We don’t hear much about facial recognition, probably because some find the subject too technical or inevitable. However, it is essential to be able to form an ethical opinion on the subject. So here are some considerations.
How facial recognition works and uses
How does facial recognition work? several techniques exist but here are the essential principles: a program equipped with artificial intelligence analyzes the presence of a face or not in a photo. If there is a face, 2D and 3D sensors measure the distance between the two eyes. Why between the two eyes? Because after a certain age, your face changes, but never the distance between your two eyes. Once this measurement is taken, the program analyzes hundreds (sometimes thousands) of biometric points (nose, mouth, jaw, etc.) and your face becomes a code. And the 3rd step comes in: this code is compared to all those around you.
There are two uses of facial recognition. ” Who am I ? Is the identification of a person. “Am I who I claim to be?” Is authentication, like unlocking your phone for example.
On the one hand, I authorize the system to compare information that I give it, with another provided before. It’s a technique like any other, even more precise and therefore safer. It seems reasonable to me, framed, I am not against it provided that the information sent for comparison is not recorded (this is rarely the case). On the other side, the system learns “who I am” by recognizing thousands of biometric points.
Some insurance companies are already using facial biometrics to analyze your health and price your premium. It’s efficient, but the most impressive remains the personalization of the user relationship: the recognition of emotions. This is facial coding, these programs use the work of Paul Ekman. In 1970, this researcher analyzed the movements of the 42 muscles of the face to define 6 emotions: joy, surprise, disgust, fear, sadness, anger. These are micro-expressions of a quarter of a second, hardly visible to the naked eye. Supermarkets like Walmart use it in their stores, Disney in theaters and more.
The positives and the dangers
Discovering these uses may seem scary, of course, but admit that there is a seductive side! We no longer need a password, keys, identity papers, bank card! And then, we like to be recognized, as in your usual restaurant: the boss, this good friend, welcomes you and sits you at his best table.
Some still claim that facial recognition opens up a safer world. This technology makes it possible to stop terrorists, thieves. Yes, of course, but the price to pay is heavy! This is the whole problem of surveillance. Yesterday we had targeted surveillance on certain individuals. We had enough doubts about these individuals to deploy exceptional means in a given time. Today we have mass identification and surveillance, to spot a few. It even becomes a real “surveillance capitalism” because certain data is sold, bought. Your information comes at a price.
Is it inevitable? I am not the one who will tell you that it is necessary to prevent the development of technologies, sciences, knowledge. I believe, on the contrary, that it is essential. Nevertheless, here is a little personal reflection.
To work, facial recognition uses weak AI. There are three types of artificial intelligence: weak AI is a system that performs the task you programmed it to do, it operates in silos and requires databases to function. It is a question of making him make a maximum of errors so that one teaches him his mistakes. The transverse AI, it can perform tasks that require several programming, it is for example the robot assistant of the human who would decide to wash the windows, or to cook … And then the strong AI that transhumanists imagine capable of thinking, of having self-awareness.
We are now in the control of weak AI, from which facial recognition is derived. Yes, in the next 5 years the demand for facial recognition or technology requiring weak AI will be significant, it is the era of surveillance capitalism. All data is necessary to reduce the risk of failure. But in 5 to 10 years, it will be transverse AI. We will then have exceeded the need for monitoring. For example, the autonomous car today requires a very important monitoring for the millions of kilometers that it is made to improve. But once perfected, there is no need to monitor the car on the road.
French companies can and should choose to reject surveillance capitalism and certain aspects of facial recognition such as identification, while investing in the future. IBM did it.