(Bloomberg View) -- Lawyers, beware. Robots really are coming for your jobs.
Exhibit A: Venture-capital firm Invoke Capital just made a multi-million-dollar investment in Luminance, which is developing artificial intelligence to automate the legal drudgery involved in corporate mergers and acquisitions. The robot lawyer is just one of many -- including offerings from Ross Intelligence and Kira Systems -- aiming to replace the overworked factotums known as associate attorneys. Without the six-figure student debt, I presume.
So how do these virtual attorneys work? Well, according to Bloomberg, Invoke founder Mike Lynch said that Luminance can "read natural language and actually understand it, using it to categorize documents, rather than just searching text to match key words or standard clauses.”
This leads me to wonder whether Lynch has ever used Google, because what he described is the exact function of a search engine. Due diligence, the process Luminance is supposed to automate, essentially involves organizing massive piles of legal documents into smaller piles based on their relevance -- the specialty domain of search engines. Even the IBM-Watson technology that powers Ross is, for the most part, a search engine.
The fact that their software isn’t quite so unique, though, doesn’t detract from robot lawyers' potential. What really matters for their “intelligence” is the data on which they are trained. Ross focuses the Watson technology, which won "Jeopardy!" in 2011, on court filings instead of the full text of Wikipedia. To prepare for its specialty of contract review, Luminance has studied "thousands of documents and contract clauses.”
Well-tuned search engines could save people a lot of time and suffering. Luminance promises to increase the efficiency of contract review by at least 50 percent. Kira Systems claims a time reduction of as much as 90 percent. If Bayer’s legal team had included robot lawyers, maybe they could have completed due diligence for the Monsanto deal in days.
So will the associate attorney, among the least satisfying jobs in the U.S., become a thing of the past? Not necessarily. Even though automated-review tools are great for organizing documents into actionable information, intelligent humans are required to step in when the computer encounters ambiguous language or unexpected cases. It’s like how self-driving cars still have human supervisors in the vehicle to deal with rogue squirrels or trolley problems.
That said, the need for humans will diminish. As machines process more data, previously ambiguous information will become recognizable. Consider tax returns. Thanks to the Internal Revenue Service's automated computer review system, trained on hundreds of millions of returns annually, humans need to show up only in the event of an audit, which happens in less than 1 percent of cases. It’s just a matter of time before Luminance develops similarly powerful pattern recognition.
The main difference between a junior associate and a partner at Cravath is the sheer quantity of data that the person has seen. A computer can process data much faster, with far more accurate recall ability. Once a robot learns to structure merger agreements and negotiate corporate acquisition deals, human-readable legal documents will be little more than an afterthought. Eventually, human lawyers will be called in to interpret only the rare ambiguity, likely a human-generated flaw that never would have existed if lawyers had used robots in the first place.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.