From the same people that brought you autonomous “corpse eating robots” so you know it’s got to be safe!
KAUFMAN: So we find this phone number. So now what you’re going to see here is the person that was associated with the phone number. This is every single website and every single phone number that that guy is connected to. So we’re starting to draw the map of what looks like a trafficking ring.
SIEGEL: Starting with that single email address, that single clue, using Memex Dan Kaufman has found a name and then a whole series of additional phone numbers linked through advertisements. Perhaps there’s one underlying group behind this long list of ads for sexual services. And as Kaufman demonstrates, the ads can be mapped. They’re on sites all over the globe.
KAUFMAN: We can also place these things geographically. Some of the ads are in California. Some are in Joplin, Mo., and we’re seeing some in Southeast Asia.
SIEGEL: Well, can you actually match pictures?
KAUFMAN: We can.
SIEGEL: The images of women leaning seductively into the camera are blurred, Dan Kaufman says to preserve investigators’ sanity, but the computer doesn’t have any trouble spotting even the tiniest details.
KAUFMAN: Some of the pictures are in blue, so that says every single picture we believe is the same person. We can also tell you if it’s the same camera. So either A – I’m seeing a woman being moved from place to place as trafficking, or I’m seeing the same people used over and over again. And again, I’m starting to see connections, so I can see this actually looks like a large, complex network.
SIEGEL: But you’re not just finding that it’s the very same picture. You’re finding that it is a picture of the same or a very similar looking person.
KAUFMAN: That’s right, we do both. So obviously, the easy one…
SIEGEL: Is the same picture.
KAUFMAN: Same picture.
SIEGEL: But you go beyond that.
KAUFMAN: We go beyond that.
SIEGEL: Dan Kaufman’s colleague Wade Shen was standing nearby helping with the demonstration.
WADE SHEN: We can also find out things like, for instance, are these two pictures taken in the same hotel room? Whether or not the lighting is similar, whether or not the room environments are similar, whether or not the cameras are similar and so one, so it’s not just the individual them self.
KAUFMAN: It then gives him a clue. It says would you like to find similar pictures? So then he can click on the button and it will then search through the exact same database and here’s all…
SIEGEL: All the same picture…
KAUFMAN: All the same picture.
SIEGEL: …Of the same woman on different sites with different phone numbers.
KAUFMAN: So now you’re starting to see the power of it. So if you were just searching for one phone number you would never have found this.
SIEGEL: What I understand from this is – she’s the trademark is what she is of some operation.
KAUFMAN: That’s right.
SIEGEL: Who this person is might be quite irrelevant.
SHEN: It’s a signature of a ring.
SIEGEL: Signature of a ring, though the first impression you would have is that this woman is in Las Vegas.
KAUFMAN: That’s right.
SIEGEL: Not necessarily at all.
KAUFMAN: That’s exactly right. And now you can ask really interesting analyst questions. How many other websites have used the same camera? Can I look at it over time? Can I see the map? And we’re just empowering the cop. The police know these questions. They know how to do this, but they don’t have tools to do it, so they’re tracking it by onesies and it’s hard. And we’re trying to make their lives a little bit better.
SIEGEL: That’s Dan Kaufman. He has now left DARPA to work for Google. We also heard from Wade Shen, program manager at DARPA’s Information Innovation Office. Well, is Memex in fact making things better for law enforcement? In New York City, a human trafficking unit has been putting it to work.
CYRUS VANCE: Memex is a Google search on steroids.