What are the current mature applications of AI in the legal industry?
He Qianming is the head of the Greater Bay Area at City Development Law Offices.
At present, the mature application of AI in the legal industry mainly has the following aspects.
First, intelligent legal advice. Traditional legal Q&A is search-based and highly accurate. In many areas, the accuracy rate can reach 90%-95%. For example, asking about labor disputes, marriage and family, criminal offenses, and corporate compliance issues can be 90%-95% accurate.
With the advent of ChatGPT, people developed deep question answering systems that are more capable than search-based systems, capable of multiple rounds of conversation, and have the ability to understand common language, which greatly improves the accuracy of answers.
Traditional search-based question answering can only perform one question and one answer, while generative systems can conduct multiple rounds of question and answer, but there is uncertainty in multi-round question answering, so the biggest difference between the legal big model and the general large model is the verification procedure that generates the results, which requires a large amount of vertical data for training. The Q&A system has been put into use and has worked well.
Second, combine the legal field database with the question and answer system to provide tools similar to the legal version of ChatGPT. When encountering legal issues or cases, enter the case situation, and the system retrieves the relevant cases and regulations and generates answers.
Third, it involves text generation, such as generating contracts and other legal documents. Training in this area is ongoing. Although ChatGPT has certain text generation capabilities, its professionalism still needs to be improved in the legal field, so it needs to be trained based on open source models.
