Later in May the Stanford Center for Legal Informatics will host the “CodeX FutureLaw Conference 2016,” a fourth annual conference focusing on how technology such as Artificial Intelligence (AI) has changed the landscape of the legal profession, the law itself and how these changes impact us all. The event will feature several panel discussions, including one relating to IBM’s Watson and related machine learning tools.
However, the law waits for no one and neither does AI, which has already made a lasting impact in many areas of business, including the practice of law. Contracts, e-discovery and overall legal research have all changed thanks to AI, but as computers driven by ever-increasing processing power exhibit extraordinarily intelligent behavior we can only assume such advances are far from over. Whether within the enterprise, partners, customers, opposing litigants or elsewhere, legal assets cannot hide from the likes of Watson—or for that matter HAL—or other budding or to-be-conceived AI platforms.
Not only has AI made these legal assets easier to find but also more accessible to attorneys, according to experts. However, the acronym GIGO applies as much today to AI systems as it did to PCs in the days of the TRS-80 and Commodore 64—garbage in, garbage out. Some things never change.
“AI is not going to work well if you only have a small amount—two or three examples—of the data you’re looking for,” says Kevin Gidney, founder and CTO, Seal Software, a contract discovery and analytics software provider. “The models would be really poor, so use an algorithm that is a fast learner and get the data using a combination of methods.”
Indeed, extensive empirical evaluation has found that a continuously adaptive machine learning approach can help lawyers keep an eye on ever-changing legal data, according to Jeremy Pickens, senior applied research scientist, Catalyst Repository Systems, which hosts and services document repositories for large-scale discovery and regulatory compliance.
“Implementing a continuous protocol involves more than occasionally retraining an existing machine learning classifier, but rather integrating it into the machine learning system at a native level,” Pickens says. “A diversity prediction algorithm in parallel keeps attorneys aware of nuances of what they don’t know. Together, continuous relevance and continuous diversity predictions allow attorneys to handle massive, changing document collections.”
Natural language processing
Despite paranoia and hyperbole surrounding AI since 2001: A Space Odyssey, intelligent computers will not take over the world, although that premise does make for exciting science fiction. While the rise of the machines is not something one should fear, AI systems and their architects have made significant strides in realizing learning machines that can adapt to dense, arcane legal terminology. One of these steps forward goes by the phrase natural language processing (NLP).
“AI has transformed the legal field—from predictive coding systems for large document productions during discovery to sophisticated NLP to provide analytics for litigation,” says Brian Howard, legal data scientist, Lex Machina, a LexisNexis company providing legal analytics to companies and law firms. “Traditional tools focused on information retrieval make users group and classify data before patterns emerge, limiting users to simple questions.”
According to Howard, NLP lets AI understand text features, enabling users to ask complex questions. For example, Silicon Valley law firm, Wilson Sonsini Goodrich & Rosati (WSGR) uses Lex Machina AI to figure innovative ways to use analytics to gather competitive intelligence. “Based on data analysis, WSGR can make better decisions, redeploy resources and align client billing to key events in a case,” Howard says.
E-discovery and predictive coding
Discussion of AI’s impact on the legal profession often frames the issue in terms of machines replacing lawyers. In reality, the biggest gains come when AI makes certain stages of legal research more efficient—not from machines replacing attorneys wholesale, according to AI scientists.
“This way, lawyers focus on solving and preventing legal problems at a higher level while leaving routine tasks to computers,” says Anna Ronkainen, chief scientist, TrademarkNow, provider of AI trademark solutions. “The work-product not only gets done faster and at lower cost but also quality is better, since people and machines make very different mistakes.”
According to Ronkainen, this transformation remains most evident in document review, where e-discovery with predictive coding has taken hold in the last decade, and reached a point where manual review sometimes is not even accepted by courts. Also, AI has made inroads into other legal practices. For example, AI trademark tools—including those from TrademarkNow—are used by Google, Roche and General Mills.
“Clients report time for a trademark search has been halved, and indication of whether to proceed takes less than a minute,” Ronkainen says. “By giving the marketing team access to these tools, trademark teams no longer have to deal with long lists of hopeless name candidates and can concentrate on finalists.”
Expert systems and neuromorphic techniques
Another AI technique that can help in legal cases is the use of expert systems, an algorithmic method. These systems can help look for particular types of evidence to support specific conclusions and provide probabilities given expert system results, according to Dr. Marco A.V. Bitetto, scientist in residence, Institute of Cybernetics Research, a nonprofit engaged in design and development of advanced robotic and control systems.
“We can also consider neuromorphic techniques that can derive patterns from video and audio stored on a lawsuit target’s computer,” Bitetto says. “Neuromorphic techniques make use of brain-like simulated circuits in software or hardware and can distinguish pattern variations that the system was trained to detect. Algorithmic and neuromorphic techniques can yield troves of evidence from a lawsuit targets’ computer network.”
Watson and ROSS level legal research playing field
While traditionally slow to adapt to technological changes, the legal profession has made an exception for AI. For example, taking advantage of AI, lawyers can keep up with changes in regulations and case law in near real time. One AI tool that really stands out is IBM Watson and its legal research platform ROSS Intelligence, according to Gurminder Kandola, co-founder of CTO Boost, a fractional CTO consultancy.
“ROSS is real-language, unstructured data mining,” Kandola says. “By asking questions in everyday English, this platform can provide a single, relevant answer in natural language, saving valuable time and providing answers at your fingertips. And an AI system is just as alert at 3 a.m. as 10 a.m. and more thorough when going through large amounts of data.”
In addition, AI systems can level the legal profession playing field. For example, using an AI platform a junior attorney could intelligently access 30 years of legal knowledge, whereas a lawyer with 30 years of experience might not turn to technology at all, according to Kandola.
Increasingly in the data-driven world in which we live professionals, including lawyers, need ways to shift through large amounts of data to get to the nuggets of gold. Artificial Intelligence solutions will only play a larger and larger role in the day-to-day practice lives for all attorneys and firms.