An artificial intelligence is preparing to help a defendant defend himself in court

For the first time, a artificial intelligence will defend an accused on questions of law in court. A company has developed a specially trained program to tackle the legal field.

“ The world’s first robot lawyer »… This is how the company DoNotPay describes the artificial intelligence specialized in law that it has developed. The latter is assumed, if we are to believe their websiteallow any user to ” fight corporations, defeat bureaucracy and prosecute anyone at the push of a button “. Their creation should be tested in a real situation in February, before a court of law. The AI ​​will be present alongside a defendant in the form of a smartphone, which will capture conversations. The AI ​​will then whisper to the accused what to answer by dictating it to him in a headset.

The specific case and the location of the judgment are kept secret by the company. However, she hinted nearby New Scientist that it was a case of speeding. The company considers this trial as a trial run, and has agreed with the “tester” to pay the fine if there should be one. DoNotPay has also played a little with the rules to ensure that it can carry out its test. Indeed, this type of “advice” directly uttered to the defendant would be illegal in most courts. The company founder, however, claims to have found a place where he could legally qualify as a hearing aid.

Originally, DoNotPay was launched in 2015 as a relatively simple chatbot. It provided users with legal advice on consumer issues, based on standard conversation models. In 2020, OpenAI released a publicly available programming interface to provide access to the capabilities of GPT-3, its language processing AI. This is what allowed DoNotPay to turn more towards AI for its tool.

More than a hundred legal specialties

The AI ​​has been trained on a large number of cases already dealt with in court to enable it to acquire “expertise”. According to the founder of the company, it is now able to tackle a large number of cases, ranging from divorce to immigration laws, including copyright protection or parking tickets… More a hundred specialties are cited on the website, and the company claims to have intervened in some 3 million cases in the United States.

Joshua Browder explained that he had tested it himself in various key situations: thanks to the use of a synthetic voice, he had his AI speak to a bank in order to negotiate the reimbursement of expenses for a amount of 16 dollars. In an interview conducted by Bloomberg, he says the technology has very high success rates when users use it. The reason given: below $10,000, big companies know they can get away with scamming people, because nobody fights back. This technology is perfect for this type of case, because it involves very clear disputes “. However, its creators had to face many challenges. In particular, they had to make sure to control the reactions of the AI ​​so that it tries to defend the cases in a fair way, and not in the most optimized way to win…