Machine Learning and Artificial Intelligence (AI) are certainly all the rage today. AI will touch everyone, it will change lives and careers in a matter of years, and definitely in your lifetime. Andrew Ng, chief scientist at Chinese Internet search giant Baidu and co-inventor of the Google Brain, explains it succinctly in his interview: it used to be that generations spent their lives in the same trade. Then rural workers moved to the cities in a lifetime. And now technologies change in ten years or faster. A truck driver may need to learn software development in the ordinary course of his career. I will add: Certainly, lawyers can practice machine learning.
Rees Morrison in his recent article explains how law firms can “draw themselves aces,” using the poker metaphor, that is, reap multiple benefits – and it is not by buying more technology. After all, lawyers are paid for their knowledge. Wouldn’t it be logical if they get paid a bit more for knowing how to apply machine learning?
The benefits one can expect are increased billing, better clients, more winning strategies, and being thought of as a progressive law firm. What’s the magic? I will explain. Keep in mind that today there are many articles about machine learning, but often they are foggy and leave the reader more confused than ever before. It is this fog that I want to dispel.
Let us start from the end – that is, from the results. Law firms are sitting on treasure troves of data that they already own. This data includes their case accounting, with all the details. It also includes data in their respective areas of expertise, such as the types of patents they have worked with, or the characteristics of the participants in an employment litigation. All this data says to you, in the language of Shakespeare:
And thy fair virtue’s force perforce doth move me
On the first view to say, to swear, I love thee.
That is, the data wants to tell you who your best clients are, or what billing rate brings you the most overall income, or who of your people should be given which cases. It is there for the asking – but you need to know how to ask. Let us start with a few basic definitions.
We have finally come to the question, what is machine learning? – It is a branch of applied mathematics that allows you to make predictions based on previous observations. Here is an example. Let us say you want to rent an office, and you want to know if the price that you were offered, in dollars per square foot, is fair. You could ask your friends or look at the prices in the area – but this would be subjective. In machine learning, you would collect the data from known rentals: office sizes, locations, lease lengths, history of problems, and other related items. Then you will derive a simple formula that will compute price per square foot, based on this data. Technically, you will use linear prediction model, part of machine learning. If your formula gives a good fit, and the prices that it calculates agree with the prices you observe, you can use this formula to predict the price for your office, and see if it is too high or too low.
Now you are a data scientist, and you are doing machine learning. If you are a landlord, doing this kind of machine learning is even more important, because it will assure that you are setting prices correctly. If you are indeed a landlord, you can do one more step: you can maximize your total income. This is how: you collect the data on your previous leases. If you can get other people’s data – even better! Then you calculate the probability of getting the deal at each of the possible prices and come up with the price which will give you the maximum expected income.
Now that you have tried the data scientist hat, how do you implement this in your law firm? Today, you are bombarded with “we use machine learning” and “we use artificial intelligence in our products” by vendors. Indeed, a task like automating privilege review in eDiscovery may be more than you would want to bite, and it is better left to the vendors. We are talking about the other end of the spectrum – doing those machine learning tasks that only you can do and which will give your firm unique advantages.
Rees Morrison suggests that you need
- Champion – someone to lead and push the project. She needs to be persuasive, understand technologies, and ideally sit on a data trove.
- Data – collected from various sources and cleaned for easy processing. Or your tools will do the cleaning.
- SME, subject matter expert, who understands the business of your law firm.
The final Rees’ recommendation is a programmer, and here I take issue. In my experience, hiring a software developer in many cases leads to frustration. He needs to understand the language you are talking, and misunderstanding here is crucial to the success. If somebody at your law firm codes for fun, it could work better.
But what is more important, you do not necessarily need a programmer. Today, there are many data science and machine learning tools that do not demand coding skills. It is enough if you understand the problem, and the tools will do the math for you.
I have been teaching Machine Learning for a while, and recently we at Elephant Scale have developed a course called “Machine Learning for Lawyers.” It is an old joke that “I went to law school precisely so that I don’t have to do math,” but that joke is indeed old by now. Lawyers and legal IT people that I meet are often highly technical. Low-hanging fruits are there for you, and even a little learning will be amplified by the tools at your disposal. Here is a list of tools that do not require you to write code. Besides, Amazon, Google and Microsoft all offer machine learning as a service in the cloud. Sky is indeed the limit.
