[MILIPOL REPORT N°2] Artificial Intelligence and predictive analysis

On the third and last day of the conferences, experts gathered to tackle the challenges involved with the Big Data – how should we deal with the terabytes of data generated every day on French territory, and what limits should be applied to the development of artificial intelligence? Should be concerned about the resilience of democracy in the face of predictive analysis?

ROUNDTABLE: Can the combination of predictive analysis and artificial intelligence be a way to deal with Big Data?

Prefect Renaud Vedel, the ministerial coordinator for artificial intelligence, started by re-iterating the missions of the Ministry of the Interior; the latter monitors any legal transgression, notably thanks to new technologies that one must learn to master. According to the Prefect, it is necessary to know how to combine the rules on which the Ministry relies with data processing in order to master these new technologies. In view of the hyper-connectivity that characterizes the French territory – 92% of the French are connected today – intelligence tends to favor the use of artificial intelligence, in order to process the growing number of data to be managed. An attack, for example, generates terabytes of data, which represents the work of about 50 people. Artificial intelligence could, for example, be used to translate conversations between orchestrators, often of international origin. However, the technology is poorly portrayed in the public debate, with concepts such as Minority Report-style predictive policing overshadowing other innovations. Laurent Allais, president of the Agora for Paris Security Directors, highlighted the tendency to attribute “quasi-magical” properties to AI, while the reality is more complex: data analysts need to know what data to retrieve, how to classify it, and train the technology to achieve better results. Luc Manigot, Director of Operations at Sinequa, reinforced this point by stating that AI is “human before being robotized“, since we write the programs and also provide the data to be managed. User confidence then comes down to the processing of the data and its life cycle, for which the responsibility is human as well.

According to LREM deputy Pierre-Alain Raphan, the real hidden cost of AI lies in the fact that the technology is replacing workers, whose tasks’ automation has erased the stimulation out of the job. Luc Jouve, president of GPMSE Installation, took the example of Crédit Lyonnais’ remote monitors, whose work unveils around 99% of false alarms upon procession of the information. AI would make it possible to assist in the processing of an incredible amount of data, but also to “hyper-personalize” the customer relationship. In this respect, facial recognition can assist armed forces, as it does for example in Sweden, where it helps identify stadium bans. The president also deplores the mistrust of the French public towards AI, evidenced by the strong criticism of the CNIL, or the Quadrature du Net against the government project “AliceM”, designed to digitize public services using facial recognition.