Artificial intelligence arouses fears and hopes. The automation of certain tasks could eliminate 300 million white-collar jobs in the coming decades. Law seems to be one of the professional fields that could be the most affected. At the same time, new analytical capacities are already offering new ways to tackle certain social problems, such as school dropout. Welcome to your near future…
One of the hopes raised by ChatGPT is that it will make the justice system more accessible to everyone. Limited by their wallets and often unable to unravel the law and its complex procedures on their own, citizens dread the courts. A “ChatGPT lawyer” has the potential to answer litigants’ questions, guide them through their disputes and even draft their pleadings, but specialists are already raising a number of ethical questions — and potential pitfalls.
The tool appeals because there is a huge gap in access to justice in Canada, established Ryan Fritsch, a lawyer working for the Law Commission of Ontario.
She hosted a recent webinar with the University of Ottawa’s Center for Law, Technology and Society to tackle a number of questions head-on, including this: ChatGPT, a type of language, backed by artificial intelligence (AI), will undoubtedly impact the legal profession, but will it be for better or for worse?
The number of legal questions that can be asked of ChatGPT (a software company OpenAI) is incalculable: do I have a case against my employer who fired me? Can I initiate divorce proceedings or a formal notice? And even this possibility, which is no longer science fiction: having a headset in the classroom and letting ChatGPT – or a similar tool from another company – whisper the answers in our ear to repeat them better judge.
And without a lawyer’s bill.
“It can save time and money,” summarized Me Fritsch during the webinar.
Here too, the tool, still in its infancy, was put to the test: after asking a few legal questions, Me Fritsch found that he was “put on the right track”, but sometimes the information was outdated or too generic to be useful, and even, utterly audacious, that ChatGPT outright made up articles legal or case law to support their answers!
The University of Minnesota Law School also got in on the action and had ChatGPT pass the end-of-year exams: it came away with a modest overall C+ grade. “Not bad, but is it enough? asked M.e Fritsch. The university would have made a close follow-up of this student by imposing strict conditions of success for the next year.
OpenAI, however, reports that its more powerful version of the tool, since made available, “GPT-4”, passed a bar exam and ranked in the top 10%.
The benefits envisioned are tangible: “It could empower individuals,” argued Daniel W. Linna Jr, a professor at Northwestern University in Chicago and director of Law and Technology Initiatives. The AI could quickly draft contracts or pleadings after ingesting all the relevant information.
ChatGPT can also help make lawyers more successful. University of Ottawa law professor with a specialty in ethics Amy F. Salyzyn sees this as a breakthrough that could help build powerful tools for the profession, including quickly summarizing and extracting information from case law. “We are at the very beginning of this story, and it will be exciting to see where it will take us. »
But enthusiasm quickly gives way to concerns.
One of the worst fears is that these tools and their responses incorporate prevailing prejudices and discrimination, warns Mr. Linna Jr, and that this harms the most vulnerable populations.
Because American studies have already shown that assessments of the potential for recidivism and dangerousness of prisoners during the parole process – which uses AI in the judicial system – particularly disadvantage blacks. To the point where the Attorney General of the United States, Eric Holder, warned in 2014 that these systems could negatively affect poor and racialized citizens.
And then, “if citizens receive inadequate legal information and act on this advice, it could be a problem,” warns Professor Salyzyn. It also raises the question of the ethics of using these tools: will a lawyer be criticized for using them rather than resorting to more traditional research methods? Or, on the contrary, will he be blamed for not having used it as a tool at his disposal?
And this headset that would answer the judge’s questions? Joshua Browder, CEO of American company DoNotPay, which started out as a parking ticket dispute app, wanted to put it to the test. His first attempt in California, using a tool that uses both ChatGPT and DaVinci, raised the ire of bars in several states. A lawsuit has also just been brought against the company for illegal practice of the legal profession.
Not to mention the risk that confidential client information could end up in the public sphere if the lawyer included it in his questions to obtain the most precise answer possible: a possible breach of confidentiality and a breach of the lawyers’ code of ethics. .
For his part, M.e Fritsch worries that a two-tier system is being created in which free tools do not perform as well as paid tools, and which only large firms — and wealthy clients — could afford.
If the experts agreed during the webinar that ChatGPT is not yet mature enough to put lawyers out of work – it invents case law! —, they believe that it will undoubtedly serve to create and integrate new tools into the practice of law. Companies have already jumped at the chance: Casetext has developed CoCounsel, a kind of “junior assistant” who can do legal research.
And why not ask ChatGPT directly if he can one day take the place of a lawyer?
“Not likely,” he replies.
“The practice of law requires not only the application of legal rules, but also judgment, interpretation and appreciation. Lawyers must also consider ethical and moral considerations, which requires a level of human judgment that would be difficult for AI to replicate. »
But AI could play a role in researching and reviewing contracts, he adds, echoing the webinar’s legal experts.