From October 19 to 22, the Milipol 2021 fair took place at the Villepinte exhibition center in the north of Paris.
The opportunity for internal security experts from around the world to present their latest technological innovations.
At the rendezvous of this 22nd edition, artificial intelligence: between fantasies and questions, this tool has become essential to the elucidation of crimes.
Interview with Jean-Gabriel Ganascia, researcher at the CNRS and specialist in artificial intelligence, who explains its use within the scientific police.
How can artificial intelligence help forensics solve cases?
JEAN-GABRIEL GANASCIA.
Artificial intelligence makes it possible to access a faster response speed, via machine learning for example: a multitude of data is given to the machine which will learn, over time, to process it automatically.
Thanks to this technique, we can follow the track of the vehicle of a fugitive perpetrator through surveillance cameras, the analysis of the videos will be done in a few weeks, when the human will need several months of viewing.
In what other cases can the police use it?
Also thanks to the cameras, it is also possible to find the faces of wanted criminals. The use of this data by the police in the case of facial recognition can be a source of concern, however, it is necessary to distinguish between the two effects of its use: the first is authentication, it is used for s '' ensure that the identity of a person belongs to him. The PARAFE system set up in airports verifies that you are indeed the same person in reality and on your passport, the software performs the same task as customs, and even more efficiently. Authentication does not really present any fundamental problems because the data remains local. The second is more controversial: it is the identification,it is about recognizing the identity of a person in a crowd using surveillance cameras. This type of facial recognition relies on databases that bring together the faces of all people, which poses an ethical problem if we do not ask for the consent of the population. It is not implemented in France. Artificial intelligence has therefore become an indispensable tool for the police, but also for criminals. We are entering a race for artificial intelligence, with criminals who can leave false traces on digital media to mislead the police, it is a permanent technological escalation.which poses an ethical problem if we do not ask for the consent of the population. It is not implemented in France. Artificial intelligence has therefore become an indispensable tool for the police, but also for criminals. We are entering a race for artificial intelligence, with criminals who can leave false traces on digital media to mislead the police, it is a permanent technological escalation.which poses an ethical problem if we do not ask for the consent of the population. It is not implemented in France. Artificial intelligence has therefore become an indispensable tool for the police, but also for criminals. We are entering a race for artificial intelligence, with criminals who can leave false traces on digital media to mislead the police, it is a permanent technological escalation.it is a permanent technological escalation.it is a permanent technological escalation.
And tomorrow ?
In the future, it will likely be possible to automatically register complaints, which will be analyzed by artificial intelligence to direct victims to the right people, and perhaps even draft responses.
The police will also be able to “predict” assaults, that is to say anticipate violent attacks and give the alert, according to compiled data.
But beware: the AI could direct the police systematically to the same neighborhoods.
The risk is therefore the stigmatization of certain places or certain populations.
Human presence then becomes essential because artificial intelligence issues generalities and does not deal with it on a case-by-case basis.
What are the risks and abuses of the use of AI by the police?
We really have to make sure that the criteria according to which the machines make decisions are well controlled to avoid discrimination: imagine a machine that could choose the people to be stopped, it would surely do so on face.
Precautions must also be taken in the use of personal data to avoid a compromise between freedom and security.
The worst would be to follow the nightmarish example of China which notes its citizens by observing their actions.
Even if we are far from this pattern, our role is to ensure that artificial intelligence is not an obstacle to our democracy.
In the collective imagination, when we hear artificial intelligence, we think of a world where machines take control of our life?
It should be borne in mind that artificial intelligence remains a complementary tool to human work, the aim of which is to make it more efficient. Artificial intelligence is conceptually not that complicated to understand, but when we explain that a machine is capable of doing tasks on its own, I find the concern legitimate. It is man who controls the machine and he can let him do what he wants: the risk is not really that the machine will turn against us, but more that it falls into malicious hands. A world where machines take the place of man in spite of his will? No. On the other hand, a company which delegates more and more tasks to machines, which uses artificial intelligence to automate everything, leaving the individual helpless, dormant:that's what we're aiming for. A sort of "laziness" of society.