Checkers, Chess, Jeopardy, Go … Law

Welcome to our computer overlords

Korean man and woman, ca. 1910–1920, playing Go.

A shudder of excitement went through the tech world recently and its epicenter was Seoul, South Korea. There, a computer named AlphaGo played five games of Go against Lee Sedol, a South Korean master of the game ranked fourth in the world. AlphaGo won four out of the five games. Long considered a difficult and perhaps impossible task, a computer winning at Go suggests that computers are moving closer to taking over some human tasks much sooner than we imagined. It also was a strong volley by Google to be the company whose algorithms drive the “thinking” behind the takeover.

Yet, of the approximately 1.25 million lawyers in the United States, it is safe to say that few read about and understood the significance of the victory, some saw the headlines but skipped the stories, and many did not even know the match took place. AlphaGo’s win over Sedol will be one of those moments lawyers will look back on and see as another tipping point they missed. The event that shook the technology world caused barely a tremor in the legal world.

What is Go

To understand the significance of AlphaGo’s win, you must understand something about Go. The game originated in China and dates back at least 2500 years. It was considered one of the four “essential arts” of a cultured gentlemen (the other three were calligraphy, painting, and qin, a stringed instrument). Two players do battle on a board which has a grid of 19 x 19 lines. Each player strategically places his pieces, following several rules, to block territory (space on the board). The winner protects the most territory.

There are several measures of game complexity, including game tree size, decision complexity, game tree complexity, computational complexity, and state-space complexity. Journalists often use state-space complexity to describe the relative complexity of games, because it is fairly easy to grasp: “the number of legal game positions readable from the initial position of the game.” The state-space complexity for Go has been estimated at 10¹⁷⁴, which is more than the total number of atoms in the universe. By contrast, the state-space complexity for chess is 10¹²⁰. The difference is not trivial. A computer can play chess using brute force. It can calculate the possible move combinations after each play and select the next move out of the universe of possibilities, taking into consideration various strategies. Because the number of possibilities in Go is so high, a computer cannot use brute force. Instead, it must do something to approximate human intuition. It has been described as the “pinnacle of perfect information games.”

Round 1: Checkers

Computers have been beating humans at games for more than 20 years. In perfect information games, players move alternately and each knows all of the other player’s prior moves. In 1994, a computer program named “Chinook” developed by Jonathan Schaeffer at the University of Alberta, was declared the winner in a match against the world’s top checkers player in the Man-Machine World Championship. While the victory was impressive, there was a hanging question about the computer’s abilities. It was declared the victor after drawing six times. Marion Tinsley, its human opponent, then withdrew from the match due to problems with his pancreatic cancer. Chinook never actually won a game against Tinsley.

In 1995, Chinook played Don Lafferty in a 20-game match. It won one game, lost one game, and drew 18 times. Schaefer retired Chinook from competition after that match, but he and his team continued working on the checkers problem (a program that a human could not beat). In 2007, they announced that the best any human player could achieve in a game against the updated Chinook was a draw.

Round 2: Chess

The next human loss to computers in a perfect information game happened just a few years after the Chinook defeat. Chess had long been viewed as a game that challenged the smartest humans. A computer beating a human would make quite a statement about the state of computer “intelligence.”

In 1996, IBM’s Deep Blue played Garry Kasparov in a six game match. Kasparov won 4–2. But in 1997 they played a rematch which Deep Blue won 3 1/2–2 1/2. That was the first time a computer had beaten a Grand Champion chess player in a match following tournament regulations. Deep Blue’s victory was significant, though the victory represented brute force more than elegant play. At the time, some believed that Kasparov did not bring his best to all the games in the second match and could have won had he done so.

Some believed if Kasparov had played with more human intuition he would have beaten Deep Blue.

After Deep Blue defeated Kasparov, IBM wanted another challenge to show off its software. Jeopardy presented that challenge. Jeopardy is more complex for a computer than chess. First, there is the format. The host gives the answer and the contestant must respond with the correct question. Second, Jeopardy involves language interpretation. As The New York Times described it, Jeopardy is “a game that requires not only encyclopedic recall, but also the ability to untangle convoluted and often opaque statements, a modicum of luck, and quick, strategic button pressing.”

In February 2011, IBM’s latest masterpiece, Watson, played Jeopardy against Ken Jennings and Brad Rutter, the two leading human contestants. After three matches, the results were a clear win for Watson: $77,147 to Jennings’ $24,000 and Rutter’s $21,600.

As with Deep Blue’s win against Kasparov, Watson’s Jeopardy win against the human contestants was impressive, but it also showed that Watson was not perfect. Computers still had a long way to go when it came to matching wits with humans.

The Final Round: Go

With the chess and Jeopardy matches under its belt, the computer world wanted another win. Go was seen as the ultimate perfect information game challenge. A win against a human would show that computers had moved beyond brute force and were taking on human “intuition.” The computer could not simply crunch numbers, it would have to do something else to beat a Go grandmaster. Two competitors took on the challenge, Google and Facebook, and Google got there first with AlphaGo.

In October 2015, AlphaGo played a match against Fan Hui, the European Go champion. AlphaGo won 5–0. While a significant victory for AlphaGo, its next match against Lee Sedol was an even bigger challenge. Sedol watched the games between AlphaGo and Hui and was able to evaluate AlphaGo’s strategies and weaknesses. Sedol predicted that, while the computer was good, he was still better.

As it turned out, Sedol was (mostly) wrong. In fact, most experts were wrong. In 2015 before the match between AlphaGo and Hui, most experts were predicting it would be another decade before a computer could beat a Go grandmaster. But in the few months between the match against Hui and the match against Sedol, AlphaGo continued improving. Unlike a person, AlphaGo could play games continuously and at a furious pace, learning all the time. What it learned gave it the edge.

How did AlphaGo learn to play Go so well? According to researchers at Google’s DeepMind:

AlphaGo was programmed to sift through a database of expert Go moves, and then play against itself millions of times to improve its performance. Researchers called that part of the program the “policy network.” Another part of the program runs through Monte Carlo simulations to evaluate board positions.

Today, Demis Hassabis, who founded DeepMind and still leads it after the Google acquisition, says his team believes AlphaGo could learn to play entirely through self-learning. As Hassabis says:

Actually, the AlphaGo algorithm, this is something we’re going to try in the next few months — we think we could get rid of the supervised learning starting point and just do it completely from self-play, literally starting from nothing. It’d take longer, because the trial and error when you’re playing randomly would take longer to train, maybe a few months. But we think it’s possible to ground it all the way to pure learning.

Because Go is so complex, during training AlphaGo had to learn how to use some measure of computer “intuition.” What do we mean when we say AlphaGo has “intuition”? Geoffrey Hinton, called the “godfather of neural networks” and a member of the AlphaGo team describes it this way:

The really skilled players just sort of see where a good place to put a stone would be. They do a lot of reasoning as well, which they call reading, but they also have very good intuition about where a good place to go would be, and that’s the kind of thing that people just thought compute[r]s couldn’t do. But with these neural networks, computers can do that too. They can think about all the possible moves and think that one particular move seems a bit better than the others, just intuitively. That’s what the feed point neural network is doing: it’s giving the system intuitions about what might be a good move. It then goes off and tries all sorts of alternatives. The neural networks provides you with good intuitions, and that’s what the other programs were lacking, and that’s what people didn’t really understand computers could do.

Does AlphaGo’s victory and use of “intuition” at some level mean computers are getting close to human abilities? According to Hinton, not for more than five years (he refuses to predict anything that he thinks is farther out than five years):

My belief is that we’re not going to get human-level abilities until we have systems that have the same number of parameters in them as the brain. So in the brain, you have connections between the neurons called synapses, and they can change. All your knowledge is stored in those synapses. You have about 1,000-trillion synapses — 10 to the 15, it’s a very big number. So that’s quite unlike the neural networks we have right now. They’re far, far smaller, the biggest ones we have right now have about a billion synapses. That’s about a million times smaller than the brain.

There will not be another perfect information game challenge that surpasses Go. While there are other strategy games, they involve human language and interactions and other dimensions which still are well beyond computer capabilities.

Oren Etzioni, CEO of the Allen Institute for Artificial Intelligence, described AlphaGo’s wins as representing “an outstanding technical achievement, … demonstrat[ing] that when the goal is crystal clear, and the rules of the game are simple … computers will dominate.” Etzioni contrasted that situation with the ones that lawyers confront:

“[W]hen the problem is ‘ill-defined,’ as in understanding a sentence, writing an article, or even comforting a friend — this is still way beyond our [AI’s] abilities.”

Lawyers should not assume Etzioni’s comments mean machine learning computers are not ready for legal services. It turns out there are many ways computers can augment what lawyers do.

When Thinking About Automation, Think Tasks Not Jobs

While no one is suggesting that board games are one step away from practicing law, AlphaGo’s significant step forward from Deep Blue and Watson suggests why lawyers must embrace a different future. I call this future the “augmented lawyer.” Lawyers leverage the growing power of computers by using them to handle “power and volume” tasks while lawyers contribute skills computers have not mastered.

Artificial intelligence is not an “either or” situation, often phrased as either the computer can do the lawyer’s job or it can’t. Lawyers’ jobs are composed of thousands of tasks. Some tasks are extremely complex, but many are very simple. Most are mixed. Legal service processes can be disaggregated, for example, during the lean thinking exercise of process mapping. Once a process is disaggregated, the question is whether computers can automate a task or even part of a task. AlphaGo’s win shows there are many tasks lawyer do that computers can learn.

We already have seen computers tackle legal tasks that require significant power or involve lots of data. Consider Shepardizing a case. Shepardizing is a term that dates back to before the modern computer era and comes from a series of books published as Shepard’s Citations (named after the original author, Frank Shepard). Shepard’s books listed all of the published case decisions by federal and state courts. They were arranged by citation. Underneath each case citation heading, Shepard’s listed subsequent cases, that is, cases referencing the case cited in the heading. Shepard’s used symbols to indicate why each subsequent case cited the heading case (overturned, reaffirmed, questioned, cited, etc.).

Lawyers want to know whether the cases they present to judges in briefs are still good law. To Shepardize using books, the lawyer would look up the main citation in the most recent bound Shepard’s volume. Then, she would move to unbound volumes covering each month since the last bound volume. Then, she would move to small booklets covering the weeks since the last monthly volume. She had to repeat this process for each case cited in her brief.

Once she knew all the cases citing a case in her brief, she had to decide which ones to review. If Shepard’s listed five cases citing her case, she might not need to read all five. But if Shepard’s indicated any of the subsequent cases had questioned, limited, or overturned her main case she had to read the subsequent cases. That meant going to the library, tracking down the book with the published case and either reading it right there or photocopying it for later reading. If there was a significant lag between finishing the Shepardizing process and filing the brief, she had to check each case against the latest small booklets. Junior associates had the pleasure of Shepardizing and it took quite a lot of time and client money.

The Shepardizing process is quite different today. Software automatically checks each citation in a brief against databases of published decisions. Within seconds, the lawyer has a list of all cases citing any of the cases in her brief. By clicking on a link, she can go directly to a citing case and the place in the case where the citation occurs. The tasks of finding the citing cases and tracking down the published decisions have been automated, significantly reducing the time spent on those tasks. If there is a lag between Shepardizing and filing, updating means clicking the button again and checking any new cases. Lawyers still must read the citing decisions and draw conclusions about whether those cases have any impact on the case they cited.

Large Parts of Legal Service Processes Already Can Be Automated

Most legal service processes involve this combination of tasks amenable to automation and tasks that lawyers still must do. The question is not whether the computer can replace the lawyer, but which tasks the computer can do more quickly, less costly, and with higher quality than the lawyer. The number of tasks where the computer excels is growing, but lawyers resist change and are holding on to those tasks. That resistance increases costs to clients while also reducing quality.

The false “either or” dichotomy masks a continuum. By using process mapping, continuous improvement, and workflow automation reviews, lawyers can construct a pipeline where processes have waste removed and tasks move to automation when the timing and cost make sense. Instead of fancy, complicated software systems lawyers can use simple, low cost and easy to modify automation tools.

A recent article by Michael Chui, James Manyika and Mehdi Miremadi in the Harvard Business Review made estimates about the percentages of daily tasks knowledge workers perform that could be automated.

30% of the tasks performed by knowledge workers can be replaced with current automation.

Add to that another 20% reduction in tasks from taking waste out of processes (waste reduction estimates range from 20% to 95%, so using 20% is conservative). That means lawyers should be able to get a 50% reduction in the tasks done by attorneys be eliminating or automating them without significant spending or overwhelming time investment.

AlphaGo’s Victory May Be A Tipping Point

AlphaGo’s victory does not mean a defeat for lawyers. But in three ways it is a warning.

First, given that only a year ago experts were predicting it would take a computer another decade to beat a human at Go, it shows the pace of machine learning development is faster than anticipated. The more lawyers resist automation, the farther behind they fall. The faster computers move up the machine learning curve, the greater the gap between what computers can do and how lawyers use them. This will accentuate the difficulty of becoming an augmented lawyer.

Second, lawyers base many of their arguments against using computers in legal services delivery on the distinction between rote tasks (find all documents with the word “discriminate”) and legal thinking. AlphaGo’s use of something akin to intuition shows that, at a minimum, computers are capable of more in law than their critics had claimed. In truth, this already was the case but the AlphaGo win makes it more apparent.

Third, lawyers argue that the cost of training a computer system outweighs the benefits received from the trained system when it comes to the legal tasks that system could do. Hassabis’ statement that AlphaGo will be able to train itself how to play Go and defeat a human player in less than a year shows that the upfront burden of bringing computers into legal services will soon be, or already is, dropping.

Today, anyone with a Macbook, a basic knowledge of Python programming, and access to a data set can turn loose machine learning on a problem.

Many machine learning programs are readily available. Google is even making some of its work available (see TensorFlow, an open source Machine Learning system, now expanded by Cloud Machine Learning “a framework for building and training custom models to be used in intelligent applications”). While good data sets are hard to come by in the law, they are becoming more accessible every day. Soon, we will have all reported U.S. caselaw available to scholars and not long afterward, to everyone. With greater dissemination of the tools and more access to data sets, the power of computers to automate legal tasks will grow and at a faster rate than it has grown before.

No one should minimize the difficulty of jumping from what AlphaGo did to understanding and manipulating language and higher level legal thinking. But assuming some barrier exists is foolhardy. The question is not if, but when. Lawyers must take heed and fully engage in understanding how to work with computers more effectively as part of an augmented lawyer practice. It’s time to go.


About: Ken is an Adjunct Professor at Michigan State University’s College of Law and a Member of its LegalRnD Faculty. You can follow him on Twitter @LeanLawStrategy, connect with him on LinkedIn, and follow him on Facebook.