Faception – AI Powered Facial Recognition Technology

You won’t find more enthusiastic supporters of the Fourth Industrial Revolution than us, as long as the technology is solid, the business plan is sound and society progresses toward a better future for all of us. And then there’s isht out there that scares the bejesus out of us, stoking our worst dystopian fears, like the return of Roseanne Barr to prime time TV. Israeli startup Faception and its facial recognition technology is another one of those things.

Started with a very modest Seed round of just $125,000 in December 2015, Faception followed with an undisclosed Series A last summer. The company came on our radar while researching a recent article on the use of artificial intelligence and real-time object recognition Terminator-style. But rather than being able to identify pretty birds, the AI algorithms developed by Faception can spot terrorists and other baddies like pedophiles, white-collar criminals, and brand promoters.

Faception – AI Powered Facial Recognition

The company claims that our personality is encoded in our DNA, which is reflected in our face. It uses AI techniques such as machine learning and computer vision to train its platform on different facial types, which it lumps into 15 categories it calls classifiers, such as terrorist or brand promoter (described on the site as someone “endowed with a high self-confidence, authoritative, charismatic and magnetic personality, with high intellect and high verbal ability”).

As the website Snap Munk explains further:

“For each of the people groups that need to be identified, the team behind Faception develops what they call, ‘classifiers’: a particular set of facial features that is consistent with the members of each group. They arrive at these classifiers by studying images of people known to be members of those groups, and identifying recurring attributes in their facial features. Once they find features that are a motif, they run images or video streams of general populations through the software to study for those attributes.”

The Faception Process

Biometrics or bias?

All of this sounds suspiciously like a high-tech version of racial profiling, by teaching a machine how to be prejudiced against people with a certain look. Faception boasts that its software was able to identify nine of the 11 terrorists in the 2015 Paris attacks. Of course, if you train your big-brained machine that people with dark complexion and dark beards, with dark, grim looks on their faces are terrorists, it’s probably going to identify photos of said grim-faced people who match that stereotype as, well, terrorists.

The problem of AI prejudice has been well documented. For example, in 2015, Google’s photo app tagged some black people as gorillas. An investigative report last year by ProPublica found software used to predict future criminal behavior—yes, just like in “Minority Report”—was biased against minorities. An AI-judged beauty contest from a company that uses machine vision and AI to study aging overwhelmingly picked white people over others in a throwback to Hot or Not.

However, we’ve already seen what AI can do when it comes to detecting human emotion or identifying your doting grandmother in your family vacation photos. The technology from Faception even caught the attention of The Washington Post, which reported that the company claims its 15 different classifiers evaluate with 80 percent accuracy certain traits. Faception’s facial recognition technology has reportedly found its way to at least one governmental homeland security agency, to the tune of $750,000. (We’re not sure if it’s the Department of Homeland Security, which at one point was working on its own Minority Report pre-crime detector, Future Attribute Screening Technology, based on remotely measuring a person’s physiology to determine malicious intent.)

Faception says its facial recognition technology isn’t just to help people identify criminals and bingo players (yes, that’s really a classifier, along with poker player). It can be used in sectors outside of security. For example, a robot plugged into Faception’s classifier system could customize its interaction based on the personality type. Another application that the company flogs is financial services. In this case, its algorithms can detect risk-takers or dishonest people, which would influence whether a bank might give that person a loan. And, of course, retailers could use the technology to gauge whether a customer is someone who might be a compulsive or impulsive shopper. Faception lists Sears and Manpower as clients, according to Computer World.

Facial recognition biometrics

Facial recognition, which is the general category that Faception falls into, is one of the leading biometric categories, with dozens of companies competing in the space. The entire biometric market could be worth $24 billion by 2020, as we recently learned. The difference between most facial biometric recognition startups and Faception is that the latter uses machine learning to teach its algorithms to detect certain types of people. Most of the other companies match faces to an existing database, usually for security and authentication purposes.

FST Biometrics

For example, another Israeli startup called FST Biometrics combines facial recognition and behavioral analytics to enable authorized individuals to be identified at a distance while in motion. In other words, you don’t have to stop and have your face scanned at a kiosk. The company has raised $20 million to date and was founded by a former head of Israeli intelligence, Aharon Farkash, and counts former Israeli prime minister Ehud Barak as a board member. Its trademarked In-Motion Identification (IMID) technology comes in different scales, including one for mobile devices and another system to lock down a remote area, such as a crime scene.


Click for company websiteOne of our readers turned us onto a London-based startup called iProov, which is scant on financial details. Its iProov Verifier is one of an increasing number of facial recognition programs that promises to replace online passwords. The company throws out the usual AI nomenclature—machine learning and deep learning—to market its platform. Biometric authentication apps are expected to be big business, with predictions of more than 770 million downloads per year by 2019, up from six million today—a nearly 13,000 percent increase.

Nirenberg Neuroscience

Click for company websiteMobile apps are cool and all, but what New York City-based Nirenberg Neuroscience is doing with machine vision and machine learning is a wee bit more sophisticated. More than just a facial and object recognition company, Nirenberg Neuroscience is basically building a bionic eye for the blind. The work is based on research by neuroscientist and founder Sheila Nirenberg, a professor at Cornell. Her company is one of four AI startups tapped last year by Ford Company to help it develop a fully autonomous ride-sharing vehicle. (Ford is pretty serious about this venture, just last month investing $1 billion in Pittsburgh-based Argo AI to steer development of its autonomous vehicle fleet.)

Finding Rover

Upload Rover to the app for facial recognition.

We’ll end this round of investigations into facial recognition startups with a less scary company. Finding Rover is selling a different trick with facial recognition software. The California-based company, which has about $2 million in venture capital from March 2014, has developed a facial recognition platform for dogs. Users can register their precious pets for free, snapping a photo that is entered into a database. The finder of a lost pup can send a photo of the animal to the app, which does a facial recognition search to match the dog with one in its database. Owner and pet are reunited—and a future dystopian hell-scape is avoided.


Leave a Reply

Your email address will not be published.

  1. What distinguishes iProov (www.iproov.com) is not its facial recognition software (though that too is cutting edge) but its patent protected anti-spoofing and anti replay attack technology. iProov superimposes a “flash mark” over each and every log-on. This creates what the company calls a “one time biometric.” If the flashmark is not present, the log-on is a spoof, probably a photo; if the flash mark is the wrong one, then the log-on is a video replay attack.. These defenses also mean that unlike competitor offerings, would be attackers can be quickly identified and locked out, preventing them from mounting repeated attempts (using machine learning) to probe for weaknesses in the underlying facial recognition software