Now that you’ve learned how to scan your own shampoo at CVS and be your own waiter at Chili’s, put on your best scowl and get ready, because self-service security checkpoints are coming.
On Wednesday a Palo Alto (Calif.)-based startup called Qylur (pronounced KI-lure) said it would begin offering automated security checkpoints next year, after running small-scale tests of the machines in airports, sporting arenas, and elsewhere over the last few months. At the heart of Qylur’s idea is the basic Silicon Valley premise that machines can do an increasing number of jobs better than people, and do so without complaining, forming unions, or taking suspiciously long bathroom breaks.
The machines, which are made of a series of honeycombed cells surrounding a sensor, automatically check for dangerous-looking items and sniff for chemicals and nuclear material. A person puts a bag into one side of the machine, scans a ticket or a boarding pass, and closes the door. The machine then scans the contents and compares their characteristics to those of every item it has ever scanned. The point is not simply match a knife in a carry-on to a knife in a database, but to understand what a knife is. As it scans bags, it stacks up more knowledge about such shapes, improving its decision-making over time, says Lisa Dolev, the company’s founder and chief executive.
Dolev and other Qylur representatives wouldn’t give much more in the way of specifics about the system’s techniques. They would only discuss effectiveness in broad strokes, saying that its system catches more things passing through while also lowering the rate of false positives—those awful moments when the security guard thinks he’s found a bomb in your bag, even though 9 times out of 10 it’s an electric razor. But the company says it has received federal certification protecting clients from liability if something does make it through.
The company was happy, though, to talk about how much cheaper the system would make it to set up security checkpoints. It says a single machine with five cells could replace five security lines at a TSA checkpoint, moving through the same amount of people in one-quarter of the space and needing only five employees, rather than 15. Further, it charges clients between $0.20 and $1 per bag, with Qylur owning and maintaining the hardware. Also, inevitably, it suggests that venues could sell ads on each of the device’s 10 screens.
Qylur thinks its machines will lead to security checkpoints in places that simply can’t afford them now. Stadiums and amusement parks are early targets. Clearly, the gems of this market are the country’s airports, but the company acknowledges that a massive shift in technology at TSA checkpoints isn’t imminent. Its first batch of scanners totals just five devices. Qylur says they’re all spoken for, but wouldn’t say who its clients are.
In a moment of irrational exuberance, a spokesman for the company said that its products could actually make going through security pleasant. It’s true: No one likes having someone rifle through their bags, and having a machine do so instead could feel more private. Then again, given the heightened sensitivity to electronic data collection, some people might prefer that the one examining the embarrassing contents of their handbags be an uninterested rent-a-cop rather than a learning supercomputer. To each his own.
A basic challenge, according to Dolev, is persuading clients to overcome the impression that people are uniquely talented at picking out threats. It’s a widely held opinion despite likely being false. “Even though we know objectively that people aren’t the most reliable, they still have a comfort zone with a person,” she says. The company is offering what it calls a collaborative mode, where people can check the computer’s judgment. Dolev thinks that will be a passing fancy.
The fact is, cutting out the middleman often comes with its own problems, which is why other companies running machine-learning projects have sometimes learned that it helps to keep people in the loop. And machines enabling self-service aren’t always a panacea. Kmart, for instance, was early to the self-checkout aisle trend, but decided to reverse course after finding that the machines ended up being more expensive and led to higher rates of theft. And even if automated security checkpoints are better on average, given the stakes, a single incident could doom the entire concept.