EY refers to the global organization, and may refer to one or more, of the member firms of Ernst & Young Global Limited, each of which is a separate legal entity. Ernst & Young Global Limited, a UK company limited by guarantee, does not provide services to clients.
Insights
Asking the better questions that unlock new answers to the working world's most complex issues.
Services
EY helps clients create long-term value for all stakeholders. Enabled by data and technology, our services and solutions provide trust through assurance and help clients transform, grow and operate.
Industries
Discover how EY insights and services are helping to reframe the future of your industry.
See more
Case studies
Strategy and Transactions
How carve-outs positioned an automotive giant for future growth
12 Apr 2024EY Global
About us
At EY, our purpose is building a better working world. The insights and services we provide help to create long-term value for clients, people and society, and to build trust in the capital markets.
See more
Top news
Press release
Vellayan Subbiah from India named EY World Entrepreneur Of The Year™ 2024
07 Jun 2024Eric J. Minuskin
Press release
09 Apr 2024Michael Curtis
Press release
EY announces acceleration of client AI Business Model adoption with NVIDIA AI
20 Mar 2024Barbara Dimajo
Recent Searches
Trending
-
How can you adapt your IPO strategy in a dynamic market?
The EY Global IPO Trends Q2 2024 covers the news and insights on the global, area and regional IPO markets for the first half of 2024. Learn more.
26 Jun 2024 Private business IPO Start-ups Entrepreneurship Private equity -
How can the moments that threaten your transformation define its success?
Leaders that put humans at the center to navigate turning points are 12 times more likely to significantly improve transformation performance. Learn More.
15 Apr 2024 Consulting Transformation Realized Workforce -
Artificial Intelligence
EY.ai - a unifying platform
16 Nov 2023 AI
Revolutionizing law with AI
Join us as we kick off season 7 of Better Innovation and our mini-series on Responsible AI with guests Benjamin Alarie and Abdi Aidid.
Related topics
Join us as we kick off season 7 of Better Innovation and our mini-series on Responsible AI with guests Benjamin Alarie and Abdi Aidid, co-authors of their new book The Legal Singularity: How Artificial Intelligence Can Make Law Radically Better. Ben is the co-founder and CEO of Blue J and Abdi teaches in the areas of civil procedure, torts, and law & technology at the University of Toronto.
Jeff, Abdi and Ben explore the concept of “The Legal Singularity,” a future state where technology facilitates the functional "completeness" of the law. The conversation spans the implications of the law’s textual nature, AI’s impact on the professional responsibilities within the practice of law, and the potential for AI to radically improve our legal processes, institutions, and society.
Key takeaways:
- Currently, the law suffers from incompleteness, uncertainty, and is under-specified.
- The Legal Singularity presents a future state where innovative technology facilitates the functional "completeness" of the law.
- AI will have profound implications for law and policy making, across the entire legal profession.
- The integration of generative AI into the law facilitates the prediction of certain legal outcomes based on vast amounts of digitized legal information.
For your convenience, full text transcript of this podcast is also available.
Jeff Saviano
Ben and Abdi, welcome to Better Innovation. We haven't had two guests together on the show in a while, so we're thrilled that you've come with us today in our virtual studio to talk about such an important topic. Let's orient our audience to your voices so they know who's talking. Ben. Let's start with you. Thanks for coming on the show today.Ben Alarie
Thanks for having us,Saviano
It's a real pleasure to see you again. It's so great to have you back.Saviano
And our audience will remember, Ben, this has been second time on the show all the way back in season two, Episode three, October of 2018. We had a TEDx recording in our office and Ben did a TEDx talk, and he was so gracious to come on the show with us. So great to have you on again.Abdi. Your first time. Welcome to a better innovation.
Abdi Aidid
Jeff Thanks for having me. Really excited to be here. Thank you.Saviano
Okay. I still remember Ben, that great TEDx event that we had. Our advanced tech lab is right in Kendall Square, right practically on campus with MIT. I don't think we've had as many people in the lab since that day. We had about a hundred folks. We had that famous red dot in the corner, and I loved your talk. It was such an important, I think, an important contribution in the legal tech space. But that was a fun day, wasn't it?Alarie
That was a fantastic day. I was so nervous. Jeff when you invited me down to participate in the event and I was intimidated and, you know, giving a TED talk is something that I anticipated was going to live on the Web forever. And I think that's true. It's still now on the Web several years later. And so, I was feeling a lot of nerves, but it was really, it was a really fantastic event. And I'm really grateful that you had me. You had me there.Saviano
It was an important step in the journey, wasn't it? I think. And we're going to we're going to talk a lot about your book today just to get this out early. I love the book. We're in this space. And I do think it's an important contribution. I love the Law Review article that I think was also a bit of a precursor to it. As we record this episode today, in mid-July, your book was published just a couple of weeks ago, right, on July 4th. So why don't we jump right into this and let's talk about the book, Ben, Maybe you want to you want to kick it off. What's the book about?Alarie
So, the book is about the legal singularity and the subtitle of the book is How Artificial Intelligence Can Make Law Radically Better. This book is an outgrowth, as you suggested, of this Law review article that I published back in 2016, and it was a really brief contribution to a symposium edition of the University of Toronto Law Journal. There were maybe a dozen pages long, and I introduced this idea of where does artificial intelligence take the law? And the idea was, well, as artificial intelligence becomes more and more powerful, it's likely to lead to the practical elimination of legal uncertainty. And that's going to have profound implications for law making, for policymaking, for the legal profession, for legal education, for adjudication and given my academic interests for tax law specifically, which is where I spent a lot of my time thinking about it, and this 12-page article, Jeff was received with enthusiasm. It was received with a lot of anger in some quarters, people were irritated by the idea that, you know, technology might have implications for the contents of the law and law making, and not everyone was as charitable and open to technology as you are. And it spurred a lot of criticism. And that led to me and Abdi sitting down and saying, you know, actually maybe we should mount a like a prolonged explication of this idea of the legal singularity. And what is the legal singularity? Let's address the critics of the legal singularity and let's try to steal man this argument, this position, and have something that can stand the test of time as technology is inevitably going to keep improving and I think is going to vindicate some of the things that we say in the book. And you know, after four years of effort, you're right. The book was released two weeks ago. I was thrilled to see it as the number one new release in data science on Amazon.com. And I'm thrilled to be here to plug the book and discuss the ideas in the book with you today, Jeff. And thrilled to have Abdi here, too, with us.Saviano
It is so it's so nice to see it out in the world. And I love early in the book you describe it as it's the story of the convergence of technology and the law, and that had struck me. I think it's a great it's a great way to explain it. And that is it is this convergence. And I also love that that you went right to the critics didn't ignore it and, you know, you didn't bypass their concerns. And I think you recognized in some cases the concerns were valid. But you devoted a whole chapter. And I appreciate that as well. Abdi, love to hear your ideas about the concept of legal singularity, that that that word singularity may resonate with some in our audience, some who are technologists, but it may not resonate with everyone. So, explain what we mean by the concept of legal singularity.Aidid
Yeah, so Ben really nailed it when he said the legal singularity is about the practical elimination of legal uncertainty. We're talking about a future state where the law is functionally complete, where legal knowledge is available to us everywhere in real time and on demand. The basic idea here is that if you think about the status quo, we do a lot of things that are great in the law, but there are some ways in which the law is incomplete, uncertain, sometimes under specified, and sometimes sort of hard to put your finger on an elusive. And so, if we leverage computing power and what it's good at, we recognize what we're good at as individuals, we sort of usher in the appropriate division of labor between human and computer. We can achieve a lot more in terms of not just steering clients, not just being able to present cases more effectively, but also restructuring our society in a way where legal knowledge is accessible and helps to inform our behavior in positive ways. And so that's really the legal singularity. It's a future state that we think we're trending towards given the proliferation of AI and now the, what you might call the emergent embrace of AI among the legal community. And so, we're really talking about a future state now that's really important. And you mentioned the TED talk. One of the things that Ben mentioned in that TEDx Talk some time ago was that it's not going to necessarily be a linear trajectory to the legal singularity. Right? There could be interruptions. There could be a messy interregnum period where we're trying to work out the contradictions and work out the kinks, so to speak. And that should remind us of it's a deeply human endeavor. Along the way, we can help facilitate it and help land the plane safely, but we can also get in our own way and slow it down. Now, the good news is that I think there's enthusiasm. I'm listening to you, Jeff, in your wonderful introduction, and I can see that there are people like yourself who are thought leaders in this space, but also with the advent of GPT and really the boom or explosion of GPT in March, you start to see people understand that this is where we're going. And so, what we're trying to do with this book is sort of make the case for embrace of legal technology, help mitigate against some of the risks which we recognize as being potentially quite real and valid, but also tell people, “look, let's come to the table and have the conversation”, because there's no way that that we're going to stop using these kinds of tools. You know, the economic cases to spawn on the social case is too strong.Saviano
I think there are some fields that embrace technology with great enthusiasm and at great pace. My own belief is I don't think the law is one of those fields. I think that that it can be it can feel as though it's a bit slow and it's complex and will unpack some of that. But I think there's an important element of this definition of legal singularity that I took from both of you article and Ben from our time, all the time we would spend together over the years. And that is that that the law will be knowable. It will be knowable with a high degree of practical certainty, but not that everyone will understand all aspects of the law. That's of course, that's you know, that's not practical, but that it's knowable and it's understandable. And I think that's an important distinction. But I just want to make sure that I'm getting that right, that in this effort for completeness and then to be able to understand it, that do I have that right, that the word knowable is an important element of the definition?Aidid
Absolutely. And, you know, the law being knowable or ascertainable is different than the law being highly simplified, we think is a good thing in many in many cases. Why? Because social facts are complex. Things that happen on the ground are complex, economies are complex, technology is complex, and you need complex laws to contend with that. So, when you hear a lot of access to justice advocates talking about simplification of the law, sometimes that can sort of undermine the law's purchase and its teeth. And so, what we're talking about here is, okay, in a world where law is complex, where the world is complex, how do we make sure that the law is understand able to us? How do we make sure that the law is as ascertainable at least? And so, when we talk about the law being knowable, we don't mean the law being highly simplified. Now, what then do we mean by knowability? Well, here's the reality of legal analysis today. You might get a bunch of different cases that stand for the same legal principle, and you have to sort of reconcile that and figure out where the sort of line is. What is the law on you know what? When is a worker, an independent contractor, an employee? Right. A classic tax law question. Right. That's in many ways a touch and feel test. You consider a bunch of, you know, non-exhaustive factors. You look to the regulatory guidance, you look at the case law, you look at past practice by the regulatory authorities, and you do your best to assess what's likely to happen. But you don't have a crystal ball. So, sometimes you're going on faith, sometimes you're going on judgment, and sometimes there's contradictions inherent in the decisions that have been made in the past. And so, what you rely on is the great expertise of lawyers to help you sort of navigate the pathway. Well, what if we supercharge that expertise with a sense of what's actually likely to happen? And how do we get there? Well, at Blue Jay we built a tool that allows you to predict how future courts are likely to rule in new legal situations. And what that does is it assesses all of the historical case law much more quickly than, say, you and I can, because we're busy people. We're people who only have 24 hours in a day. And rather than us trying to read a sampling of the cases and make a judgment, the compute, you know, using computing power, you can synthesize all of the historical case law in the area. That's an example of helping us get a legal answer rather quickly. Now, the good news is that as a lawyer or as a tax professional, you have much to contribute to that answer making because, you know, it's not just the case law that gives you.Saviano
Yeah.Aidid
The basis for your legal advice. You also want to be able to reconcile that against your judgment and your experience and your expertise and your knowledge of your client's case strategy or your client's business environment. And so, what we're talking about, we talk about unknowability in that way is how do we get to a place where we reduce the uncertainty by being able to take a broader, more accurate look. The other thing I will say about the law being knowable is it is true very often that the law is inaccessible. And when we talk about inaccessibility, there's inaccessibility of our actual practical difficulty accessing the information. Right. You know, think about that many times in your life where you've been doing some tax research, but the case is hard to find or there's a revenue ruling somewhere that isn't digitized orSaviano
Right.Aidid
You were trying to recall the great tax treaties that might be, you know, collecting dust in a library somewhere, but has the great explanation for the particular problem that you're confronted with. It's sometimes the stuff is hard to find.Saviano
It can be really hard to find. And we're going to come back to this notion of accessibility. I totally agree. First of all, I love all the tax law references. Of course, taxes. The most important aspect of I'm biased in that, but I appreciate all of these tax law references and which was also struck me and again, this was early in the book, is that I think the way that I had been thinking about it and this was from knowing the Blue Jays story from the earliest days to other aspects of computational law that I think it's easy to think about it from an output scale. Right. How do you interpret how do you aid in the interpretation and in the prediction? But also, in one of the elements of the book that I found fascinating is that there's opportunities for better legislation and for better policymaking on the front end. I want to take, I want to mark that because it just it really struck me and I had it actually conversation with a member of my team today about what I think one of the great aspects of the book explaining legislative intent, an opportunity to have better policymaking for better societal benefits. I want to go back to, and Ben I think to be interesting to hear your take on this as a legal educator and also, you're a legal tech entrepreneur and we've talked a bit about Blue Jay already that, you know, both you and Abdi are in a great position to address the complexity of the topic that we're discussing today. I'm just so curious, Ben, what motivated you to explore this idea of legal singularity? Not as much to write the book. Going back even earlier, the earlier days of I think it was 2016 when you wrote the Law Review article, is that right? 2016 Right. So, your interest goes back a number of years, but where did that motivation come from?Alarie
Gosh, it's complex. And to use a word that you just used in your question, Jeff, I think it's anything important that we do is it there are complex antecedents in the causal chain, but I'll tell you a story about it. It actually is. It grew out of my work as associate dean of the faculty of Law here at the University of Toronto. So, I am a full-time law professor here. I hold the Osler chair in business law here at the University of Toronto at the time. Back in 2012, 2013, I was spear heading a curriculum committee here at the law school. And I guarantee you this is going to come back to your question. But I was chairing this curriculum committee, and we were we were aspiring to do an overhaul of the J.D. program and how we're delivering a legal education. And it hadn't been done for 40 years. So, it was overdue. We needed to do this. And I started reflecting on the fact that it hadn't been really done for 40 years, and that caused me to have the thought, Well, if we're not going to do it very often, and if it's going to be another 40 years before we do this again, maybe we should be future proofing the curriculum to make sure that whatever we change this to, it's going to be safe in the future. And then that got me to thinking, okay, well, what are the big trends that might change the world of legal education over the course of the next several decades? And that that then I felt like I was hit by lightning, Jeff, because everything else follows from that, you know, artificial intelligence machine learning. We're at the University of Toronto. Remember, it's ten years ago, Jeff Hinton had, you know, the godfather of AI had been working on deep learning. So, in the headlines, DeepMind was starting to get working on, solving Atari video games. There's a lot of stuff in the air in 2013, 2014, 2015 leading to this Law Review article in 2016.Saviano
And you also found yourself judging a Watson competition. I love that story. I love, I think you told that on the show when you were first on about how you were judging an IBM Watson competition on campus, and something struck you that there was an application you saw you where I think later on appreciative that the dean came to you and asked you to fill those shoes. But that always struck me that that you never know where inspiration will come from.Alarie
Exactly. And it's so interesting. I mean, we're just coming out of COVID and, you know, we all and I'll speak for myself, I desperately missed being in the presence of others, didn't see so many colleagues at the university, these haphazard intersections, you know, running into people, having conversations with people, being invited to speak across campus at an event. These accidental collisions are so important for how things turn out in our lives. And so, you're absolutely right. And that's another part of the causal chain here about the origins of the legal singularity.Saviano
Well, and I also came away from the book feeling this such a strong sense of boldness, a sense of optimism that A.I. has the potential for a positive aspect on society. And we're going to talk about some of the risks and challenges. And Ben you're an optimistic guy. Did you always have this outlook on the potential of A.I. that your glass is half full? I know from knowing you and from reading the book and your other work, was there anything in particular that that shaped that optimism?Alarie
Gosh, that's a difficult question. I, I think I generally have a ton of faith in humanity, Jeff and I have a ton of faith on health and technology. And, you know, I'm not blind to all of the all of the problems that we create for each other and scientifically and, you know, climate change and existential threats to humanity, you know, nuclear proliferation, war, famine, pandemics, I'm not immune to noticing those things, but I also have a deep admiration for human ingenuity and creativity in rising to the challenges that we're confronted with and yes, new technologies have always come with new powers, new problems. And I always tend to focus on the benefits and how can we use these new tools to improve well-being. And I think it's kind of a story of the enlightenment, really. I mean, if you go back to, you know, the 1516 hundreds right up to today, it's not a straight line in terms of improved well-being, human flourishing, but I would challenge anyone to defy the claim that life is on the whole for the average human, significantly better. Now than it was several centuries ago. And for me, I spent a lot of faith in our ability to invent and to address challenges with new and improved tools, including A.I.Saviano
And you also believe in the rule of law. That goes without saying. As a law professor, you believe it's not a perfect system, but believe in the legal system. Sticking with some of the principles early in the book and Abdi, love to get your perspective on this in the early chapters, you discuss how everything in the law is confined to the text. Maybe it's a bit overreaching, but I think that's the language that you that you may have used, that certainly the vast majority of it finds its way into the text. There was a great quote from Lord Camden back in 1765 saying, quote, “If it is law, it will be found in our books. If it is not found there, it is not thought.” end quote. I thought that's a pretty absolute statement. Explain this idea and the text-based nature of the law.Aidid
Yeah. So, I'll start with encouraging yourself and your listeners to, you know, engage in a thought experiment here or rather just picture a courtroom or a law school class 150 years ago and try to imagine one today. Other than, you know, a couple of laptops that might be on the desks if you're imagining a courtroom other than maybe a screen where there might be some evidence being presented. By and large, they're going to look the same. Right. So, in many ways, what we've what we do in the law today and what we do in legal institutions like courts, like law schools, it bears a strong resemblance to what we have been doing all along. And so, if someone says they're engaging in legal research, they really mean they're reading, right? When someone says they're preparing their legal advice, they often mean they're writing. Right. Now, the way that they access that information might be different now. You know, you might be entering some search terms as opposed to navigating the stacks of brick-and-mortar library for hardbound case volumes. But for the most part, what we're doing hasn't been significant technologized. Right. And that's really the essence of our claim in the second chapter of the book, which is the history of legal information, is really the history of text-based systems and then proliferating. Now, what's the challenge with law being a text-based system? Well, number one, you know, the person who has the privileged access to the texts kind of runs the table. Right. But it also means that folks are, of course, limited by how much text they can access at a given time. So, we talk a little bit about I think this is in Ben's original Larvae piece. This legal scholar Oliver Goodnough talks about Abraham Lincoln as a lawyer and says Lincoln was only limited by the number of books he could carry on his horse.Unknown
Right. That's what his practice is limited by. And that was a strong but clear statement about how the world, the universe of what's possible in the law really depends on how much of the text you can access at a given time. So, if you're a small legal service organization and you're not able to afford the expensive subscription or the expensive spiral and often leather-bound volumes, then you have you're unfortunately considerably less capable. And that's a reality that we're still sort of dealing with in navigating through the technology provides a partial answer to.Saviano
And of course, it's I think it's particularly challenging when we think of cases and still in court. And so, passing the rule of law, framing judicial opinions. There was another you have to say that a great quote in the book from Oliver Wendell Holmes said, quote, “The life of the law has not been logic, it has been experience. And I think that's recognizing that the case-by-case approach, case by case approach of the common law. So, of course, at its core, the law is a text-based system of constitutions and codes, statutes, regulation, but also the secondary sources as well, and the treatises and academic articles. And we've mentioned Ben's Law review. Okay, let's set the stage for our listeners. I want to get to the why on why the application of artificial intelligence to the law is so important maybe to start with. What are the most significant shortcomings in the law today? Not a high, but for shortcomings in the law and our legal institutions that we think make the legal singularity necessary? Do you want to keep going? Abdi?Aidid
Yeah, happy to. So, one of them is something we've been talking about all along, which is the significant amount of legal uncertainty. Some of that uncertainty is because of ambiguity in the law, which sometimes is desirable, right? You don't always want the law to overprescribe because, you know, legislators can, and rule makers can't always anticipate everything that's likely to happen. But sometimes that ambiguity or uncertainty leads to significant consequences, like people not knowing exactly how to plan, businesses, not knowing what kind of regulatory environment or legal environment they're operating in. So, one of the issues is, of course, uncertainty. Another issue is inaccessibility. The fact that so many of the core bits of information that one might need in order to perform effective legal analysis is sort of difficult to identify and reach. Another issue is that the law is quite often, you know, let's say, not achieving its desired effect. So, for instance, you have laws that don't have the deterrent effect that you might aspire for them to have. You have some laws that are so under specified that they're not capturing behavior and, in the domain, they're trying to regulate. You also, of course, have the issue of laws sometimes being easy to elude, laws being easy to contract around to not comply with. All of these things are functions of that background uncertainty and the underlying elusive nature of the law. And so, technology has a lot to say to this. Number one, in allowing us to have improved predictive capabilities and the ability to sort of foresee how the law is likely to play out, it helps us not only plan better laws, but gives us a much more effective mechanism by which to understand and comply with laws, which I think is crucial now, all of that is that's even before you get to the legal institutions. That's really just about knowledge of the law. Well, what about when we get to our legal institutions? So, what about when you get to courts, which, you know, one of the few things that the legal profession can sort of agree on altogether is that there's an access to justice crisis. Some of that is that courts are clogged. They have more cases than they know what to do with. Some of that, you know, a significant proportion of those cases are cases that maybe should have settled, maybe cases that shouldn't have entered our courts in the first place, cases that maybe should have been resolved through some other means privately through our normal sort of social relationships. Right. So, if you can get to a place where you have an understanding of how cases are likely to play out, we can really reserve our courts or rather preserve our courts for the kind of tough, sticky situations that they're well positioned to help us navigate. Yes, and not the garden variety disputes, which really, we might have better mechanisms to resolve.Saviano
It's such a great way to put it. And we're going to we're going to come back going to spend more time on this idea of prediction and it'd be helpful, Ben, if you could talk about and we haven't really introduced the concept of computational law yet and I thought in the book was a great description of the history and the foundation of the concept in order to, I think, explain computational law, you have to explain what we mean in the analog era, what was the digital era of legal information and how that now is has shifted to computational law. But if you could then just tell that story of how we went from analog to computational and what that what that means in terms of legal information.Alarie
Sure. This ties back into the idea of laws text Jeff that you’re discussing about the earlier if we think about law as principally being about text of course you know the discussion is about books as physical artifacts in the world commonly housed in in law libraries and law offices elsewhere. This is really what we mean by the analog age of legal information, information captured in books as physical artifacts, capturing words relating to law as legal information. There's a massive change of course with the introduction of digital computers back in the 1940 and 1950s. It took decades for this to play out really practically for professional services firms and for lawyers generally.But there was essentially a re-platforming of a lot of legal information from these physical artifacts, from these books into digital formats. And so, this is the mass digitization of legal information, the scanning of countless legal texts, making those available initially on, you know, I'm big servers that that folks will have to dial into and then later in CD-ROMs on DVDs and then eventually in the cloud and then different search technologies allow you to find the right digital information, the hyperlinks that link all of these legal texts. But fundamentally, we're just talking about taking legal text from a physical artifact and recreating it digitally. Now, when we talk about computational, this is a step further, and I think we're really starting to see the power of this with generative A.I. And it's so exciting to see what's happening in the world with generative AI. But the whole idea behind computational, the computational turn in legal is that now that we've got all of this text, this legal text digitized, we can apply intelligent algorithms to all of this data and do so much more with this legal information than we ever would have conceived of as being possible 100 years ago. It's unless you're over Oliver Wendell Holmes. Right. Who actually did foresee this 130 years ago? Virtually nobody else did. And it's really exciting. And what does it mean? It means we can start to predict legal outcomes based on a vast mountain of digitized legal information that make inferences. We can draw concepts and really fundamentally engage in creative legal research, leveraging these tools. And that's the computational term.
Saviano
What's exciting too, is that it's not as though there are multiple decades or certainly centuries that separate this analog to digital, to computational. I graduated law school in 1992, I remember jeopardizing cases with the books. I actually had a friend in law school say to me a short time ago, Aren't you glad that we've had that experience? And I looked at it and said, Well, not really. I don't really look at that. I don't really look at that fondly. But I guess you do learn something from having to go through. But think it wasn't that long ago even that analog to digital transition. And I think the point that you made earlier about I love the analogy of, you know, what, Lincoln could carry on his horse, but let's face it, lawyers rely on limited resource. And, you know, any human law, you can only read just a handful of cases that that may relevant. It's the intuition and the experience of lawyers that will really, that will really make a difference. But you can only absorb just a handful of what's in the law. There was a study that you cite in the book that really struck me by Grove and Neil, I think I'm pronouncing that right, that algorithmic prediction outperforms human judgment and like 128 studies. So, it's not a matter of question whether there are some things that machines are just better. They're just better than we are. We as humans, as human lawyers, have biases and blind spots. And, you know, of course, the machines don't have that.Alarie
I mean, I, I will grant there's a ton of evidence, including the meta study that you cited, Jeff, just a moment ago, showing that that predictive algorithms kind of perform human experts, you know, in a wide array of settings for sure. I might want to mount a bit of a defense of humans, though, in response, which is to say those models are going to be as good as the information that they have available to them. And so often, yeah, we humans are bringing a slightly different information set to the predictive task. And so, this this gets us to the pairing of really sophisticated machine learning models with human judgment together, human expert judgment as probably being the best solution, the best method for coming up with predictions, at least for the time being. Right. So, certainly in chess, this was true for a long time that a grandmaster playing against a chess computer could be quite a rivalry. And eventually, you know, Kasparov lost to Deep Blue and then that made headlines around the world. But it remains the case that that computers plus a grandmaster would beat either the grandmaster alone or a chess engine alone. For some time now, it's gotten the point where no grandmaster is having any value to the top chess engines. There's too much computing power being deployed there. Those algorithms are now so sophisticated of the top chess engines. I think because law is such a deeply human enterprise, it's going to be quite some time before an algorithm is able to completely supplant the added value of humans in the law.Saviano
But maybe not ever, maybe not in it. Perhaps we may never we may never get there. I love the terminology collective intelligence. I think it's a great way to write the best a man and Machine and I also love the explanation. If you in in many fields, I'd say the law certainly applies here. If you're worried about your job as a lawyer being replaced by an algorithm, that's probably not a real fear. But the fear may be that you could be replaced by another lawyer who uses artificial intelligence more adeptly. And I think that's like that nuance, I think, is quite important. Okay. Let's go back to Abdi. Abdi, You were taking us down the path and then you just alluded to this as well about prediction. And let's go all the way back 1897, you provided this in the book, which again, I'm go back to 1897, that there was another quote from Oliver Wendell Holmes referred to the law as, quote, “systematize predictions”. I thought that was fascinating. He was way ahead of his time. You could mean that that terminology applies certainly today as we're discussing computational law. You discuss the prediction theory of the law, discussed that in the book as this notion of how legal rights require us to foresee what the outcome may be. So, Abdi, let's talk a bit about how these computational law tools are helping us make legal predictions today. How is that important? And talk a bit about how you think that will extend into the future.Aidid
Sure. So, the very essence of the prediction theory of law, which starts with Oliver Wendell Holmes and Holmes, like you said, was really ahead of his time on this and a bunch of other matters we won't get into today. But the essence of the prediction theory of law is that the very thing we do in the law is trying to foresee outcomes. All right. And that all legal statements are sort of imbued with this predictive element. So, the example that I like to share is if I say I own a home, I might be saying that my name is on a deed somewhere or my name is on a document somewhere, or I carry a mortgage or something like that. But what I'm really saying is if it came down to it, Jeff, and you challenge me for my home, I believe that a future court or some official body would vindicate my property interest in the home. Right. That's what I'm really saying in many ways. And so, this idea that all legal statements are inherently predictive is a core notion here, and it really helps us understand what we do in the law. So, if you have a client that walks into your office and they ask you a question, you know, does my, is my transaction likely be respected by the tax authorities?They're not asking you to describe the contours of the economic substance doctrine. Right. What they're really asking you to do is tell them what's likely to happen. Right. Am I going to face regulatory scrutiny? As you know, if this went to court, would I be in trouble? How much would I'd be on the hook for the same way that if I if someone walks into a criminal law office and they say, I've got into some trouble, they don't want you to, you know, describe the elements of an assault. They want you to tell them what kind of trouble they're in and what's likely to happen and what the risk is that they're facing. And so, the very thing we do as lawyers and legal professionals is we try to predict outcomes. Now, here's the thing. The essential truth of this, and this is what Ben was alluding to earlier, is that machines can help us do this better and more effectively because they can synthesize and assess more information than we're capable of much more quickly.
Saviano
And how hard that is and how difficult that is, especially for these vague notions of the law. And the law is loaded with legal terms such as bona fide or reasonable and that are developed through sometimes decades or longer centuries of judicial interpretation. Of course, it's impossible for lawmakers to enact narrow statutes that define every unreasonable act. So, is it fair to say that for these vague notions of the law using computational tombs as tools are even more important? Because it's difficult to pass that, isn't it?Aidid
Absolutely right. And some of that stuff, especially, you know, reasonable ness. So, I teach torts. And one of the things we talk about in torts, tort law is the reasonable person standard. Well, that's been developed over generations across multiple jurisdictions, across multiple centuries. And so, if I just look at a sampling, no matter how representative I think it might be of the case law in the area, I don't have the full view of the reasonable person standard. Well, if I run, if I have a predictive tool that enables me to synthesize all of this sort of a case law in that area, and then I reconcile it with my own judgment of my own experience, I just have a better answer. And the key thing here is that it's about using technology to start you off at a higher floor. Why wouldn't you want to help, especially when you're in a client situation where not only are clients often being faced with problems that could be existential in their lives, but you also know, often, you know, when we teach our students, we tell them, you know, very often when people are consulting a lawyer, it's because they're dealing with something they'd rather not be dealing with.Saviano
That's true.Aidid
Why not try to get them the help that they the maximum help? But also, if we're stewards of the rule of law, if we recognize that our profession is different because we swear an oath right, people don't always wear oaths in their jobs. You know, lawyers, believe it or not, we do. And part of why we do so is because we're supposed to be you know, we have some fidelity to this bigger thing, which is the rule of law. Why wouldn't you want to have more effective answers? Why wouldn't you want to deliver more effective client service? And then, of course, there's the fact that clients are scrutinizing the bills and they're saying, yeah, research. You know, I thought you knew the answer. Why am I charging? Why am I paying for 40 hours of legal research? Well, you can cut that down and deliver clients better value, too. And all that's possible through predictive tools.Saviano
I'm sorry I got stuck on when you said you were torts Professor. I remember. Still distinctly remember spending a week on this poor old woman. If I have her name right. This is Palsgraf who was on a train platform and was it a clock that had fallen on her?Aidid
It was just scales.Saviano
It was the scale that generated that. Yeah. I remember feeling so badly for Mrs. Palsgraf 's good, but it was it was her audience.Aidid
That's Palsgraf versus Long Island Railroad if you're interested.Saviano
There we go. There we go. And there were no I can assure the audience there were no notes that Abdi was looking at these things.Aidid
It's an important case. Sorry Jeff, I want to say one thing about that case. Yeah. What was so fascinating about that and is that it was a chain reaction that started with someone carrying a nondescript box onto the train that turned out to be fireworks and down the line, you know, there was a chain reaction. The person got pushed onto the train, the box came ajar and then exploded. Then later on the other side of the train platform, the scale hit Ms. Palsgraf, and that's how we get our modern law. So, this thing that Ben talked about, the happy accidents and the collisions, they lead to great things. That's actually how our law is made to in many ways, the law is a function of a bunch of accidental interactions that we then make pronouncements about. And there's ways for us to do that a little more intentionally. Now with technology, it's a great way to put it.Saviano
And this discussion is a really good segway into and Ben, if you could pick it up on this idea, what we've talked about a bit already, the notion of the completeness of the law. The law is complex. There are many conflicts and areas, of course, where it's incomplete. But then how can the fields of computational law improve the completeness and provide this clarity?Alarie
Gosh, you know, Jeff, I think the law as complex as it is, and normally the prop that I hold up at this point is a copy of a partial copy of the Internal Revenue Code or the Canadian Income Tax Act. And I talk about look how dense and, you know, how much how like just the density of the rules here. Look how complex the system is. And it really is. But I think it pales in comparison to what complete law would look like. And there are so many gaps. There are gaps everywhere in the law. And this is this is why we're kind of, we keep inventing the law and we have different tools doing this. One is the common law. So, we have this idea that we're not going to be able to foresee every single thing that is going to happen and come up with rules ahead of time to govern what's reasonable, what's unreasonable. It's why we have these amorphous, kind of ill defined, vague notions in the law. I think with artificial intelligence, I think what we're going to see, and you know, this is apropos of the current attention being paid to generative AI, we're going to be able to generate answers to novel situations, leveraging computational power, leveraging computational law going forward. And this is through a gap, filling an algorithmic gap, filling role that AI will play and basing it on, you know, a model assembled from all of the, all of the past experience. And before you said, you know, the life of the law is up in logic. It's been experience. I think the really, really interesting promise here is pairing the two. Pairing, really deep logic building these what are effectively sophisticated models of how the law has grown, how it has responded in the past gaps and then how did judges address that? And if we can build a model for how judges have been filling gaps and that model as a successful model, suddenly you have a system that's capable of in real time and on demand coming up with a solution to a legal problem that seemed unanticipated, so very difficult to actually foresee. But turning those models in on themselves and saying, well, how would a how would an astute judge fill in this gap with the resources that the judge has available? If we can predict that, then suddenly we have a way of filling in the gaps and we kind of make the law close to airtight.Saviano
And this may be obvious to our audience, but one of the problems with incomplete law, as you point out in your book, is that it will be an insufficient signal to the world and if there's little predictability in what outcomes are, then, then, of course, that will lead to chaos and it will lead to individuals not having a clear understanding of what ramifications of that signaling element, I think is incredibly important. Other point that really was interesting to read in the book is that is that laws may be clear when they're enacted, but over time as the world changes especially with as we're seeing in one of the areas our team is very focused on or some of the policy and regulatory issues with innovative technology and you can't draft the statutes fast enough to adapt to technology. So really, really difficult. Okay. I want to come back to and then stick with this, if you could, because I feel like we've come in and out of this issue about how legal institutions are striving for equity and in how the law is understood, but also access this question of accessibility, of legal support, really difficult to achieve in practice. If you're at lower or even the middle class, the percentage of people who need a lawyer but don't have access to one, it's just way too high. What are the implications of this disparity on the accessibility? How bad of a problem is this today?Alarie
I think it's a terribly challenging problem. I think it does have real social costs, social implications. I think it tends to do things that even undermine laws on claims to being a normal, highly desirable system. So, the rule of law presupposes that its subjects are equal before the law. If it's so difficult to get access to legal information, to legal advice, to even know where you stand on a particular matter, it opens the door to opportunistic behavior and opens the door to more powerful, more well-advised parties. You know, leveraging that position of strength and power to, you know, perpetuate unfairness, you know, the pattern of results that we see socially. And of course, those things then can feed on themselves and actually be perpetuated into the future as more advantage parties exercise legal powers in order to, you know, tunnel into those positions of privilege. It perpetuates unfairness and inequity. And so, I think it's a profoundly important point. And it manifests in different dimensions. It manifests with economic inequality, racial inequality. If you look at rates of incarceration across different ethnic groups in Canada, in the United States, elsewhere around the world, you can see it manifested in a bunch of different ways. And so, I think there's a real human cost associated with the inaccessibility of the law. And it's pernicious. It's extremely difficult to address. And I say this also as someone who's, you know, cognizant that the law is probably about as, you know, good as it has been historically right now in our leading jurisdictions around the world, the level of performance of the law is actually really quite good in historical terms. So, I'm not saying, you know, there was some, you know, past time when law was perfect. I'm not viewing the past rose-colored glasses. I'm just saying there's a lot more work that we need to do to allow the law to really thrive and show what it's capable of.Saviano
And go back to how we open the show. If it's competition law. It's also this notion of legal singularity means that the law is more knowable and it's more understandable than as that cascades than perhaps legal services will be more affordable. Perhaps there's even self-service that individuals, if it's knowable, that maybe you don't need high priced representation. Okay, as I'm watching our clock and we're coming towards the close of our show today. Abdi, want to come back to you because I know that this is an area that you're passionate about. I want to make sure we highlight the risks associated with AI and this practical and the impact that I has perhaps on professional responsibilities of various stakeholders, judges, lawyers, what have you. We've talked a lot about the application of AI in the law. Now, let's get to some of the ethical implications. What are those ethical concerns and these surrounding the use of AI and algorithms in legal prediction?Aidid
Yeah, it's a really important question. So, in our book we talk about as a first step, we need to disentangle the problems and try to have a framework for thinking them through and the problems that are posed by the use of AI in the world of law are some problems which you might call reflection and amplification problems and some problems which are really about the technology. The reflection and amplification problems are about the various ways in which the use of predictive algorithms, generative AI, sort of reflects back to us some things that are in our society, either in a subterranean way or already on the surface. For instance, if you're building a predictive algorithm, you're using historical data to inform it. And so, the outputs are going to be, of course, bounded by that historical data. Well, if that historical data is full of all kinds of bad social practices like racism or sexism, and the outputs will be reflective of that, and you saw this in some specific examples where algorithms are being used for things like predicting the right sort of bail and post-release conditions for people who've been convicted of crimes or helping to perform risk assessment for criminal sentencing. You saw very often the data that was being used to inform the algorithms were things like arrest data, were things like criminal histories, where things like previous convictions and judgments from courts about what makes someone likely to be to recidivate. And so, all of that, of course, is intersecting with the story of racism in the United States. And so, the outputs were not ones that people could fully countenance, right? They said that this is uncomfortable for us. Well, part of what we are saying in the book is don't allow technology to take us off the hook socially for resolving those problems. Right. The problem is a problem of racism. The problem is a problem of sexism or classism or the distributional problems of economic inequality. And technology is at best holding up a mirror to us in those kinds of situations. Right. And so, it's really, really important that we not lose sight of it, because if we try to tinker with the technology and say we're going to we're going to improve the tech such that it resolves that underlying social problem, well, at best we're masking it, right? I give we give the example in the book about imagine that the FHA in the U.S. were building a tool that was trying to predict credit worthiness for mortgage applicants. And they used all of their own historical data to do so. Well, that predictive tool would structurally preclude African Americans from being creditworthy because of the FHA’s own admission of its redlining practices in the past. And so, it's a predictive mirror. Right. And so, what's really, really important here to say we need to still do the hard work of resolving these social issues and not try to just lay that technology's feet, because technology is, of course, reflecting back to us.Saviano
And unfortunately. I'm sorry. Unfortunately, there have been examples of not only a system from the private sector, but a system that have been launched from government to provide. To quote a few examples in the book. One in the state of Arkansas use algorithmic decision making to decide whether disabled people should receive home health care. And in the Netherlands, they use an AI system to discover fraud and benefit systems. And each of those examples were found that the data that was driving those algorithmic decisions were targeting the poor and disadvantaged. And in the case of the system in the Netherlands. And so, you know what we're learning about these systems, though, Abdi that it's hard to rectify that. It's hard to adjust the results of those systems. I think there's still a lot of work in the world to, number one, recognize when those biases exist and that the data that's fueling the system is actually producing a biased result. But everything we're learning about is that these are not easy fixes, are they?Aidid
Not at all. And the second part of that sort of reflection and amplification story that I was telling is that if you're going down the path of using technology to make predictions in these areas that involve these kinds of fraud questions or ugly histories, then you need to be attentive to it. So, the examples in the Netherlands and Arkansas are great actually illustrations of this story. So, in Arkansas was really interesting. It was eliminated and the Legal Aid Society of Arkansas brought this lawsuit against the state. They eliminated the bureaucrats who usually, who were previously making the decision as to whether someone was entitled to a home personal support worker. And instead, they used this algorithmic assessment tool and they found that the algorithm assessment tool was excluding people who, of course, should receive health care. And, you know, perhaps that was reflecting some negative history back to us. But it was also poorly technically designed according to the Legal Aid Society. So, for example, it asked questions like, do you have foot pain? And the idea is if someone answers yes, then they're more likely to receive in-home personal support. If they answer no, they're less likely to be because that is indicative of someone who doesn't have mobility issues for example. And so, they were forcing people into the yes or no answer. But if you're someone who's a paraplegic or a quadriplegic or an amputee, how do you answer that question? You might answer, no, I don't have pain. And that would, you know, trend you towards not receiving in-home health care. And so, part of the solution here is designing the technology intelligently, being critical about the questionnaires and the information that we're eliciting from the sources. And that that I think is much more possible in a world where we're all involved and accepting that our features technologies so that the people who design those tools are not just technologists trying to solve the narrow technical problem for the government, but for workers and the nurses who are privacy involved in health care administration. And so, coming to the table in many ways helps resolve some of the ethical quandaries that we find ourselves in.Saviano
And another reason perhaps. I love that explanation. And it's another reason why we're not ready to hand the keys to the machines just yet. And that, you know, these systems must be kind of mad about it. The systems must adapt to our norms and values and how important that is. And I think as we're learning still, the early days of generative AI. Okay, we are just about at the close of the show and we have a regular feature of Better Innovation. We end the show with three quick questions and quick answers. What do you say, guys? You guys up for it?Alarie
Let's try it, okay.Saviano
All right. Let's do the first question I'm going to pose to each of you. Then we'll split the last two. Okay, Ben, why don't you start us off? What's a book that has greatly impacted you?Alarie
I love this book called Superintelligence Paths Dangerous Strategies. That's a subtitle by a guy named Nick Bostrom, who's a philosopher at the University of Oxford. It's about what makes superintelligence possible and what should we do about it? What are the risks? And it's highly creative. It's super provocative. It'll get you thinking about the dangers of runaway artificial intelligence. Yeah. For me, it's like catnip. I loved it.