When I was asked to beta test its AI research bot, I informed a major legal research provider that it worse than sucked. It was dangerous. Not only did it hallucinate, which could be ascertained with minimal fact checking, but it conflated almost all the critical distinctions that make law work. It failed to distinguish between jurisdictions, both states and state and federal, as well as majority, concurrences and dissents. To AI, it was all the same, words about law, all of which melded into one amorphous mass of misbegotten but seemingly accurate “law.”
It will improve, I was told. Perhaps, but enough that it can be relied upon to tell us who wins and loses?
We are on the path to the legal singularity. Advances in technology, especially the improvement and widespread proliferation of artificial intelligence (AI), are driving us relentlessly down this path. By legal singularity, we mean a stable and complete legal order, capable of addressing and resolving practically all types of legal uncertainty in real time and on demand. Over the coming decades, the emergence of this legal singularity will fundamentally transform our existing legal systems and, with them, our societies.
In the minds of geeks, this is the glorious future that will both end the need for lawyers as well as resolve all disputes with the flick of an “enter” button. How one frames the question, of course, might produce directly contrary results, because AI has no judgment beyond responding to the query asked. Want to win? Ask the right question. Don’t like the result? Ask a different question.
But the point is that the people pushing this notion, and creating the algorithms necessary to enable AI to play judge, jury and executioner, aren’t lawyers but computer nerds. They wouldn’t know a legally correct answer if it bit them in the butt. At least they have a grudging grasp that there might be issues.
The stakes are high. Navigating the path to the legal singularity safely is necessary for humanity to flourish during the rest of the twenty-first century and beyond. For society to evolve and leverage these new technologies effectively, we will need to develop an ever deeper and more responsive legal infrastructure.
Is it necessary for “humanity to flourish”? Granted, lawyers and courts haven’t done a particularly great job of providing reliable opinions at the snap of a finger, or even after years of argument. To the lawyer, the old joke that the only correct answer to a complex legal question is “it depends” isn’t a joke, but the real response. To courts, the answer is often hard and contentious. Indeed, the most celebrated cases, whether for progress or otherwise, have reflected a paradigmatic shift in the law. Think Brown following Plessy, or Heller or Roe, and then Dobbs.
The good news is that the very technologies that are upending our existing practices will also enable us to construct the deeper and more responsive legal infrastructure that is sorely needed. The stability and resilience of the legal singularity will require more adaptability in our legal systems than they exhibit today. Fortunately, if we can get things right, the technology for a profoundly beneficial legal singularity will be in place just as we need it most.
There is much to be said about the institutional virtues of stare decisis, legal stability such that we can rely upon the state of the law in fixing our legal duties to one another, knowing that the law will not flip on its head every few years making our critical decisions and actions correct at the time but wrong soon afterward.
On the other hand, even assuming AI had the capacity to create a “legal singularity,” some homogenized version of law that applies to all, putting aside myriad jurisdictional distinctions and perhaps failing to take into account the nuances that upon which the law regularly depends, it would turn stability into an impenetrable wall through which no novel argument could survive.
Granted, there would be a great many people who would willingly forsake the potential for change in the law, and even the misguided ruling coughed up by Justice Google, for the benefit of ridding their world of lawyers and courts. The access to justice issue of affording a good lawyer would be solved, and no one would ever have to suffer reading a law review article again once we were confident AI read it for us.
And if the geeks are to be believed, the law will not be static, providing at least stability, but will evolve with society, as if society evolves in some clear and agreeable fashion.
The legal singularity will evolve and be able to absorb and accommodate changes to our social, economic, and technological contexts. If it is successful, it will not be dogmatic. Indeed, the nature of the legal singularity will be to provide quiet confidence that justice will prevail. Disputes will be resolved justly and in the best interests of society. Powerful actors will be held accountable to a greater extent than they are even in today’s most advanced legal systems. Weaker parties will have their positions bolstered.
The appeal of a legal singularity to the “weaker parties” is obvious, and, indeed, has some virtue.Then again, if you get an adverse ruling from the Court of AI, to whom would you appeal?
The shape of an eventual legal singularity can undoubtedly be influenced for the better by careful monitoring, concerted action, and thoughtfulness (or, for the worse, by the abuse of technology to oppress or suppress populations). It is in our collective interest to work to forge and secure the best-possible path to legal singularity.
While the general concern about AI taking over the world is a real one, the concern about a “legal singularity” raises many distinct problems. Good as it may be to know that “careful monitoring, concerted action, and thoughtfulness” will influence its creation (as opposed to the alternative), are you prepared to hand over your freedom and fortune to the best efforts of thoughtful coders?
I, for one, welcome our new AI masters.
Perhaps the geeks conflated “legal” and “mathematical” singularities?
In mathematics, a singularity is a point at which a given mathematical object is not defined, or a point where the mathematical object ceases to be well-behaved in some particular way …
For example, the reciprocal function 1/x has a singularity at x=0, where the value of the function is not defined, as involving a division by zero. [Wikipedia]
“. . .we will need to develop an ever deeper and more responsive legal infrastructure.”
Sounds more like a geek full employment scheme to me.
Some of the social media platforms run with AI moderators. Their systems aren’t capable of grasping devices like sarcasm, even where the world’s stupidest living human would recognize it. If we adopt such a system for judging legal issues, it will only add another element of randomness to a system already burdened by randomness. Some evidence will still go undiscovered, some witnesses will not show, some parties will still hire phony experts. AI judges that impose an ultimate result will further justice only in the sense that a roulette wheel or a lottery furthers justice.
I question whether these “legal innovators” truly believe this, or are just looking to make a bundle from the IPO and retire to bunker in Montana. I have seen AI in action and don’t it trust for uses beyond playing the oppent in video games.
Perhaps AI could conduct some truly routine hearings, e.g., arraignments. It is hard to see it doing well with much beyond that but who knows?
As someone who works in the field, allow me to make something abundantly clear: computer programmers don’t trust other computer programmers. We honestly don’t even trust ourselves.
I’ve been a programmer almost as long as our host has been a lawyer, and these folks should know better. Just ask a programmer what’s the best data structure to use, and any one of them worth a flip will answer, “It depends.” (We have that joke, too.)
But many programmers are cursed with the belief that anything outside their highly technical area of expertise is simple. Most of the pretty good ones have become accustomed to always being the smartest person in the room, and have rarely if ever been confronted with the extent of their ignorance, not just of non-programming subjects, but even of computer science at all. The smart ones are those who have had their noses rubbed in their ignorance.
The best thing to do with them is to pat them on the head, and to say, “I’ll be glad to take a look at it when you can tell me at least five major problems this program currently has.”
I’m pretty sure they mean “singularity” in the Ray Kurzweil sense that once progress reaches a certain point it takes off at a speed that no human can ever hope to achieve.
From the text quoted by the post: “Fortunately, if we can get things right,” — whoa, stop right there. This one sentence fragment is doing a mountain of heavy lifting. How many lives are going to be destroyed by “move fast and break things” programmers who think it’ll all be fixed in the next iteration of the software? How many iterations before the “singularity”? And, while we’re iterating, what would regression testing look like for a Legal AI, anyway? Meaning, how would we know that we haven’t broken anything with the next release?
Madness.
“Disputes will be resolved justly and in the best interests of society.” – The “best interests of society”? How old are these people? What does that even mean? We’ve been arguing over this for thousands of years and haven’t settled on one answer yet. Are we going to get a good answer with enough VC money? Unlikely.
There is a danger of the move fast and break things mentality influencing law reform.
When I’ve pointed to obvious consequences in the local attempt to create offences about coercive control, I’m often met with the answer, “If there are unintended consequences, the legislation will be amended.” This complacent faith in iteration expresses no concern for the lives it will ruin, and no understanding of the political and legislative process
Pingback: Twenty-First Century Delphic Oracle - Fresno Criminal Lawyer