If you have published a photo of yourself on the web and made it public, i.e. visible to everyone and not just a small circle of friends, there is a good chance that your face has ended up in the archive of Clearview AI. And, above all, that you don’t know it.
Clearview AI is like a search engine. Like Google, in short. Instead of typing a series of words, you upload someone’s photo. At that point the AI examines a database of 30 billion images collected from sites and social networks such as Facebook. As soon as a perfect match emerges, all available photos of that person appear, with links to the web pages where they were found. And therefore to names – often surnames – and details of the private lives of the people identified.
In many countries around the world what Clearview AI does is illegal. Collecting and storing is not permitted biometric data of a subject – including the conformation of the face – without his consent.
In Italy Clearview AI has already received a fine of 20 million euros by the Privacy Guarantor. But its founder, Hoan Ton-Thatis convinced that he is acting for good: “We want to help solve serious crimes such as human trafficking, murder and child abuse,” he told us.
Clearview AI’s software, in the USA, according to Ton-That “has been used a million times by law enforcement”. “Thanks to our facial recognition – the entrepreneur explained to us – a recent operation by the US Department of Internal Security made it possible to identify over 300 minor victims of abuse in three weeks”.
Hoan ton-That, founder and CEO of Clearview AI
Hoan Ton-That is an enigmatic character. The CEO of Clearview AI is 34 years old. In the past it was a part-time model and a developer of not-so-memorable apps. Then in 2017 he founded Clearview AI. His CV tells us a lot about the future we are facing.
Ton-That knew nothing about facial recognition. Yet he created a very powerful tool. It was enough for him to use “open source” technologies – that is, accessible and modifiable by everyone – and to study the “instructions” – tutorials, articles and academic studies – available online. When we asked him what the “key” moments of his career as an entrepreneur were, he replied elusively: “I grew up in Australia and I have always considered Italy a world capital of beauty and excellence”.
Unlike Facebook, Google e Apple, who have been developing facial recognition for years solely for their services – for efficient organization of photos on your smartphone, for example, or to unlock a device using only your face – Ton-That has decided to commercialize its invention. With disruptive effects on society, for better or for worse.
AI and war Ukraine plans to stop the war by sending photos of dead soldiers to their mothers in Russia by Pier Luigi Pisa 22 April 2022
Ukraine employs Clearview AI to identify possible spies, for identify the corpses of Russian soldiers and to track down those guilty of war crimes. The UK also believes facial recognition is key to tackling crime. “It will have a similar impact on investigative work as the use of DNA had 30 years ago,” Mark Rowley, head of the Metropolitan Police Service, said recently. one of the most important positions in the country.
“The Clearview AI algorithm – claims Ton-That – can correctly identify a face in a set of more than 12 million photos with an accuracy rate of 99.85 percent according to NIST [un’agenzia federale Usa che promuove l’innovazione, nda]”. But this technology it is not infallible.
In the last months Kashmir Hilla New York Times reporter, reported how six US citizens were arrested following incorrect facial recognition. They were all African Americans. It’s not a coincidence. Facial recognition they are 5 to 10 times more wrong when they test a black person. It can happen if the algorithms were trained on a dataset in which black people are poorly represented, or if the technology used reflects social biases.
Artificial intelligence AI between racism and discrimination: two years later, we are worse off than before by Emanuele Capone 27 March 2023
The risks don’t end there. An authoritarian government, for example, could use tools like Clearview AI to surveil political dissidents, monitor demonstrations or persecute minorities. “AI must not become an instrument of control and we must not allow its use, perhaps initially for a good purpose, to turn into abuse” he said Brando BenifeiMEP of the Democratic Party and rapporteur of the AI act with which the EU banned real-time facial recognition in public places.
“I am saddened by the misinterpretation by some in Italy, where we have no commercial activities, of the impact of Clearview AI technology on society” told us Ton-That, who with his company is preparing to take a further step forward . “We have developed a version of Clearview AI that works on augmented reality glasses, but it is not yet public,” said Ton-That. We believe this product will improve access controls to military bases and other locations that require extreme security.”
An example of scanning the biometric data of a face
The augmented reality glasses for facial recognition will be produced by Vuzixan American company specializing in smart glasses, and they will cost mille dollar.
The number of photos in the archive of the “identified” person will appear on a lens. By touching the temple of the glasses, you can scroll through the images. A tap will allow you to select one and view the associated information. For Thon-That these devices will allow “saving lives”.
But the possibility that anyone could access this technology in the future is frightening. Thanks to one of the many augmented reality glasses already produced today by Western and Asian tech giants. Clearview AI is not offering its archive – for now – to the public. It mainly targets contracts with governments, armies and law enforcement agencies.
But anyone can get an idea of how – and how powerful – AI-based facial recognition works. Just click on the PimEyes website, which offers technology similar to that of Clearview AI at 35 euros per month. Also in this case the research involves an archive of public images collected online.
The study Why artificial intelligences cannot emulate common sense by Giuditta Mosca 24 July 2023
It is difficult to confine new technologies in a fence. Or promise that they will be used exclusively for good. Rules will always be needed. And the common sense of those who use them.
@ppisa