In the United States, an AI would make it possible to predict the next crime scene


Researchers at the University of Chicago have implemented an algorithm that is supposed to be able to predict crimes a week in advance, with a success rate of 90%.

Each of us probably has the reference Minority Report on your mind. Researchers from the University of Chicago have unveiled an algorithm that would be able not to predict crimes, but at least the places where they could be committed. And this with a week in advance.

The model put in place would have two functions: to predict crimes, but also to reveal shortcomings in police interventions in the United States. Indeed, the study highlighted weakened police protection in some poor neighborhoods of several large cities, including Chicago and Los Angeles. The purpose of the algorithm would therefore be to optimize the interventions of the police by adding “to a toolbox of urban policies and policing strategies to fight crime.”

Professor Ishanu Chattopadhyay, head of the research team, explained to the scientific magazine New Scientist that law enforcement resources were not “infinite steps” and therefore should be “optimally used”. It would then be “ideal to know where homicides are likely to occur”.

A 90% success rate?

For the prediction of crimes, the tool was tested and validated using historical data from the city of Chicago (collected between 2014 and 2016) concerning two main categories of events: violent crimes (homicides, assaults, beatings and injuries) and property crimes (burglary, theft). In an article published in the journal Nature Human Behaviorthe researchers explain that the precision of the localization of their tool has a radius of 300 meters, with a week in advance and a success rate of approximately 90%.

Besides Chicago, the algorithm would work in seven other major American cities (Atlanta, Austin, Detroit, Los Angeles, Philadelphia, Portland and San Francisco). The AI, on the other hand, would not make any predictions concerning the suspects, to avoid any type of discrimination. The group of researchers has also made its data public, as has its software, so that it can be analyzed by other specialists. Any malfunctions could therefore easily be detected and reported.

Advertising, your content continues below

Advertising, your content continues below



Source link -98