America's best lawyers fought against artificial intelligence and found out which of them would handle the interpretation of legal documents faster and better. As a result, the robotic system turned out to be more savvy than experienced specialists.
An experimental competition was held on the online legal platform LawGeex under the supervision of professors from Stanford University, Duke University School of Law and the University of Southern California. 20 highly qualified lawyers and a specially trained system fought for the right of the “best”. In four hours, the participants had to study five non-disclosure agreements and identify 30 violations, some of which demanded an arbitration court and damages. They were judged by how accurately they could identify each case.
As a result, the neural network achieved 95% accuracy in the interpretation of various clauses of the agreement, while the accuracy of specialists was 85%. Moreover, the artificial intelligence completed the task in just 26 seconds, while the lawyers took more than an hour and a half - 92 minutes. The highest indicator of a computer is 100% accuracy in solving one of the questions, while in humans it did not exceed 97%.
Grant Gulovsen, an intellectual property expert and one of the contestants, said the challenges they faced were very similar to those he and his colleagues have to face every day. By the way, he admitted that he does not feel defeated, as he believes that machine learning in the future will facilitate the work of lawyers and give them the opportunity to focus on those cases that will always require a "human" approach.
Erika Buell, one of the initiators of the experiment and a professor at Duke University, agrees with him. She is confident that the machine will be able to perform monotonous work without any problems, but only people will continue to consult. “I think law students and aspiring lawyers need to understand how artificial intelligence works in order to use it in their work,” says Buell. "And customers will be delighted if the computer helps solve their problems faster."
Recently, consultancy McKinsey published a report entitled Jobs Lost and Found: Workforce Movements During Automation, which said that within 13 years, an estimated 800 million workers around the world could be replaced by machines. According to the authors of the report, it is already possible to automate about 60% of the work performed by people. At risk are mainly catering workers and machine operators, but bots are less likely to replace gardeners, plumbers and nannies.
Evgeniya Chernysheva