“Néhány AI-rendszer olyan magabiztosan hazudik, hogy azt gondolhatjuk, biztosan igaza van…” – interjú Theodore S. Boone-nal (2. rész)

Theodore S. Boone amerikai ügyvéddel, a Dentons Budapest Of Counseljével, a Corvinus Egyetem Gazdasági Jog Tanszékének oktatójával és az Amerikai Magyar Kereskedelmi Kamara korábbi elnökével készült interjúnk második részében azt vitattuk meg, hogy milyen kihívások elé állítja a jogi szakmát a mesterséges intelligencia, és hogyan kell a jogászoknak ezekre a kihívásokra reagálniuk. Az interjú első részében Theodore S. Boone Illinois-ban kezdődött, New Yorkon és Washingtonon át Budapestig tartó pályafutásáról és a szakmában megjelenő egyéb kihívásokról beszélgettünk.

How would you compare the past challenges that the legal industry has faced with recent developments involving Artificial Intelligence?

In Shakespeare’s play Henry VI, an unsavory character and pretender to the throne named Dick the Butcher suggests the following as one of the ways the country can be improved: “The first thing we do, let’s kill all the lawyers”. Is Artificial Intelligence today’s Dick the Butcher? Is AI going to kill the legal profession? Will that be a tragedy of Shakespearean proportions (at least as far as the lawyers are concerned)? I don’t think so. AI’s potential day-to-day use in the legal profession is certainly a new situation. However, I still see AI as part of an ever-ongoing evolution of the use of technology in the practice of law. Do not get me wrong, the rise of AI is a fundamental and important technological development, but it is not in a category by itself. In the late 1990s the Internet began being accessible at every lawyer’s desk. That happened in about 1998 at the large Washington, DC law firm that I was working for at that time. Can you imagine a time when lawyers did not have ready access to the Internet to conduct research and daily business? Think about that. Another example is handheld email. Prior to the early 2000s, there was email, but you did not carry it around with you. Starting in about 2002 mobile email came into widespread use. This was also a big jump because lawyers could maintain constant written contact with their clients and others. There have been many technological developments that have been significant and have had a strong impact on the legal profession – you can even point back to the invention of the telephone itself if you like.

Technology impacting the legal profession develops the same way a child grows, not always at an even pace but sometimes in rather dramatic growth spurts.

The rise of AI is such a growth spurt. However, I do not see this rise as some sort of existential threat to the practice of law, but rather another important step. In 1950 the logician A.M. Turing wrote a remarkably prescient article titled “Computing Machinery and Intelligence” in which he discussed what we now term supervised learning and unsupervised learning and in which he analyzed the question as to whether computers can think. In the article, Turing set forth and discussed what has now become known as the “Turing Test” and which he referred to as “The Imitation Game”. The Turing Test involved putting a computer in one room and a human in another and having a second human, who was separated from both in a third room, ask the first human and the computer questions in an effort to determine which room the human was in and which room the computer was in. Turning predicted way back in 1950 when he wrote the article that by the year 2000 the human questioner would have about a 70% chance of determining which was which after five minutes of questioning. Well, what would happen if you performed a lawyer version of the Turing Test today? That is, what would happen if you put a lawyer in one room and a computer in another room and asked a client who was separated from both to determine which was which? I think the client would be able to tell which was which fairly quickly, particularly once it came down to asking for actual legal advice such as analyzing an ambiguous statute or contract term or posing questions on matters such as negotiation strategies and risks. What lawyers and law firms have to do is embrace this new technology quickly but thoughtfully.

Some of the areas where I believe there will be an even stronger use of AI in the near future is in litigation risk analysis, transaction due diligence document reviews and the preparing of initial redrafts of contracts proposed by another party to a transaction.

Lawyers must utilize AI creatively to serve their clients even more effectively and to make themselves even better lawyers. This is certainly something that needs to happen and is going to happen. Lawyers and law firms which stand on the sidelines do so at their own peril.

Those prior developments you mentioned helped do the work, but they were not threatening actual legal positions. Now there is a fear among lawyers and other professionals that in a few years AI technology will replace our jobs. What are your thoughts on that?

According to legend, a reporter once asked the American writer Mark Twain what comment he had about the rumor that he was dead. Mark Twain is said to have replied ”Reports of my death have been greatly exaggerated.” I think that goes for AI’s impact on the legal profession as well. I developed and now teach a seminar at Corvinus University’s School of Business here in Budapest titled “Business Leadership in the Age of Artificial Intelligence.” In this seminar we examine such issues as digital transformations of companies and the business, ethical, economic and legal aspects of our new Age of AI. Part of what we examine is the impact of AI on jobs. AI will be a very useful tool for lawyers but will not replace lawyers en masse. There will however be significant adjustments as to how lawyers work and how law firms are structured and operate. Law firms will go through their own form of digital transformation, what one might call an “AI transformation.” When I was in law school in the US back in the late 1980s if you found a court case that you wanted to use for your legal arguments or writing, you had to by hand hard-copy book by hard-copy book research in the library if that case had been subsequently overruled. It could be a major project to be able to simply figure out how that case was treated in subsequent court decisions. Then, a few years later, you could do much of that research with a touch of the button on a computer. To me, here is the analogy:

if you have a serious pain in your stomach, would rely solely on the best AI system imaginable for your diagnosis? Or would you also go see and discuss it with a flesh and blood doctor?

This analogy, although not perfect, holds for the legal world. At the end of the day for anything really important to you, you are going to want a well-qualified human lawyer at the top of the pyramid overseeing any work of an AI system and making the ultimate analysis, recommendations and decisions.

But if the matter was not that important, maybe we would trust just the AI. Take Property Purchase Agreements or Standard Terms and Conditions which are very common and not the hardest tasks that are addressed by lawyers. Some could think that their drafting could easily be replaced with AI without the need to see a lawyer at all. What are your thoughts on that?

I understand what you are talking about – let’s call them the “everyday transactions”. For example, you are referring to agreements such as a Non-Disclosure Agreement or a Purchase Agreement for a simple real estate transaction. I understand your point, but I would respectfully not entirely agree with it. Even before AI became so sophisticated people have tried to use form legal agreements. Companies have tried to sell form agreements online for years and you can find them on the Internet. But if you are a young person starting out with not too much money and you are buying an apartment that is a very important transaction for you, even if it is not the most expensive or luxurious apartment in the world. For such a transaction I would suggest you would still want a lawyer involved. You can use a form agreement or you can use an AI generated form. But almost certainly you are going to want a lawyer to review it and that lawyer will almost certainly need to adjust it. I have never been involved in a transaction, whether it is a “big” or “small” one, where there have not been particular issues that have come up which need to be discussed, negotiated and addressed in drafting with pin-point accuracy and in a manner which is keyed to the specific transaction. You want for example to buy an apartment that you love, but it turns out there are some water spots on the ceiling that need to be addressed – there are always specialized issues.

The likely growing use of AI as a legal assistant is however an area that will make it more difficult for recent graduate lawyers looking to start their careers. AI systems soon may be able to do some of the junior work that beginning lawyers used to do. The AI systems will be able to do this work at times quicker and much more cheaply, both for the law firm employer and for the client. For example, in the past a junior lawyer might have drafted the first draft of a simple Share Purchase Agreement which was then reviewed by a more seasoned lawyer in the firm. Now, an AI system might draft that first draft and the AI draft would then be reviewed by the more experienced lawyer. In the past and still now, junior lawyers did work that was very useful for the client, but they also learned through that process and they learned under the supervision of a more senior lawyer. Junior lawyers coming into the workforce need to understand this issue and think about how to attack the problem head-on. Junior lawyers need to be able to get to the point where they have the knowledge and the skills to be comfortable as someone who is supervising not only younger lawyers, but supervising AI driven work product as well. How can junior lawyers as well as law students do that? Here are a few suggestions. First, read as many basic forms of agreements and as many actual final fully negotiated agreements and other real world work product as you can. The US’s Securities and Exchange Commission has an amazing web site called EDGAR at www.sec.gov which contains a treasure trove of complex real world final form legal documents that have been filed by public companies. I suggest using that data base as a source. Second, beginning lawyers should very actively seek out more senior lawyers as mentors. Senior lawyers are generally very open to helping educate the next generation if you just ask. Once you have those advisors and mentors do not hesitate to ask them “why” – why, for example, a contract clause or court filing or other document ended up the way it did. Third, become part of projects where human client contact and negotiations are an important part of the process.

It is hard to imagine how AI could ever replace the human dimension of the practice of law – so develop your skills in this area.

Fourth, seek out opportunities to speak at legal conferences and participate in substantive subject matter-oriented bar related activities and moot court and case law competitions. Fifth, stay on top of the most recent legislative proposals and enacted laws in the areas of your expertise or interest. When a new law is proposed or enacted it is almost invariably subject to interpretation and you will be able to contribute in this area. Indeed, work to write articles for legal publications about new and proposed laws (and ask your mentors to comment on your drafts of these articles). Sixth, dig deep into the AI technologies themselves to consider how they can best help you yourself do your work. Seventh, think about how the rise in the use of AI in the legal professional is itself going to create new legal issues and legal work that must be addressed. For example, how does one protect client confidential data and still make use of AI systems which run on and learn from technology and systems external to your firm?? How are intellectual property rights of authors going to be protected when some AI systems could be using copyrighted articles to provide responses without attribution? AI is a tool but remember there have been many other new tools throughout history. The wheel was a fabulous ground-breaking invention too. But I do not think most people were too upset that it put some horses out of business. People adjusted, methods of transport changed and many more new jobs were created. The bottom line is that junior lawyers and law students are going to need to be very proactive in working to get the experience they need to advance. By the way, be careful about those so-called “Standard Terms and Conditions” you mentioned – what is “standard” to one party could be dangerous, risky and far from standard for the other.

From what you say I see two areas of the practice of law that will not or probably will not be replaced by Artificial Intelligence. One is experience and the other is the legal judgement. What else do we need to keep in mind?

In Ancient Greece Pythia was the high priestess of the Temple of Apollo at Delphi. Pythia was known as the Oracle of Delphi. She was the most prestigious and authoritative oracle among the Greeks and was among the most powerful women of the classical world. Pythia was said to have provided prophesies in her own voice. A lawyer supervising work product generated from AI is going to need to have to have the confidence to question what has been generated and not view the AI work product as have been generated by some sort of infallible Oracle of Delphi of our digital age. Some of the AI systems that exist now, according to a term being used today, “hallucinate”. I have a more direct word for this phenomenon.

Some AI systems lie to you and they do so with so much confidence that you are tempted to think what the AI system is saying has to be right because the AI system is saying it proudly and without reservation as if it is a fact.

A lawyer has to have the knowledge, the skills and perhaps most importantly of all the confidence to question those results. Otherwise, you could think, who am I to question that result? Who am I to question today’s Oracle of Delphi? By the way, certain AI systems in use today provide absolutely no source footnotes for their supposed factual statements, making it hard to check the origin of the alleged facts they present.

In terms of additional skills, one of the most important, which I cannot picture that the best AI system is going to be able to do anytime soon, relates to the human element. At the end of the day, we are human beings living in a society and encountering situations in which we need to make decisions. Human beings are going to want to deal with a fellow person of that same species when examining critical risks and making important decisions. This does not mean clients will not want to keep their fees down. This does not mean that clients will not want efficient work. But at the end of the day, clients want the comfort of being able to look a human being in the eye and discuss key points and make sure an expert, a seasoned human expert, in that field is there helping make critical decisions and advising on important decisions. Another critical skill to develop is knowing how to work effectively as part of a team. This skill can be developed not only through legal work but also in other aspects of your life. Some of the most important skills I apply in my legal work came from being a member of sports teams and then coaching sports teams.

Do you prepare yourself for the new era? Are you taking webinars or doing other training to improve your skills so you can stay competitive during these times? And what would you recommend for other lawyers?

Absolutely. I always try to improve myself. No matter what level one is at – whether you are starting out or somebody that has a number of years of a career, whether it involves AI or another field, continuous learning is key. I am certainly doing a lot to think through how AI can best be used as a legal tool. One needs to not be intimidated by AI systems but rather approach them with the question: how can this system help my workplace and me in my work? I certainly do that, both in a practical sense and in keeping track of relevant legislation. For example, the European Union issued about two years ago the first draft Act of the EU seeking to regulate AI. This draft is currently being debated within the EU.

Looking into the future, what could you tell us about that draft legislation?

If the draft EU AI Act is finalized, enacted and enters into force, it has the potential to have a fundamental impact not only on all the countries within the EU, but on countries outside the EU as well as it would have an extraterritorial application for any country around the world outside the EU doing business in the EU. If the EU AI Act comes into force according to its current framework, it would place AI systems into three possible groups. The top group would be AI systems that are prohibited, such as real time facial recognition systems or systems that without your awareness manipulate your consciousness. The second group would be AI systems that are deemed to be systems operating in high-risk areas, such as energy, infrastructure and education. These AI systems would be subject to extensive oversight. How much do you want an AI system looking at applicants to a university, making predictions as to who would be successful and making recommendations as to who should be accepted? The last group addressed under the draft EU AI Act would be low-risk AI systems. These low-risk AI systems would not be subject to a lot of oversight.

To show you how difficult these issues are from a legal perspective let me touch on two of the main subjects which over the last two years the various bodies in the EU have not been able to come to agreement on yet when debating the draft EU AI Act. These two subjects are first defining what actually is an AI system and second what AI systems should be in the high-risk/heavily regulated group I mentioned.

What is an AI system? Is an AI system ChatGPT or does it also include your Spotify account? Your Foodpanda app?

You do not want to overregulate AI and stifle innovation but at the same time you want to protect people. The second major open area of discussion within the EU is what AI systems, once they are defined, would go into that high-risk basket because those are the AI systems where there would be a lot of oversight and reporting and review requirements. Breaches of the EU AI Act if it is enacted could result in fines as high as 30 million EURO or 6% of the total worldwide annual turnover of the violating entity for the year preceding the violation, whichever is higher. The EU AI Act, because of its subject matter and its proposed extraterritorial effect, has the potential to not only to have a strong impact in and of itself but also to serve as a model AI law for other countries outside of the EU much in the way the EU’s General Data Protection Regulation has served as a Gold Standard model in the data privacy area in the eyes of many. Very interesting. So, there are a lot of ways that a lawyer can stay on top of matters related to AI. It is important to work hard at it, but also not be intimidated by it. As a general matter, as one contemplates this new AI driven world into which we both as lawyers and simply as persons are being born I suggest it is useful to keep in mind the words of the renowned Hungarian born mathematical genius János von Neumann who among his many accomplishments developed one of the first computers in the basement of the Institute for Advanced Study at Princeton, New Jersey using what is now known as the von Neumann architecture. Von Neumann wrote in 1950 the following words in his manuscript The Computer And the Brain which words I submit are as true today as they were then: “The human brain is, after all, the best example we have of an intelligent system.”

Az interjú első részében Theodore S. Boone életútjáról és karrierjéről, illetve a mesterséges intelligencia megjelnéséig felmerülő, a jogi szakmát érintő kihívásokról beszélgettünk.