A world first. Chinese scientists from the Chinese Academy of Sciences have developed artificial intelligence capable of replacing a prosecutor in certain tasks. This AI, currently being tested by the Shanghai Pudong People’s Procuratorate, can recognize certain offenses and lay charges. However, this innovation is not unanimous, several prosecutors and specialists believing that its reliability is not sufficient for a real application.
Researchers say they have successfully developed a machine powered by artificial intelligence that can charge people with certain offenses, according to the South China Morning Post, who had access to an article published in Management Review. They claim that she can lay a charge with almost 97% accuracy, relying only on ” a verbal description of the case “.
European Commission approves Microsoft’s takeover of Nuance
A “trained” machine since 2015
Its development started years ago. Researchers have it ” trained Between 2015 and 2020. During these five years, the machine has been able to study 17,000 cases, enabling it today to identify the eight most common offenses in Shanghai. These include the offenses of illegal gambling, credit card fraud, theft, dangerous conduct, as well as intentional injury, obstruction of official duties, fraud and ” arguing and causing trouble “. The latter surprising and imprecise terms are said to be used most often to hush up cases of political dissent.
This ” attorney Is tested by the Shanghai Pudong People’s Procuratorate, which is the largest and most active district attorney’s office in China. Scientists say he will soon be able to recognize other crimes or offenses, more complex, and bring several charges against the same person.
However, its reliability does not convince legal experts. ” The accuracy of 97% may be high from a technological point of view, but there will always be a risk of error. Who will take responsibility when this happens? The prosecutor, the machine or the designer of the algorithm? “, Asks a prosecutor in Guangzhou, in southern China, questioned by the South China Morning Post. ” AI can help spot a mistake, but it can’t replace humans in decision-making He argues.
A real use of this device, and its possible distribution, would also raise the question of the autonomy of prosecutors. This new technology would however have several advantages for them. It would reduce their workload and could significantly improve their speed in dealing with a case. For Professor Shi Yong, main actor of the project, it would also give them more time to devote themselves to the more difficult tasks.
Artificial intelligence is already used by prosecutors in China
Chinese prosecutors have been using artificial intelligence technology since 2016, and many of them are said to be using the ” system 206 “. It allows to evaluate “ the strength of the evidence, the conditions of an arrest and the degree of danger of a suspect to the public “.
Several problems arise when using such devices in the legal field. The AI must be able to identify all the important information in a case and delete all the others. She must also understand human language, which is constantly evolving. It must therefore be translated into mathematical language, but an imprecise translation would mislead the AI and thus lead to bad verdicts. There are already software and other tools that allow this translation, but they require the use of powerful computers. Computers to which prosecutors currently do not have access.
These AI-based tools therefore have a very limited impact according to the researchers. ” They do not participate in the decision-making process of laying charges or suggesting sentences », They explain.
The AI developed by researchers at the Chinese Academy of Sciences can, for its part, be installed on a conventional desktop computer. Concretely, for each case, it is based on 1000 « characteristics »Based on oral testimony. To determine the guilt of the suspect or not, it then resorts to the 206 system. It would therefore be more accessible and more efficient than the existing tools.
China already uses a lot of artificial intelligence to improve the efficiency of its administration or to monitor the population. This technology seems inevitably bound to develop. Faced with this trend, Unesco recently adopted the first world standard on the ethics of artificial intelligence and the European Union is working on a common regulation on this subject. The objective being in particular to lay down a framework to avoid drifts.