Facial recognition: Nowhere to run, nowhere to hide

A conversation with Kashmir Hill, author of the chilling new book, "Your Face Belongs to Us"

People who write about the dangers of AI tend to drift pretty quickly toward scenarios out of the Terminator movies: Super intelligent machines decide they no longer need annoying carbon-based life forms and set out to exterminate them.

While that scenario is theoretically possible [1], the more immediate danger is both less cinematic and more likely. Thanks to advances in AI, facial recognition has gotten really good, really quickly. And if used indiscriminately, it could exterminate whatever privacy we carbon-based life forms have left.

New York Times reporter extraordinaire Kashmir Hill knows this better than probably anyone. Her new book, "Your Face Belongs to Us: A Secretive Startup's Quest to End Privacy as We Know It," should be required reading for anyone interested in protecting what's left of their civil liberties.

It's about a startup called Clearview [2] that set out to build software that could identify every human being on the planet. And while they haven't quite gotten there yet, the company's software is already being used by thousands of law enforcement agencies all across the United States [3].

I persuaded Kash to take time from her hectic book promotional tour to sit down for a quick video Q&A with yours truly, in a new segment I'm calling "15 Minutes of Infamy." That video, including a theatrically timed interruption from the National Emergency Broadcast System, can be found here.

For those of you too lazy to watch the video, here's an edited summary of our conversation.

Tell us about this secretive startup, Clearview.

It's a New York based startup launched by an odd young man named Hoan Ton That and an older founder, Richard Schwartz, who used to work for Rudy Giuliani when he was mayor. These two men scraped billions of photos from social media sites and the Internet, and used them to build a facial recognition app that's now being used by thousands of police officers all over the world. [4]

Clearview has roots in Trumpland, yes?

While in New York Ton That fell in with a conservative crowd and met Chuck Johnson, an infamous conservative provacateur (troll). They ended up attending the Republican National Convention in Cleveland that year (2016) where they met Peter Thiel [5], who ended up investing $200,000 in the company. He was their first investor.

As you note in the book, after you got a tip about this company, they tried to hide from you, blocking you from their photo database and instructing their customers to not talk to you.

At first Clearview did not want to be written about and tried to hide from the public. The address on their website was for a building that didn't exist. There was one employee listed on LinkedIn, a fake profile for a man named John Good. I mean, I've seen companies operate in stealth mode before but this was quite extraordinary. They were actively tracking me, telling police departments if I called to not talk to me. But eventually they hired a crisis communications person who convinced them to talk to me, because she knew I was going to write about them anyway.

So why should people worry about facial recognition? What's the potential for harm?

What concerns me is how accessible facial recognition technology is getting, and the possibility that you are no longer a stranger to the people around you. You will no longer be anonymous as you move through the world. When you're having a sensitive conversation in a hotel lobby or a restaurant, somebody could snap a picture and know who you are. Or if you go to an abortion clinic and protestors snap a photo of you, and then they can look up your name and address. We've already seen it being used at Madison Square Garden in New York, where the owners used it to ban people they didn't want to attend events, mostly lawyers from companies that were suing them. And then there are the more chilling examples of how it's being used to track people in China and Russia.

What's the biggest misconception people have about facial recognition?

That it doesn't really work that well. And yes, there is a long history of the technology being flawed and making mistakes, especially when trying to identify non-white non-male subjects. But I'm not sure people understand how powerful this technology really is. And when I demonstrate it by feeding one of their photos to the Clearview engine, they're shocked at what other photos it can pull up.

People who defend the use of this technology say it's a valuable tool for identifying criminals and locating missing children, among other uses. Is there a way to get the benefits of facial recognition without the dystopian elements?

I think it's still possible to choose the future we want to live in. We don't have to let this technology determine what the world looks like. That's part of why I wrote this book now, because this is the time to come up with the regulations and laws to reign in how the technology is used. There are activists who want to ban the technology entirely, but I don't see that happening — it's too useful for police officers when they use it appropriately. So, yes, I definitely think we can find the 'golden middle' with facial recognition.

Final question: Do you use facial recognition to, say, unlock your phone?

Yes, definitely. I open my phone with my face, open my apps, pay for the subway by looking at my phone and putting it next to the card readers. While I was doing reporting for this book, I traveled to London. When I landed at Heathrow, instead of waiting in a border control line for hours, I went to a kiosk that scanned my face and walked right in. It's a very alluring use case. At the same time, just because we want some of the beneficial uses of the technology doesn't mean we have to accept all of them. I don't want to be tracked everywhere I go by every surveillance camera I pass by. I think we can get the good without having to take the bad.

If you've ever posted a photo of yourself on the Internet, your face is very likely part of Clearview's database. Tell us how that makes you feel in the comments below. And be sure to share this post with your friends and neighbors.

[1] AI is already being used by advanced military drones to identify targets and blow them to smithereens. So far, there's still a human involved to OK the kill order. (At least, for the US Air Force.) Let's hope it stays that way.

[2] Also known as Those Evil M*therf*ckers.

[3] With the exception of Illinois, the lone US state with a biometrics privacy law that forbids companies from using your photos or other biometric markers without your permission. Even then, it was in use by dozens of police departments there before Clearview was sued and forced to stop.

[4] Per the book: The company, originally called Smartcheckr when it was launched in November 2016, "would be a tool that could theoretically identify and root out extreme liberals." (Page 52.)

[5] If you were looking to cast a villain in a superhero movie, you could do a lot worse than fascist-friendly billionaire Peter Thiel. He's a less charming combination of Lux Luther, Ernst Blofeld, and Thanos, with a touch of Voldemort.