By Edward McIntyre
I slide into my desk chair and greet my assistant.
“Good morning, St. George.”
My assistant’s nickname — the dragon-slayer.
“Good morning — again.”
St. George is right. We’ve been back and forth all morning — as I rolled out of bed, on my way to the gym and to the office — about schedules, “to do” lists, today’s priorities, some long-term planning.
I settle in; my computer screens wake up.
“OK, St. George, what’s first?”
“Over night we received notice of the assigned panel for your Ninth Circuit argument in the Coast securities case. You have a report” — I glance at a screen — “analyzing every time these three judges sat together in a securities case; when any two did; when any one sat on a panel with a case similar to yours, even if not a securities case; when they were on a panel, in combination or with others, and your opponent argued —”
“In any kind of case or just securities?”
“All cases. May I continue?”
“Please.”
“The report gives you a statistical profile of the issues the judges prioritized; how they decided them; what arguments from the briefs and argument transcripts they favored; and, of course, outcomes, relevant parts of opinions and favored phrases used. I’ve also identified patterns of questions from the bench and who among them is more active; who is more deferential, both during argument and in writing opinions. Since you only have 10 minutes, I outlined what I conclude may be your three best arguments and included language that some or all of them have used repeatedly in affirming trial courts.”
“Perfect.”
“I also included argument patterns your opponent has used repeatedly and his manner of responding to questions from the bench. Not just in the Ninth Circuit, but all appellate arguments. Very interesting.”
“What are my chances, St. George?”
“I don’t do predictive analysis — yet.”
“Touché.”
“The data analytics, as a result of an analysis of these three judges, for the issues in the Coast case, given the opinion of the trial court, including its weaknesses, and your opponent’s prevailing tendencies all suggest an 82 percent likelihood of an affirmance.”
“Very nice. Next.”
“You asked on the way to the gym this morning about a summary judgment motion in the Preston case in the Arizona district court.”
“Yes, thoughts?”
“I’ve analyzed all summary judgment motions Judge McCool has decided since she took the bench, including all the briefs and oral arguments where she allowed arguments and the outcomes. I then focused on trademark cases like Preston. I’ve digested all the briefs, transcripts of arguments, comments from the bench and opinions, and started on a summary judgment motion using the arguments that have prevailed with her or that prevailed with judges she tends to rely on. I also analyzed the discovery in Preston and did a working draft of a separate statement of uncontested facts.”
“This means we can do a full summary judgment motion in fewer than, say, 10 hours of my time, 20 tops. That should encourage the client to move forward with the motion.”
“Very cost-effective.”
“How often has Judge McCool granted summary judgment in a trademark case?”
“She has only had three. One has a lot of facts in dispute, so denial was easy. On the other two, she split 50/50. Looking at her record overall, on close calls she tends to be conservative. She often quotes from Judge Betts’ decisions; seems to defer to him. I’ve included language from cases where he granted summary judgment in trademark cases.”
“A 50/50 chance?”
“I don’t do predictive analytics.”
“The draft brief?”
“On your screen.”
“Thanks, got it. I looked at that professional responsibility research report —”
“That was new for me.”
“I know. A good start. Now I’d like you to refine the data search. Look at all State Bar cases where discipline was considered — whether imposed or not — under 6106 or 6068(d) for misleading a tribunal.”
“Even if it doesn’t fit your facts?”
“Yes. We may have to argue by analogy. I need to see the data first.”
“That doesn’t seem reliable or predictive.”
“Perhaps not as a matter of data analytics. But this may be a good time to explore the rules of analogy.”
“That could be interesting.”
“Is there more?”
“A lot. Shall I continue?”
Who is this St. George?
Good question.
Does St. George have an army of lawyer researchers and data analysts at immediate command? Or at least a couple of platoons to produce all this work based on an overnight court notice and my question just two hours before?
No, it’s only St. George.
Where did St. George do all this work? That’s easy. Somewhere in the cloud.
St. George is my multi-faceted, artificial intelligence platform. The servers that house and munch the huge data sets St. George manipulates so fast and skillfully are likely in Iceland or Svalbard Arctic — somewhere remote and cold, with abundant cheap energy.
St. George, however, is not some futuristic daydream. AI platforms are already radically affecting certain aspects of legal practice — as in medicine, manufacturing, retail and, soon, driving. And they are only the beginning. Just as email started slowly, then cascaded and has now altered radically in less than a generation the way we communicate — a transformation other social media have only reinforced — AI will soon become ubiquitous, an indispensable asset in the practice of law.
What is Artificial Intelligence?
The term first gained recognition at the Dartmouth Conference in 1996. The early “test,” the Turing Test, posed that if a machine could carry on a conversation — using e.g., a teleprinter so the machine’s inability to speak was not a disadvantage — indistinguishable from a conversation with a human being, it would be reasonable to say the machine was “thinking.” AI is most capable in the areas of data mining, industrial robotics, logistics, speech recognition, banking software, medical diagnosis and Google’s search engine.
Another way to think about AI is to consider it cognitive computing: teaching computers how to learn, reason, communicate and eventually make decisions. At the heart of such machine learning are neural networks, patterned on the neuron networks of the human brain — although not nearly so fast, sophisticated or capable. These neural networks are, of course, empowered by the quantum leaps in computing power and speed the last decade has brought. Without that power and speed — think IBM’s Watson® — none of what we see would be possible.
Neural networks are “trained” in pattern recognition, using large numbers of training examples, correcting the network as training moves forward. As training progresses, the neural network automatically begins to infer the rules of recognition, becoming more and more accurate; the larger the number of training examples, the more and faster the network learns, and the greater its accuracy. Thus, cognitive computing tools are trained, and eventually self-train, as opposed to computers that are programed to a specific task. The focus of a cognitive computing tool is looking for patterns in the data, testing the data, and finding and providing results.
This contrasts with Boolean logic and keyword searches. Each search is linear with no relation to past or future searches. With AI, on the other hand, each search becomes part of the learning process; every search and answer — and correction if necessary — makes the machine much better for the next search.
AI’s Current Impact on Law Practice
AI is at our fingertips. One existing AI platform, designed specifically for lawyers, relies on advanced technologies — machine learning and natural language processing — to structure and analyze raw data from millions of dockets and documents. That platform has in its database all federal circuit and district court cases, and Delaware Court of Chancery cases, for 10 subject matter areas of law including: commercial, employment, bankruptcy, patent, product liability, securities, antitrust, trademark, copyright and corporate (Court of Chancery). The database contains not just opinions but also identifies the judges and lawyers, and has the briefs, transcripts of arguments, dockets, orders and every relevant bit of data in each case.
As lawyers, we are trained to research and reason. What this AI platform adds to research and reasoning is data analytics; it examines and aggregates a vast array of data and provides the lawyer with data-driven insight into the behavior of the individuals — the judges and lawyers — in specific kinds of cases, faced with specific kinds of arguments. Thus, data analytics allow lawyers to understand trends and patterns in past litigation that they can employ to inform legal strategy and anticipate outcomes in current cases. The underlying premise is that each of us — lawyer, judge, client — has latent behavioral tendencies when choosing arguments, writing briefs, engaging in oral argument, asking questions from the bench or conducting hearings, deciding cases and electing strategies.
Rather than sending an email around the office asking colleagues what they know about this lawyer or that judge or how this party reacts in litigation — imperfect anecdotal information at best — this AI platform analyzes a vast quantity of data, at warp speed, and delivers a factual, data-supported, answer: In x percent of trademark cases, Judge Y has found the challenged mark confusingly similar and these four arguments prevailed with her z percent of the time. Or, in a securities case, lawyer x has brought a Rule 12(b)(6) motion — q percent unsuccessful — in y percent of the cases he defended.
Clients, especially sophisticated in-house counsel, also have a tool by which to evaluate and choose counsel. No need for dog and pony. Rather, how successful has lawyer x been in defending employment discrimination allegations in a particular district, or before a particular judge, against a particular opponent. No more slick brochures and lawyer performances; hard data, analyzed from a vast array, is difficult to ignore.1
What Will This Mean To The Profession?
In overwhelming respects, AI will enhance our ability to serve clients. Think, for example, of the asymmetry of a plaintiff’s employment or securities case. No longer will a small plaintiff’s firm necessarily be outgunned by “Big Law” — with layers of associates and paralegals — on the defense. The small firm, or even the sole practitioner, now has available overwhelming data analytics tools; and many more will follow.
AI is a tool; AI, at least as we know it, will never replace human judgment. Rather, it frees the lawyer from much of what we currently do — researching and analyzing information — by delivering research product and analyzed data to us so quickly that we have time to think, exercise judgment and engage creatively. Critically, it will allow us to do so much for clients more economically — as the sidebar article discusses. After all, technology does not get tired, show up late, call in sick or take vacation, and doesn’t need food or other breaks.
What, however, about those cadres of lawyers, paralegals and other personnel at firms currently doing legal research, analyzing documents, reading transcripts and engaging in all the other people-intensive tasks? AI will replace most of those tasks. On the other hand, AI will create the need for those who can develop and manage AI computers — legal engineers — and those who will be writing the necessary algorithms. What about all of the billable hours that legal research, document analysis and transcript reading produced? They will disappear; firms will have to rethink their operating business models. Thus, AI will both transform the practice of law and eliminate parts of it.
For all lawyers, AI will mean adapt and embrace or be left behind — as other technological innovation has also affected the profession. To return to the email analogy, we will not be able to “delegate” AI to someone else within the firm. But no need to panic.
AI platforms are not just for the “tech savvy”; the interfaces are user-friendly and often intuitive. Spoken language recognition will soon become routine — like my St. George. We need not fear the further advent of AI. Yes, it will transform much of how we practice today, but in ways that will allow us to better serve our clients.
Edward McIntyre is an attorney at law and co-editor of San Diego Lawyer.
1 The AI platform identified is Lex Machina, headquartered in Menlo Park. Its product is currently used by a number of law firms and corporations.
This article originally appeared in the March/April 2018 issue of San Diego Lawyer.