The Oxford English Dictionary defines artificial intelligence (“AI”) as the theory and development of computer systems able to perform tasks typically requiring human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. AI can be applied across many disciplines, but I have chosen to focus on the relationship between AI and the law.

AI has been around since at least the 1950s and is a broad term often used to describe things that are just complex algorithms rather than actual intelligence and consciousness.

AI systems work by internalising vast amounts of data, analysing the data and then providing answers and predictions to specific questions. It is designed to supply the most accurate results as quickly as possible.

The limitations of AI

You may have heard of ChatGPT, a controversial chatbot released last November that can communicate with humans through conversation. Imagine using a voice-activated assistant to ask a question and receiving the answer within seconds. Sounds great, right? Unfortunately, studies demonstrate that chatbots are not always trustworthy or accurate.

In a recent study, ChatGPT answered 95 multiple-choice questions and 12 essay questions in four different legal exams. The answers were marked blindly by professors at the University of Minnesota Law School. The bot passed all the courses but performed on average at the level of a C+ student.

Another study compared American- trained lawyers with years of experience and an AI algorithm. The task was to review non-disclosure agreements (NDAs) and find issues within the agreements. NDAs were chosen as they are one of the most common contracts used in business.

There are about 30 proposed issues that could appear in an NDA and this list was put together by prestigious law professors and veteran lawyers to test the lawyers and AI. The study involved input from academics, data scientists, legal and machine-learning experts, and an independent consultant lawyer.

The results were that AI was faster and more accurate. However, the average accuracy rating was close, with the AI achieving 94% and the lawyers achieving 85%. The speed was a different story because, on average, it took the lawyer 92 minutes, and the AI did it within seconds.

I took from this research that AI technology is a powerful tool as it can answer simple legal questions, but it struggles with more complex ones.

Advantages of a human advocate

While AI can help lawyers immensely with mundane tasks such as research and sifting through court judgements due to its speed and accuracy, it will inevitably struggle with emotions, fears, aspirations, desires and unpredictability. In addition, AI cannot match human intellect and analytical skills for accuracy in fundamental legal work. There is very little room for error in legal drafting, and the law is incredibly complex and continually evolving, so it remains to be seen whether AI technology will keep up.

The best comparison so far is to think about AI as being more like an actor who is playing the part of a lawyer in a film versus getting legal advice from a person who has studied and practised the law. I know which one I would trust when you think about it in these terms.

Currently, chatbots are ill-equipped to compete with their human counterparts because there is no substitute for experience and pragmatism. While lawyers will not be replaced anytime soon, there is certainly scope to suggest that there are some tasks that AI can perform faster, saving a lawyer many hours and a client many dollars. Outsourcing some work to AI seems to be a viable alternative that could be utilised safely, with both parties playing to their strengths.

Conclusion

While the speed and accuracy of answers given by AI is alluring, it has not reached the point where it can understand things and make its own decisions. We need to take care of how we apply AI technology, especially in its infancy. One of the biggest problems with a chatbot now is that while its fluency is exciting, it can sometimes confidently state something as fact when it is incorrect. This can be hard to spot when AI gives an answer that is plausible but not necessarily correct.

AI does not need to compete with lawyers. Instead, it could augment a lawyer’s ability to provide trustworthy advice promptly. This way, you get the best of both worlds: a sociopathic response that incorporates the effectiveness and efficiency of AI. One day, it could go beyond regurgitating content and actually understanding what it is saying. For now, the best way forward appears to be a partnership. Lawyers are still best placed to deal with the subtle nuances of the law. But it must be acknowledged that those who do not upskill to utilise the technology effectively to optimise their practices will get left behind.