By Jason Fields
WASHINGTON, Feb 11 (Thomson Reuters Foundation) — Scientists are hoping artificial intelligence (AI) can help identify and rescue child victims of sex trafficking via an app and more than a million crowdsourced pictures of hotel rooms.
Researchers launched an app in 2016 to collect photographs of 50,000 hotels around the world that could be matched up with online adverts placed by traffickers, who often use selfies taken by their victims in hotel rooms.
Hotels-50K can be used to identify where trafficking victims are being held — and ultimately to rescue them, according to a paper written by a group of scientists and presented at a conference on AI in Hawaii.
A huge number of pictures are put through an artificial intelligence engine called a deep convolutional neural network.
The engine "learns a set of filters," said Abby Stylianou, co-author of the paper. "Here's what a headboard looks like if it's from this hotel. Here's what a lamp looks like in that hotel."
The initiative comes as hotels around the world are making efforts to stop sexual slavery, teaching employees signs to look for, including frequent sheet changes and "do not disturb" signs that are never removed.
There are 4.5 million people in sexual slavery around the world, according to the International Labor Organization.
The idea is to be able to pick out the specific hotel room and use the information to attempt a rescue.
By getting 150,000 people all over the world to download their app, TraffickCam, and take hundreds of thousands of pictures, it is easier to get the same view that a trafficking victim's selfie taken for an ad is likely to have.
Once Hotels-50K was ready for testing about eight months ago, the scientists shared it with the National Center for Missing and Exploited Children (NCMEC), a US non-profit involved in the fight against child sex trafficking.
It is unclear whether anyone has been saved directly because of Hotels-50K.
"We've had a couple (of ads) that have given us leads and information," Staca Shehan, executive director of the case analysis division at NCMEC, told the Thomson Reuters Foundation.
"There's a lot of contributing factors that go into a recovery of a child."
One of the key things that the scientists are not trying to recognize is the people in the images. Before an image of a hotel room where a victim is being held is uploaded, certain steps have to be taken.
"The first thing that they have to do is erase the victim from the image. The first reason is to protect the victim, so that their image isn't going back and forth across the internet," Stylianou said.
This technology could easily be abused, Stylianou said, and that is why only organizations like NCMEC and the police will be allowed to use Hotels-50K for the moment.
"We have created a piece of software that helps locate people. It has a good use case here. We're trying to rescue people."
(Reporting by Jason Fields; Editing by Claire Cozens. Please credit the Thomson Reuters Foundation, the charitable arm of Thomson Reuters, that covers humanitarian news, women's and LGBT+ rights, human trafficking, property rights and climate change. Visit http://news.trust.org)