A Lawyer's Guide to Using OpenAI's API (Part 1)
A great excuse to get familiar with code.
[This is the first of a multi-part series on how lawyers are best positioned to ride the next technology wave, and what skills you need to build to be ready. This series was free to subscribers of Future of the Law and for a limited time. The motivation to start this series came from your responses to this post. Part 1 continues to be free to read, and the rest of the series is available here.]
Learning used to be harder, now its not.
In early 2020, I took a course in learning how to code for the first time in 20 years, most of which I had spent in a legal career. Applying technology to legal problems means learning how natural language processing (NLP) works, and then understanding how machine learning (ML) and AI models work.
To tackle the first, I took a basic course on how to read and write code using the Python programming language, which my research indicated is best suited for solving NLP problems. Understanding ML and AI, on the other hand, turned out to have no connection with computers and programming languages whatsoever. It's all about math, geometry, statistics and calculus. For those who might be intimidated, Andrew Ng's seminal courses are presented both theoretically and technically; even if you fail in detail, you gain in perspective!
At the time, I was in two minds about investing this time. On one hand, while it's one thing to understand, it's another to excel. On the other hand, without enough understanding of the technologies behind products you want to build, how would you bridge the gap between the consumer (whom you understand) and the solution (dependent on technology)?
Time has answered this question for me, and how!
Language models are changing the world.
Then, with my basic knowledge of programming, all I could accomplish was reading code or attempting to read code, often struggling my way through. Writing would take ages compared to a seasoned professional who does it daily.
Now chatbots can write code for you.
You just need to be instruct them in human language in the manner of a fine doyen of ideas. And then, check their work by being able to read, correct, and review.
A basic understanding of programming can now take you a far.
AI is levelling the playing field.
Software is now truly getting to the point where it is about ideas and execution, not as much about spending money on gatekeepers - coders who know the "language."
Anyone can build!
Lawyers are already "programmed" for code.
Legal drafting has much in common with programming.
Good legal drafting involves ordering language in a logically consistent way, avoiding ambiguities of language, creating and applying rules, and predicting how rules will operate in real-life problems. Programming is no different.
Programming and legal drafting also follow similar core strategies, such as defining terms and using them consistently, proper punctuation/syntax, hierarchy of rules, niche libraries, and recognizing that slight changes can have drastic consequences.
Code has the power to turn words into color, while legal opinions can strip a person of their life and dignity.
Why lawyers haven't found coding approachable is not because the subject is alien to them, but because it's in a language they don't understand.
How good of a lawyer would I be if all the laws, legal opinions, and judgments suddenly switched to Mandarin? I might be able to solve a problem if I could find its description accessible and have my opinion translated. All this would make the whole exercise seem expensive, tedious, and unapproachable, better left to others, and I would rather drive trucks or write blogs for a living.
AI is galloping toward you, learn to ride.
What if I told you there is an approachable entrance into programming that will, at the same time, be fun and useful to you as a lawyer and not only that - it would dramatically transform you into a capable programmer capable of applying your "new" knowledge to legal problems? Fantastic AI is out there, coming your way, ready to carry you on its shoulders if you're just good enough to recognise the opportunity?
Here's what you need for the recipe, assuming you know nothing about technology but can use a computer (type, move the mouse on the screen):
Willingness to fumble through 3 hours trying to get this right.
A computer or laptop not more than a decade old with 1 GB of free space.
An internet connection.
USD 2 or INR 160 on your credit/debit card for OpenAI.
Knowing you're not too old for this and that this is a small step forward, which is also a giant leap, depending on how you look at it.
A lot of this time will involve boring tasks like figuring out how to install and set up your system. Remember that gatekeepers have moats and castle ramparts that you need to cross, but once you do… you're in the castle!
Programming isn't approachable to most of us because there isn't a pre-installed app on our computer that just works and invites you to a playground.
Doing it yourself means you start getting a sense of the nuts and bolts while also equipping yourself to take on greater challenges as you go.
Not to mention - OpenAI's API is super fun, easy to experiment with, will be the base to understand other language model and their APIs, accessing which will be considered an essential computing skill in the years to come.
If you are ready to do this, I am ready to write the next post in this series which will focus on, hold your breath.. installation!
P.S: For those of you following my work on the Founders’ Toolkit, thank you for the fantastic response, there will be updates soon!