Artificial Intelligence Just Defeated Top 100 UK Lawyers

CaseCruncher Alpha, a legal inquiry scoring system, has proven more accurate in dealing with real claims than the top 100 private lawyers in London. The competition was informal and was intended to assess the effectiveness of a new tool based on artificial intelligence.

AI and human lawyers were given a relatively simple task: based on real data on disputed payments, to assess the likelihood of success when filing a claim under the payment insurance system. In total, 775 predictions were made on each side and the AI ​​accuracy was 86.6%, while in humans it did not exceed 66.3%. For those who want to order a legal service, this is a significant and obvious difference.

As a result of the competition, the BBC journalists raised the topic: whether in the coming years AI will replace junior clerks in law offices who are mainly involved in processing statistical information. The answer is no, but I really want to. Machines are guaranteed to outperform humans if the problem statement includes an "exact answer", but they are not able to independently revise the conditions of the problem and think flexibly. But most of the small employees do extremely boring, routine work and, in addition, very slowly.

CaseCruncher Alpha is the brainchild of three Cambridge University students, Joseph Marushak, Rebecca Agliolo and Ludwig Bull. It is noteworthy that they all study at the Faculty of Law and do not have deep knowledge of IT. Initially, they designed a chatbot for quick answers to typical questions about lawsuits, but then they extended the algorithm and adapted it to analyze rather complex cases.