During the ILTACON 2016 Session “AI in Law: The State of Play in 2016”, Michael Mills, co-founder and chief strategy officer of Neota Logic, gave compelling insight into the economics, significance and potential of AI in law firms.

According to Mills, law firms have traditionally operated on a time-based billing model, simply scaling their work by “throwing people at the problems” clients hire them to solve. Current business model trends have begun to deviate from this philosophy to one where firms are now productizing their expertise.

Types of AI Applicable to Law

Effective uses of AI in law require a combination of techniques, some of which Mills described during the session. These included:

Supervised Learning: This system is based on the method of “teaching” a system how to recognize patterns by labeling data. For example, telling an algorithm that thousands of pictures are all of a cat, until eventually the algorithm determines that a certain formation of pixels is likely to represent a cat. Supervised learning has already been quite useful in law, particularly in e-discovery.

Unsupervised Learning: This is the converse of the method noted above – a system learns patterns only with unlabeled data. The algorithm will cluster data that are alike and try to come up with conclusions based on statistical analysis of the data. Similarly, this method has been applied to e-discovery as well as to produce concept clustering.

Reinforcement Learning: This type of machine learning combines the leveraging of existing knowledge with the exploration of uncharted territory without explicit guidelines for determining optimal patterns. It still requires input from another source to learn, but it can develop unsupervised predictions over time. This method represents a new frontier in law.

Each method requires the presence of a subject matter expert – like a lawyer – to develop, particularly in tuning the algorithms to be effective against the legal dataset and use cases.

Mills stated that most of the ways AI has been applied to law can be grouped into five categories:

Knowledge Automation: Using intelligent systems to break down complicated sets of rules. There have been many attempts to automate the US tax code, most of them simply coding statically against it rather than dynamically applying an algorithm. Newer companies are utilizing some of the machine learning techniques described here to try to assist in the creation of more interactive legal systems.

Legal Research: According to Mills, progress has been made in this area, but the industry has yet to see a game-changing application.

Prediction: A number of companies are trying to predict the outcome of cases and decisions. Dan Katz noted several examples of how the industry is trending here in his day-one session.

Contract Analytics: Organizations are using AI tools to make sense of corporate documents, similar to what has been done in e-discovery with technology-assisted review.

Mills, along with other speakers at ILTACON, gave an insightful and well-informed overview of the state of AI in law. But in the end, the industry is seeing sizable investments to bring innovative applications of AI to the practice of law.  And Mills suggests it may be less about whether the underlying technology meets the strict definition of artificial intelligence, as much as whether it is a superior way of solving a problem.

I’ll leave you to ponder a quote that Mills shared with us from Dutch computer scientist Edsger Dijkstra: “The question of whether a machine can think is no more interesting than the question of whether a submarine can swim.”