Back Home 5 News 5 ChatGPT: what does accurate AI-powered machine learning mean for lawyers?

ChatGPT: what does accurate AI-powered machine learning mean for lawyers?

9 Feb 2023

| Author: Steven Moe

It’s been a long day. As you are shutting down your computer a client calls and asks for an urgent memo on a complicated legal issue. You sigh deeply as the sun is setting. You would rather be at home, and this will take lots of research and effort.

But instead of starting work on the memo, you use a voice activated assistant to state the question to an interface which spits out a comprehensive answer within seconds. Science fiction or a coming reality?

This scenario has been teased and talked about in various forms for a while now, often with comments on the demise of lawyers. After all, the client could easily get the answer himself. I think lawyers will be needed for a long time, but let’s look at the latest developments in this area to get a sense of where we are and where things are headed.

OpenAI is a company which was founded to focus on artificial intelligence with a mission “to ensure that artificial general intelligence benefits all of humanity”. It’s governed by a not-for-profit but also has investors such as Microsoft and is fundraising on a reported valuation of US$29 billion. The company has hit the headlines due to the release of ChatGPT in November 2022, a conversational chatbot that generates text responding to questions.

The GPT stands for “Generative Pre-trained Transformer” and the reason for the hype is the speed and accuracy of the answers. Also capturing attention is its ability to respond with poetry or song or computer code. Just this week Google has announced its response to ChatGPT, in the form of Bard AI. The pace of adoption is likely to increase as these tools become more available.

So, what does accurate AI-powered machine learning, of which this is just one example, mean for the legal industry? First let’s address some lazy use of terminology. “AI” is a broad term that needs to be broken down. Artificial Intelligence is often used to describe things which are just complex algorithms rather than actual intelligence and consciousness.

The term “general AI” is used to differentiate those more basic abilities from the AI that can understand things and make its own decisions. AI has not reached that point…yet.

Just a parrot?

Critics seize on this point, noting that these forms of algorithm-driven programs, which have been trained and tweaked with human input, are essentially operating like a parrot, drawing on vast quantities of source material (literally hundreds of billions of words).

Where it gets truly interesting is if the parrot understands what it is saying and not just regurgitating content. The co-founder of OpenAI puts it best, saying in a tweet, “ChatGPT is incredibly limited, but good enough at some things to create a misleading impression of greatness.”

Since Microsoft has invested in OpenAI it seems likely that the software none of us can easily avoid – Microsoft Word – might soon include features that are based on ChatGPT-type innovations. For example, what if there was an ability to have assistance with inserting specialist clauses directly into a contract after it considered your context, and had evaluated millions of other clause examples? Law can be incredibly nuanced, so even then I still see the place for a lawyer to be steering the ship of such drafting, and for a long while to come.

Three questions

At present, using ChatGPT is free so I registered and asked it a few questions to experiment with how it might be used to prepare summaries of information:

  • First, I asked it, “What is the role of the Auckland District Law Society?” Its instant response was, “The Auckland District Law Society is a professional association for lawyers in Auckland, New Zealand. It was founded in 1857. The society’s main purpose is to ‘promote the administration of justice and the interests of the legal profession.’ It does this by providing services and support to its members, and by advocating for the legal profession and the interests of justice.”
  • Next, I asked it to write a summary of the court system in New Zealand. It responded within a few seconds: “New Zealand’s court system is divided into two branches: the district courts and the High Court. The district courts are the main courts of first instance and deal with most criminal and civil cases. They also hear appeals from some tribunals. The High Court is the highest court in the country and hears appeals from the district courts. It also deals with certain criminal cases and applications for judicial review. All courts in New Zealand are independent from the government and the judiciary is appointed by the Governor-General. The Supreme Court of New Zealand is the final court of appeal for all matters in the country, and its decisions are binding on all lower courts.”
  • I then asked it to add sarcasm and it responded with this addition: “So if you’re looking for justice, you’ve come to the right place! Our court system is the perfect example of efficiency and fairness – unless, of course, you’re a defendant.”

I was impressed. It seems to be like a much smarter version of Google because it doesn’t simply give you links to sources of information – it provides you with the actual summary itself. In other experiments the summaries were generally accurate although some details were incorrect. But they were errors I see humans get wrong all the time too, like referring to the Charities Commission instead of the current name of Charities Services.

Other developments in this area are interesting. For example, recently in the United States the company DoNotPay, which arose out of an app helping people get off parking fines, has offered US$1 million to anyone willing to allow their AI app to argue a case in the Supreme Court (with it delivering directions to the human lawyer via an earpiece).

That won’t be permitted due to rules about what can go before the court but it’s interesting to see where the technology is headed. The company says this year it will argue a speeding ticket case, with its AI listening to arguments and instantly telling a defendant what to say in response via an earbud. So, what about for you? Would you welcome an AI assistant that scoured every legal decision and provided an answer?

Perhaps it would free us up to concentrate on novel situations where we can bring decades of wisdom to bear for the benefit of our clients while freeing up time to watch the sunset instead of diving into memo preparation.

I like to think our role as lawyers is to catalyse positive impact for our clients, so tools that might assist that are welcome. It is a fast-moving space that will surely see more innovations in 2023 now the concept introduced by ChatGPT is being mainstreamed. ■

Steven Moe is a partner at Parry Field Lawyers, a member of the ADLS Technology & Law Committee and hosts Seeds podcast with 330 episodes ■

Subscribe to

LawNews

The weekly online publication is full of journalistic articles written for those in the legal profession. With interviews, thought pieces, case notes and analysis of current legal events, LawNews is a key source of news and insights for anyone working within or alongside the legal field.

Sign in or
become a Member
to join the discussion.

0 Comments

Submit a Comment

Your email address will not be published. Required fields are marked *

Latest Articles

Loading...