alt.legal: Can Computers Beat Humans At Law?

If artificial intelligence (AI) can beat the greatest strategy champions at their own games, will it reach a point where it can beat a lawyer at the practice of law?

Ed Sohn

Ed Sohn

A good friend recently told me that it takes a special kind of nerd to appreciate what Google’s AlphaGo did to international Go champion Lee Sedol: a nerd that is both a Go nerd and a computer nerd.

I guess that’s me.

For Go nerdiness, I am recently enamored with the massively complex game that has exponentially more outcomes and dimensions than chess. (If you want to learn more about this ancient game, start here, then get here and eventually, if you’re my special type of nerd, watch this entire series.) As for the tech nerdiness, many of us assumed that after DeepBlue beat Kasparov in chess, any other game was a foregone conclusion. But actually, it’s taken twenty years for a computer to rise to the level of top-ranked Go players, because high-level Go incorporates less calculation of a limited set of future outcomes and far more intuition.

Challenges like this are not just an interesting competition. They are yardsticks measuring how far the machines have come, and the impact on our lives is never far away. After IBM Watson beat former Jeopardy! champions Brad Rutter and Ken Jennings, IBM productized the deep learning and natural language interaction to form a level of artificial intelligence (AI) now known as “cognitive computing.”

We’ve seen AI beat the greatest strategy champions at their own games, and we’ve seen AI thrash trivia champions in their own language. Now Watson-powered applications are doing everything from assisting in healthcare to helping financial advisors plan our futures.

Will AI reach a point where it can beat a lawyer at the practice of law? Just as Lee Sedol battled AlphaGo, will legal robots of the future battle attorneys, seeking to supplant them? I offer an investigation into this for my second post in the series on AI and the law (part 1 is here).

Sponsored

A Brief History of Artificial Intelligence and the Law

In gathering sources for this article, I realized I hadn’t Googled “artificial intelligence” and “law” for quite a while. The very first hit:

AI and the Law

There is an international association devoted exclusively to promoting the study of artificial intelligence and the law! And, after thirty seconds of browsing, I came to discover that Jack Conrad, former president of the IAAIL, was also a colleague of mine at Thomson Reuters. Bully for Thomson Reuters.

Conrad is a true scientist. It’s his actual job—lead research scientist. He holds several advanced degrees in engineering, linguistics, and computer science. He is widely published in areas related to linguistics, information retrieval and artificial intelligence. His role with Thomson Reuters included two years in Europe researching with our Swiss labs. His name is associated with at least four patents.

Sponsored

So this guy is no newbie. I connected with him on one of my trips to the frozen tundra called Minnesota, where Thomson Reuters has a massive campus. Sitting one chilly afternoon in the on-campus Caribou Coffee, Conrad was soft-spoken and friendly. While ever the scientist, he offered his opinions with the tone of someone used to taking complex concepts and making them into plain English.

“There were a lot of interesting things happening in areas of search and conceptual search, even trying to model the law using technology, back in the late 70s and early 80s.  The first international conference for artificial intelligence and the law (ICAIL) took place in Boston, in 1987.” A twinkle of pride snuck into his smile. “This community recently celebrated 25 years of AI and the law.”

According to Conrad, the IAAIL has three main pillars: academic researchers, government people seeking new solutions, and corporate research labs, like at Thomson Reuters. The association’s primary mandate is holding the biennial ICAIL conference, spurring on academic research and providing a forum for encouraging and presenting findings in papers.

But the most interesting part for Conrad has been when some of their research found real-world application. The early creators of natural language search at Westlaw came out of this forum, and the late Peter Jackson, who is highly revered at TR for his work on WestlawNext, was a prominent contributor in the IAAIL—there’s now an award for best innovative application named in his honor. Recently, one of the early chief researchers at LexMachina published papers through their conference as well.

“These weren’t just abstract ideas,” Conrad emphasized. “People were putting their ideas into practical applications.”

One practical application of AI in the law is the good work Thomson Reuters recently announced in partnership with IBM Watson to create a center of excellence for cognitive computing. To explain cognitive computing, Conrad gave a quick overview of how in the 80s, people differentiated between “hard” and “soft” AI. The popular theory at the time was that hard AI was coming soon to completely replace human beings. But soft AI is more of an extension of human cognition, an assistant. This, Conrad explained, is the way we have to think about the AI in legal practice: a cognitive assistant to help us learn, search, retrieve, and analyze information.

robot lawyer computer attorneyWhich brought me to the big, big question. “Given the advances you’ve seen and the advances you anticipate, are the robots, in fact, coming for attorney jobs?” I asked.

Conrad paused, then he answered with the precision of a scientist. “Cognitive computing does not replace the capabilities of the human brain. There’s a lot of hype and hyperbole about this.

“But it may make an attorney more efficient. And to the extent attorneys perceive a threat to their practice, it may be because there’s been an inefficiency there.”

Hear, hear.

ROSS Intelligence

I had some answers from Jack Conrad, but the alt.legal in me went searching for commercial application in the startup world. I found a new legal tech startup, ROSS Intelligence, and I was able to speak with Andrew Arruda, the CEO and co-founder of ROSS.

You might be wondering what ROSS stands for. “It doesn’t stand for anything,” explained Arruda enthusiastically. “We just gave it a human name, and it’s taken really well. People say that they are going to go ask ROSS something and it’s really natural.”

Fine. ROSS is not quite like HAL 9000, in that sense.

So what is ROSS, then? “Ross Intelligence is an AI legal researcher that allows lawyers to do legal research more efficiently, in a fraction of the time. It does that by harnessing the power of natural language processing and machine learning to understand what lawyers are looking for when conducting their research, then get smarter each time to bring back better results. It grows alongside our lawyers.”

It all started when the CTO and co-founder, Jimoh Ovbiagele, was a young boy. “His parents had unfortunately gone through a divorce, and this kid saw all these legal bills and saw how much added stress was in their lives.” As Ovbiagele grew up, he was a “total prodigy, coding at the age of 10.  But law was always in the back of his head.”

Arruda and Obviagele met at the University of Toronto, then Arruda went on to law school. “I got a call from [Ovbiagele] after I finished law school and he told me about this idea of using natural language processing and machine learning to enable attorneys to do their legal research.” Just like that, Arruda went into the company full-time and together with a third co-founder, a visiting computer science scholar from Brazil named Pargles Dall’Oglio, they started ROSS Intelligence.

ROSS was immediately successful within the canon of Canadian employment law, and the team generated a lot of fast interest internationally. “The U.S. was a prime place to grow our business, given the cost pressure firms were facing from the Great Recession. As we started packing our bags to go to the States, one of our advisors asked us if we had gone to Y Combinator yet.”

That’s like asking a budding violinist if she’d tried taking a few lessons at Juilliard yet. Indeed, Arruda had not yet been to Y Combinator, the famous American startup fund referred to as a spawning ground for emerging tech giants like Dropbox, Airbnb and Stripe. Of course, ROSS did apply, and their application was accepted. Just like that, they got Y Combinator funding. They also received funding from a prominent legal tech investment group, NextLaw Labs, a global innovation platform primarily backed by Dentons.

So fresh investment in hand, in 2015 ROSS entered the U.S. and, with a beta group of 20 firms, learned everything about American bankruptcy law. “In its current state, ROSS has learned bankruptcy law. But it’s actually also learned a lot about how to answer legal research questions generally.”

It sounded promising but also confusing. How did this thing work? At its core, Arruda explained, their stack included IBM Watson (of Jeopardy! fame) and other open source and custom-built tech. However, Arruda assured me, “the ROSS experience is simple. We’ve designed it carefully to get attorneys up and running really easily. You query the law in the way that you would ask any research question. As you ask it more queries, the machine learning component kicks in and ROSS becomes smarter over time.”

And ROSS isn’t just the tool that goes and finds—it’s the tool that reminds. “Upfront investments into legal research take forever, and then as time goes on, subsequent research tasks continue to pop up in the same area. ROSS can actually monitor your questions and provide alerts when there’s updates in the law. It handles and automates that information retrieval task.” Just like that.

At this point in our conversation, you could really hear in Arruda’s voice the familiar passion I’ve heard from other legal entrepreneurs. It’s the conviction that the world will be a better place, and he was personally going to help it happen.

“We see ourselves as the company building AI solutions in law. When we hear that lawyers don’t like change, that’s not quite right. Lawyers can accept change, lawyers just don’t like risk. When you can show that the technology really works, they become huge advocates. We are part of a group of companies that can come in and bridge that gap and enable lawyers to provide more services to more people.”

Arruda was on a roll. I already knew what his answer might be, but I had to ask him the same question I asked Jack Conrad—are the robots coming for our jobs? As expected, Arruda was impervious to pessimism.

“If you look at the state of our profession, we should be very happy about what’s happening with AI,” Arruda enthused. “Right now, 80% of people who need lawyers can’t get one. We have record numbers of law graduates who are unemployed and in record debt. What we at ROSS have prophesied is that systems like ROSS will be able to bridge this divide between humans who need services and the lawyers that can provide it. ROSS is one tool that lawyers can put in their toolkit and reach a really strong market that is in need for greater access to justice.”

Again, as with Conrad, the emphasis was on assistance. To illustrate the point, Arruda started explaining this phenomenon called centaur chess. On further research, “advanced chess” (also known as cyborg or centaur chess) is something invented by Garry Kasparov after being defeated by IBM’s DeepBlue to advance the game to new heights and complexities by pairing humans together with computers, thus amplifying their combined performance.

Arruda’s energy was, at this point, radiating through the speakerphone.  “When you pair the computer with the human, you get something way better than either the human or the computer. If you look at it from that formula, humans will always be on the winning side.”

Centaur lawyers?

That’s the best answer I’ve found, consonant with Conrad’s view on lawyer efficiency. If attorneys aren’t pitting themselves to compete with tasks that AI can do, there’s nothing to worry about. Why compete with the computers when you can collaborate with them?

When you look at AI and the law from that perspective, there is no fearsome force coming for us. There is only a formidable force that’s coming with us, to work alongside us.

Together, we’ll create a centaur legal world.

Earlier: alt.legal: I Still Haven’t Found What I’m Looking For!


Ed Sohn is a Senior Director at Thomson Reuters Legal Managed Services (formerly Pangea3). After more than five years as a Biglaw litigation associate and more than two years in New Delhi overseeing the delivery of managed document review, Ed now focuses on managing the new e-discovery solutions with technology managed services. You can contact Ed about e-discovery, legal managed services, theology, chess, Star Trek: The Next Generation, or the Chicago Bulls at edward.sohn@thomsonreuters.com.

CRM Banner