Welcome to a new mini-season of Read Law Right, where we pivot from reading skills into the practical task of writing authoritative legal arguments. So it is time to address the elephant in the room that I have ignored so far – generative AI.
Looking back over the first two seasons of this blog, gen-AI is conspicuous by its absence. We’ve discussed reading the textbook, law reports, academic journals and more, and my advice has always been based on what you need to focus on when you do your reading.
That guidance has been highly practical, giving you the tools to efficiently read and digest the copious amounts of content you will be required to address as both a student and a professional lawyer.
Never once have I suggested asking gen-AI to summarise a legal document for you.
Nor will I ever suggest that you ask gen-AI to write any law for you.
Yet you will be aware of your peers using gen-AI to give them ideas and even write their assignments. You may even have done so yourself or at least been tempted.
In this first post of the mini-season, I’m going to explain in laypersons terms how gen-AI constructs its output. This may start to undermine the confidence you may have in it to produce reliable legal content. But my central thesis is not that gen-AI is forbidden fruit that you shouldn’t touch.
Then, in subsequent posts, I will focus on three more reasons why you should embrace the hard work of understanding, analysis and evaluation – professional reasons, academic reasons and personal reasons – and not outsource this to gen-AI – which given the proliferation of gen-AI use, will potentially set you apart from the herd in years to come.
Finally, I will offer some guidelines for using gen-AI and some questions you should ask yourself before employing it.
But first, I’m going to summarise briefly how gen-AI constructs its output. Once you know how it works you will understand why we cannot take as given everything it says, and why it is essential be in a place where you have the capacity to reliably evaluate its output.
Generative-AI works by repeatedly determining the next-most-likely word that makes sense in the context. That context is based significantly on the prompt you feed it, plus the rules it has been provided with by its owner. Gen-AI most simply put is an immensely sophisticated word-prediction machine.[i]
What gen-AI can do is undeniably impressive, however the statistical nature of how it produces its output should serve as a warning to those using it for material that you will rely on professionally or academically. It will sound plausible, even convincingly authoritative. It may even be correct a substantial proportion of the time. But as a student of any subject, you will still be developing the tools you need to assess its accuracy. Developing those tools is one of the most important reasons you are studying at university.
And if you are in a situation where you are professionally or academically responsible for the AI’s output, you will need to fact-check it step-by-step. That is likely to be time-consuming, particularly if you do not have a baseline understanding to rely on. In other words, you may as well put the hard work in at the front end. I’ll explain what that looks like in the later posts of the series.
Now this hasn’t stopped students, litigants and even lawyers from relying on it – often to their detriment. That’s what we’ll discuss in the following posts, and what you need to do instead. But to start with, I want you to embrace this maxim:
Don’t outsource your thinking to the machine.
In the next article I’ll explore why lawyers who outsource their judgement to gen-AI often discover that responsibility cannot be outsourced with it.
Subscribe below if you’d like a notification when it is published.
[i] Cal Newport, ‘What kind of mind does Chat-GPT have?’ (New Yorker, 13 April 2023) (Read article)
Enjoyed this article?
Read Law Right is a blog about reading effectively, writing persuasively, and developing the understanding, analysis and evaluation skills that help law students to succeed.
Subscribe below to receive notifications of future posts.
Leave a comment