The Continuing March of AI Through Law

Last month I had the pleasure of presenting a paper at the International Conference on Artificial Intelligence and Law. It was fifteenth such conference held every two years.   Artificial Intelligence and Law thus is a discipline that is already thirty years old. Because of the exponential increase in computation (doubling in power about every 18 months), the theoretical advances made over the years at such conferences are now yielding practical results.

In this initial post, let me mention the most interesting practical application of AI to law at the conference– A Legal Citation Recommendation Engine. This program embeds a research mechanism within Microsoft Word. As a lawyer types his or her brief or memo, the program suggests case law and other material relevant to its arguments. Thus, the program acts as real-time, imbedded virtual legal research assistant.

The inventors of the product hope to distribute a prototype by next year. The program seems relatively rudimentary now. But at their beginning speech-to-text programs were also rudimentary and made many errors in transcription even when I used my PC and an excellent microphone. Now I use such programs to dictate all e-mails into my smart phone even with ambient noise. And the legal citation engine will improve by learning from the way lawyers use its recommendations.

Such programs will continue to disrupt the legal market in the next decades.  They will help extend the reach of the talented in law and put the job of the journeyman lawyer at risk. The lawyer capable of writing creative memos and briefs, particularly in high stake matters, will still be offering valuable services. But more routine work,  like finding authorities and precedents, will be become more and more automated. Of course, it is true that there can be creative use of precedents, but even here we should expect machines to get better.  They will have the advantage of surveying more authority that offers odd angles on a case than humans can easily digest.

As Benjamin Barton describes in Glass Half-Full, his wonderful new book about the legal profession, small law has been suffering for over a decade, in part because routine, low-stakes work, like setting up a corporation or generating a simple will,  has become a commodity that can be automated. Currently, even in big law, the best firms are becoming far more profitable than others for similar reasons. Developments like the program unveiled at this year’s conference will only accelerate that trend. As I tell my students, however, it is easy to have great career in this brave new world: be a superstar!

Reader Discussion

Law & Liberty welcomes civil and lively discussion of its articles. Abusive comments will not be tolerated. We reserve the right to delete comments - or ban users - without notification or explanation.

on July 11, 2015 at 18:33:58 pm

Forgive me but I feel like i just recieved a hand job. I thought AI programs like this already existed? Am I wrong?

read full comment
on July 11, 2015 at 19:27:35 pm

Yes, you are correct! The Black Robes of SCOTUS have been employing this method for years now - what do you think all that fiddling beneath their robes is about, anyway?

read full comment
Image of gabe
on July 14, 2015 at 15:07:41 pm

The next question is: Who is responsible for legal malpractice, the lawyer, or the authors of the software?

This is a general issue for all AI systems. At what point is AI software free enough from human intervention that the software itself is responsible for a harm?

If the AI software is responsible, would it be charged for damages?

There was an automobile accident recently where the car had one of those automatic parallel parking AIs installed on it. The car hit a pedestrian. The car had an option for a backup safety sensor AI to monitor for persons standing behind it, but the owner had not installed this option.

Who was responsible for damages to the pedestrian? The owner of the vehicle because he owned the car and was its operator? The makers of the parallel parking AI, for not building the safety feature into it? The owner for not installing the safety AI?

Hopefully the lawyers' legal AIs can come up with a good solution to the issue. If this occurred in England, perhaps the new EBay-style adjudication being tested there could come up with a fast and cheap solution.

We are entering an interesting age.

(I hope the spell-check AI in my browser caught all my typos.)

read full comment
Image of Scott Amorian
Scott Amorian
on July 14, 2015 at 15:18:00 pm

(Ahem, gentlemen!)

read full comment
Image of Scott Amorian
Scott Amorian
on July 14, 2015 at 15:26:10 pm

AI systems have existed since the 1950's. It was originally expected that they would be more powerful, but they have not lived up to the anticipation around them.

Some aspects of AI died out because because they are unfeasible. Some areas have grown as computing power has increased.

Making a computer doing human things is kind of like making an elephant out of crystal. You can shape the crystal into an elephant shape. You can build crystal mechanisms that make the crystal move kind of like an elephant. You can cover the elephant with crystals. But in the end, a crystal will never be an elephant and it will never do all the things an elephant does.

AI systems are limited by their mechanics. They can repeat the same computations over and over accurately. They can access large amounts of data quickly. They can perform complex computations quickly. When we try to make them do more than that the AI systems tend to bog down.

That said, did you realize that cars are becoming AI driven robots?

There are a number of tests being done in the US today where cars and trucks are driving around on a few public roads with no human driver intervention. The vehicles have humans at the wheel, just in case. But cars and trucks are starting to drive themselves now.

read full comment
Image of Scott Amorian
Scott Amorian

Law & Liberty welcomes civil and lively discussion of its articles. Abusive comments will not be tolerated. We reserve the right to delete comments - or ban users - without notification or explanation.