Artificial intelligence (AI) is the most exciting area for innovation right now, as it has been for the last decade, at least. AI has been around since computers moved from science fiction to laboratories, especially the laboratory of Alan Turing. But AI research and innovation really took off after IBM’s Watson beat Brad Rutter and Ken Jennings in Jeopardy in 2011 and then became the first commercially available “AI as a service.” After that, it didn’t take long for legal innovators to get on board.

What is AI?

A general definition might be when computers do things that show human-like intelligence. However, AI is a moving target, especially since humans have a tendency to constantly raise the bar. The “AI effect” is an observation that if a computer is able to do something, humans will probably decide that being able to do that thing doesn’t count as intelligence after all. It’s how humans protect our egos from AI encroachment.

If we take a step backwards, though, we can see that AI has accomplished quite a lot of tasks that people once considered a clear indication of artificial intelligence. Here are just a few examples:

  • Natural-language search.
  • Optical character recognition (OCR).
  • Voice recognition and virtual assistants (Siri, Cortana, Alexa, and Google Assistant).

In other words, humans used to think it would be a really big deal if computers could understand us. Well, now they can, and we just ho-hum about it. Hey Siri, where did I leave my phone?

Today just about everyone uses AI-powered technology, every day. For law, most of the innovation seems to be in the areas of expert systems and machine learning, with natural-language processing often playing a key role.

When you want to transfer your knowledge and decision making process into a system that can analyze issues and give advice pretty much the same way you do, that’s an expert system. An expert system is like a really sophisticated decision tree, built by someone (or more likely a team of someones) with expert knowledge of the issue.

When you have a lot of data and a question about that data, that is probably a job for machine learning. Machine learning is when a computer analyzes a dataset and looks for connections in order to draw conclusions. The dataset could be public records, user data, or judicial opinions.

AI technologies are complementary and can be used together, of course. You might have an expert system that uses natural language processing to understand user inputs, and machine learning to determine how to respond.

1940s–1990s
Origins of AI & the “AI Winters”

AI first moved out of fiction with the work of Alan Turing in the 1940s. He explained that a computer can simulate any process of formal reasoning. AI research began in earnest in the 1950s, heavily funded by the military. But in 1974 progress had slowed and the US and British governments cut off funding for exploratory AI research (sometimes called an “AI winter”).

While private AI research continued throughout the 80s and 90s, the popular resurgence of AI is probably when IBM’s Watson beat Brad Rutter and Ken Jennings in Jeopardy in 2011. Watson was a question-answering system with commercial potential, and in 2014, IBM launched the IBM Watson Group. Watson soon became a cloud-based “AI as a service” with users in medicine, teaching, fashion, weather forecasting—and law.

2015–Today
AI in Practice

Powered by Watson, ROSS Intelligence (now closed) launched in 2015, promising better results, faster, by using Watson to process questions and deliver the best cases in response. Rival legal research startup Casetext also incorporated AI technology to power its cite-checking algorithm and, later, its CARA and Compose assistants. CARA can read an uploaded document and identify the most relevant cases. Compose can actually produce draft briefs.

Although ROSS and Casetext, have grabbed most of the headlines for AI in legal research, Fastcase, Westlaw, and LexisNexis have steadily added AI to their legal research products, as well.

Document Review & Predictive Coding

Machine learning is a powerful tool for reading and categorizing large batches of documents, such as when companies need to do due diligence before a merger or acquisition, or when litigators must evaluate documents in discovery. Kira Systems is one of the first products built primarily for legal needs.

Published on January 14th, 2021. Last updated on March 23rd, 2021, by Sam Glover.