On March 5-6, 2018, in Los Angeles, California, I was a co-facilitator at the first “Legal Ops” Institute of Today’s General Counsel. As a former California litigator (38 years) and the founder and CEO of an AI startup, Intraspexion, I was there for being cross-functional as to AI and the Law.
My takeaways consisted of two insights that I want to report here.
My first insight was that, just before I spoke, I realized that when asked about what AI does, attorneys are befuddled. Blank stares abound. Sure enough, when I did speak, I couldn't generate a conversation amongst the attendees.
I tried my insight, but I was casting water to a desert. So I might as well write about my insight and hope for the best.
From experience, I know that one thing AI does well with text is pattern-matching.
So here’s the translation: what attorneys learn to do in law school is to match new “fact patterns” to precedents.
Ah ha! I’ve already said that what a computer can learn to do is just that: learn patterns in one context and, in a different context, engage in “pattern-matching.”
What do we do? We know that precedents (appellate decisions) collected in the context of a litigation category are patterns. When a prospective client brings us a new factual situation to consider, we match the new “fact pattern” to a category of precedents that we’ve previously learned.
So what is it that's going on with the text-based form of AI called Deep Learning? Yes, it's pattern matching, and computer scientists would call this an "A to B" matching. There's one big difference between what we humans can do and what a computer can do: a computer can learn to match patterns not only with a huge (and ever-increasing) digital memory, a computer can also process all that data at very high (and ever-increasing) speed!
In fact, the inventive concept undergirding Intraspexion (and its patent portfolio) is that, first, we select a category of litigation (say Nature of Suit code 442 for Civil Rights: Employment, which is employment discrimination), extract the factual allegations from a large clutch of discrimination complaints, and feed those examples into a processing system (aka an “algorithm”) to train it.
Yes, a Deep Learning system can “learn” any category of litigation by digitally digesting examples of that category!
When the training (with factual allegations drawn from discrimination complaints) is complete and the algorithm is “tuned,” the system can recognize the pattern (or classification) we call discrimination.
So now a computer can carry the memory of all those previous discrimination lawsuits that we attorneys previously filed in the classification for discrimination, i.e., Civil Rights: Employment.
But where will we get the new fact patterns to consider?
Answer: a company’s internal electronic communications, e.g., emails.
So when the computer “pattern-matches” emails (in near real-time, say, from yesterday) to the discrimination category, it reports an “alert,” surfaces the risky emails, and scores the emails for how risky they are.
It’s as simple as that. The attorneys who work for the Corporate Law Department of the Future will deploy Intraspexion’s patented system and let it fish for the risks in yesterday’s emails.
So now if corporate attorneys read this post, or hear me speak to them in the future, they will know that the computer is doing is what we all learned to do in law school: match new fact patterns against the appellate precedents.
The second insight was an inspiring analogy. It came from Aleksandra Zimonjic, an attorney with the Los Angeles law firm of Landau Gottfried & Berger LLP.
After I explained what Intraspexion was about, she exclaimed, “Oh, I see. Better a toothbrush than a drill.”
I asked Aleksandra for permission to use her brilliant analogy and she said yes.
So, for visualization purposes, here’s different way to understand what Intraspexion means for the Corporate Law Department of the Future:
This, the toothbrush?