with Evolv’s weapon detectors, there are also questions about effectiveness. The system requires a weapon to be visible for a significant period of time before firing begins. How often does that happen? As this technology spreads, future shooters will simply hide their gun(s) from the AI until the last moment (which might be easier than hiding them from a human). As IPVM industry analysts conclude, video analytics “cannot stop [violent] events and will not add significant advance warning time to an event.”
Fourth, it’s also important to understand a larger context here: video weapon detection is part of a larger trend to use computer vision as a cheap substitute for humans in enforcement and surveillance work. We’re already seeing AI-powered cameras being deployed to monitor things like driving behavior and cellphone use, and (as we’ve explored in this report) this is likely just the tip of the iceberg to come leading to intrusiveness and oppression could levels of surveillance throughout our lives.
Finally, both ZeroEyes and Evolv raise questions about transparency. It is difficult for policy makers and the public to make decisions about technology without accurate and complete information about their false negative and false positive rates. Evolv actually refused to let IPVM buy a scanner for independent accuracy testing.
Ultimately, ubiquitous surveillance is not the solution to gun violence. Americans should say no to intrusive technologies that threaten privacy, bring dubious benefits, and have negative side effects like racial profiling and disruptions to education.