Artificial Intelligence - the Terminator or Change Maker of Legal Industry?

Chenchen Zhang

This article was published in Chinese (in October 2016); here I translated it into English.

One of the most gripping events across the world this March would be the Go match between Lee Sedol, one of the world’s top Go players, and AlphaGo, a specialized artificial intelligence (AI) for Go developed by Google DeepMind. At first, people were astonished at how well AlphaGo played; then they were frightened by it, as the AI beats the world champion at 4:1. People once thought that Go was the most complicated board game ever designed, and that computers would never be able to beat the best human players, simply because the computational resources required to find a best move is ten of the six hundred power. Equipped with Monte Carlo Tree Search algorithm and two powerful neural networks, AlphaGo can improve itself through learning for other matches and self-play, and is able to acquire a powerful strategy over variables of exponential growth. This competition between human beings and machines forces us to consider several serious problems. Who will control the future world – human beings or machines? Will machines seek to enslave or even exterminate human beings? If this day finally comes, what can people do to protect themselves?

From a narrower perspective, recent advances in artificial intelligence force us to think what kind of career we should seek in the future, so that we will not be replaced by artificial intelligence? It seems obvious that practicing in law is not one of the options.

Let us take a look at the most mundane work in legal practice - drafting contracts and other legal documents. It is hard to imagine how many such documents a lawyer, a prosecutor or a judge has to draft during his or her entire career, while most of the content of these documents is repetitive. Now, these processes have been simplified with automated systems - for example, for an employment contract, the system compiles the document automatically, where the employee only needs to fill out a questionnaire to provide the basic information such as the name or age. More complicated documents, including the legal documents for large loans, can also be compiled in such a manner. Lawyers can save considerable amount of time when their law firms are equipped with such systems; needless to say, it will further save money for law firms. As for the standardized documents that are much simpler, hiring a lawyer to draft them becomes unnecessary.

This has come closer to reality. A company called LegalZoom provides individuals and small businesses with an easier, less expensive and more efficient option when they need legal documents. Using its online platform, one would only need 15 minutes and 69 dollars to draft a will. Early this May, a law office announced that they would “hire” Ross, an artificial intelligence “lawyer” based on the technologies provided by IBM. Ross’s main responsibility is legal research - its human colleagues (usually professional lawyers) will ask questions to it (or should we use “he”?) in natural language, and it will rapidly respond with the precedents and articles with the highest relevance.

Machines can even do more. Last year, Joshua Browder, an undergraduate from Stanford University, created a website which used artificial intelligence to answer simple questions about appealing unfair parking tickets. This year, Joshua improved the website; now it is an artificial intelligence lawyer chatbot. It will first decide whether an appeal is possible by asking the user some questions and collecting relevant information such as whether there are clearly visible parking signs around, if positive it will guide the user through the appeal process. More encouragingly, it can also answer some simple legal questions, such as “I have experienced this accident. How can I request payment from my insurance company?” During the development, the biggest obstacle is to let machines “understand” how different expressions can have the same meaning. Finally, he used machine learning to allow machines to learn people’s expressions. This chatbot would extract features from the text, such as keywords and the sequence to perform “self-study”. The more people use the system, the more data it obtains and the more comprehensive it will become. Using the system is not only efficient but also free of charge - until the end of June in 2016, Joshua’s machine lawyer has helped people successfully appeal 160,000 unfair parking tickets, which saved over 4 million US dollars. In other words, the machine lawyer starts its competition with human lawyers through “learning” by itself.

Creations of this kind are unprecedented. The obstacles people once assumed to be insurmountable during the legal automation will be overcome in the near future. There are two key technologies for these tasks - one is the natural language processing capabilities that the machines have yet to fully possess, the other is the expressing “the ability of thinking as lawyers” in the form of algorithms, which legal practitioners always take pride in. The only problem between the prospect and reality is how developers can obtain a variety of datasets under the monopoly of publishers such as Thomson Reuters and Elsevier. Some companies are already making efforts to fix the problem. A company named Ravel is cooperating with the Harvard Law School Library, scanning and uploading almost all the judicial decisions in American history. Searching for the results is free - the premium is only required for more complex analysis tools. The database is still being compiled.

Now, Joshua’s system has made a huge step in terms of natural language processing and expressing “the ability of thinking like lawyers” in the form of algorithms . These two factors will no longer be obstacles in the near future, especially when artificial intelligences with the capacity of AlphaGo become widely utilized. And the foreseeable development of open-sourcing that allows the legislation, precedents and standard documents of every country to be accessible online will break the monopoly posed by publishers. Legal practitioners, at least people who depend on law to make a living, can no longer feel safe under the shield of the expensive legal databases.

No matter whether we are willing to believe or not, we must accept this possibility, that machines will take over most of the work done by today’s legal practitioners. They will never get tired, they work more efficiently, they do not demand promotions or pay raises, and are impartial and incorruptible. Lawyers will still be needed at court, but the demand for junior level lawyers will be significantly reduced. As for judges and prosecutors, given that the judgment of machines might not be trusted by the public but at least it can serve as a reference point, the demand for judges and prosecutors will also decline.

These revolutionary changes will indeed frighten law school students. But from another perspective, law itself never functions of providing law students with a means of livelihood; it never does. One of the most important functions of law is to help people resolve conflicts. And if artificial intelligences can resolve conflicts more effectively, more efficiently and cost less, why is it not a good sign? Take Joshua’s artificial intelligence lawyer chatbot as an example, it can now provide legal aid to homeless people in the United Kingdom, helping them to get temporary shelter aside from helping people appeal parking tickets, starting from this August. This AI tidal wave might cause many law students to become unemployed, but will benefit the masses as a whole.

Meanwhile, to ensure its survival, legal education must discard its old practice of focusing on statutory interpretation and applying precedents, and discuss policies and politics where the rules are blurry. These discussions are what the mainstream legal education in China has been avoiding all the time. Only through these transformations will legal education show new signs of life.