Given the emergence of GPT-4, the launch of first AI legal assistant CoCounsel, and the acquisition of Casetext by Thomson Reuters, 2023 will without a doubt stand out as one of this team’s most momentous. The kind of truly intelligent AI that experts have said was a half-century away now seems likely within the next half-decade. Until this year we didn’t think we’d get to see developments like these at all in our lifetime. Now? Almost nothing would surprise us in the months and years to come.
To make AI’s new capabilities a bit more tangible, consider this: CoCounsel has, since its launch on March 1, 2023, read (with the skill of a postgraduate) 179,832,832,748 words. That’s the entire Harry Potter series nearly 166,000 times over. The average college-educated person reads 250 words/minute, so it would take 5,167 lawyers (more than Baker McKenzie’s entire staff) reading for 8 hours straight for every one of the ~290 days between launch and now, to do the same.
2023 in the rearview
These recent leaps forward have been almost more than even those of us in the tech and legal tech industries could process. So making sure our customers could make the most of today’s AI quickly became our priority. We’ve spent, and will spend, a lot of time helping people understand what an LLM is, what distinguishes the LLM GPT-4 from all those that came before, and what separates using an LLM-powered product, (like CoCounsel) from ChatGPT.
This work is of course ongoing, and the more we learn about the opportunities and risks the newest generation of this technology brings, the more clear it becomes that we all need to continue promoting AI literacy—our own and our customers’—if we’re to ensure the greatest number of people realize the greatest possible gain, while holding the risks at bay.
And while this might seem like quibbling, “generative AI” is the wrong name for the revolution that’s underway, because it focuses us on the wrong aspects of the technology. GPT-4’s ability to generate text is impressive—and no doubt entertaining—but we have a long way to go before it can write legally useful text. What is astounding, though, is its ability to read, annotate, analyze, and seem to reason about text.
In the legal space, specifically, it’s important to understand that, as astounding as 2023’s advances have been, some of the most powerful applications of this technology to the profession are happening behind the scenes. For instance, we’ve had the privilege of partnering with prominent firms to validate and prototype for CoCounsel use cases in high-volume due diligence and e-discovery—with some truly amazing outcomes that we’ll be able to share more about in 2024. Applying GPT-4’s reading and analysis powers to documents, and particularly tremendous volumes of documents, is hands-down the most significant development that’s ever been seen in legal tech. These capabilities will have a tremendous impact on, for instance, how long litigation takes, and its level of quality. And these are still early days.
In fact, the stories about AI that so far have gotten the most attention don’t accurately represent what’s really going on with the technology itself and the legal industry’s embrace of it. Everyone knows who Steven Schwartz is (even though his failures had nothing to do with technology), and there’s no lack of headlines about how courts are restricting use of AI rather than positively focusing on disclosure. Meanwhile, other courts are quietly exploring how they can use GPT-4 to streamline certain aspects of their workflow, intelligently evaluating AI’s advantages, rather than handing down reactionary rules. And the California Innocence Project is confident CoCounsel helped them cut their case backlog in half—meaning less time in prison for the wrongfully incarcerated eventually freed by their work.
What 2024 holds for AI for law
Ubiquity
By this time next year, AI should seem downright boring, if those of us working on it do things right. We’ll see a flip from “Why are you using AI?” to “Why aren’t you?” If you’re a lawyer in the midst of discovery and get slammed with two million documents, not using AI to process them will be the much bigger risk.
We’ll see better research briefs, better evidence uncovered, more efficient deals made more inexpensively. Which means we’re also likely to start seeing cookie-cutter briefs from people over-relying on AI tools. Like any tool, AI adds value only if used well and correctly, and of course not everyone will do that.
Multi-modal leaps
Now that AI can interpret photos and audio, its applications for law are staggering. For instance, it could be directed to pick out all the safety violations in a photograph of a work site, and then, coupled with the right content—like in Thomson Reuters AI-Assisted Research on Westlaw Precision—identify exactly what laws and regulations are being broken. Add CoCounsel Core capabilities, and you’re well on your way to efficient enforcement. These are just some of the ways the regulatory agencies named in President Biden’s recent Executive Order on AI could employ AI, in their mandate not just to curb its misuse but almost more important to skillfully employ it to better fulfill their existing directives.
And beyond …
The uncertainty of scale
Right now literally billions of dollars are being funneled into seeing what will happen if the parameter size and training data for LLMs is increased by an order of magnitude. It’s quite possible their capability will plateau. It’s quite possible they’ll experience a tremendous jump in capability. Could we see an AI that truly can reason? Given what’s happened in 2023, we’re not ruling anything out. And honestly the question is not whether, but when.
Disruption of the billable hour
Every time the fears that AI will disrupt the practice and business of law bubble up, we’re reminded of one of our Chief Innovation Officer Pablo Arredondo’s favorite books: former Yale Law School Dean Anthony T. Kronman’s The Lost Lawyer. In it, Kronman posits that as oath-taking professionals and officers of the court, we should welcome disruption. We’re in a unique and privileged position and should be unafraid to step back when developments happen in the profession and the world, and allow the possibility that shaking things up could be good for lawyers, the law, and, most important, citizens.
Certainly the billable hour is fundamental to how many firms operate, so it’s understandable that they’d be spooked by anticipated changes to that structure. But the legal profession has weathered a lot of change—and weathered it well. Five years from now timesheets are going to look different. That doesn’t mean, however, that the bottom line will necessarily be less. Just different. And we’d argue that’s a good thing.
Ultimately in the United States and similarly globally, we’re charged with upholding Rule 1 of the Federal Rules of Civil Procedure: “These rules govern the procedure in all civil actions and proceedings in the United States district courts … They should be construed, administered, and employed by the court and the parties to secure the just, speedy, and inexpensive determination of every action and proceeding.” Because the law is so language intensive, it’s uniquely positioned to benefit more than most professions from responsible use of this new technology. A greater number of just outcomes will be made possible by delivering service faster and for less money, which AI is already proving it can do.
Lawyers aren’t going anywhere
Though so much of what we in the industry are talking and thinking about these days still seems surreal, and anything seems possible, AI won’t be replacing lawyers anytime soon, if ever. The lawyer’s role might continue to change, but that’s for the better: more time and brain power for more valuable, strategic work machines cannot do, once we can delegate a lot of the most time-consuming and rote tasks to AI.
In fact, there’s a danger of losing our ability to do that higher-level work if we delegate too much of the writing crucial to the practice of law, because wrestling with a blank page is vital to the work of thinking through a problem.
Law is an evolving organism, and it can only work with what we feed it, like novel ways of articulating ideas—which is exactly what LLMs don’t and can’t do. It’s crucial that we keep mutation and creative sparks alive, the processes that lead to persuasive advocacy before a court or nuanced judicial opinions. These are the problems lawyers can and should be working on, and the justice system will only be the better for it.
The post AI’s biggest year (yet) appeared first on Casetext.