Large Language Models (LLMs) were the hot topic of the LegalWeek conference last week in New York. Even if you didn’t attend LegalWeek, you have likely heard of LLMs, or ChatGPT, or OpenAI, or generative AI.
To level set:
Large Language Model developed by... OpenAI, which is a company developing various forms of...
I heard these 4 terms used pretty interchangeably throughout the week, although I get the sense that people are settling on LLM as the generic reference to the technology. This is particularly confusing in legal, as an LLM is a Master of Law degree, but I am not convinced that “LLM” is the term we will ultimately land on, so stay tuned...
It seemed that nearly every vendor at LegalWeek was incorporating LLMs into their technology. The topic came up in nearly every conversation I had with both law firms and tech vendors.
At a separate KM conference I attended on Thursday, the speakers and attendees managed to ramp up the LLM hype even further. By the end of the keynote presentation attendees were visibly whipped into a frenzy of excitement, possibility, anxiety, and a palpable fear of either drowning in the new tech or missing out entirely.
At one point the conversation facilitator asked for a show of hands.
“How many of you have already or are currently in the process of buying a product to incorporate LLMs into your firm?”
About a third of the hands go up.
“How many of you are actively training your own Large Language Models?”
A different quarter of the hands go up.
“A lot of you didn’t raise your hands, so what are you doing!?”
First, I find it hard to believe that a quarter of the firms represented in that room were actively training their own LLMs. I don’t know if they misunderstood the question, or if they were over-zealous and wanted to be seen as proactive in front of their peers, but that response seems highly unlikely to me. And if they did answer accurately, I suspect there are a lot of firms poorly training their own LLMs. Secondly, I am pretty sure the last question was a rhetorical joke from the facilitator, but it drove home a point that the keynote speaker made earlier: “This is moving so fast that if you’re not on top of it now you will be left behind.”
Many of our clients have asked for my thoughts on LLMs over the last several weeks and I have openly shared my thinking. By nature, I am a skeptic. I abhor magical thinking outside of fiction. And I always seek to understand truth and reality to the best of my ability. My thinking about LLMs has not substantially changed after this week, but my fear of hyperbolic peddlers and the ways they will seek to take advantage of my clients has grown substantially.
A few things I heard last week and my perspective:
1. “LLM Technology is incredible, amazing, revolutionary, and will fundamentally change our relationship to computers forever!”
I think this is quite possibly true. There is no doubt that this technology is remarkable and is already doing incredible things. But it is important to remember, that even though ChatGPT has set user-adoption records (”Fastest technology adoption in history!”, I heard at least 3 times last week.) and nearly every vendor at LegalWeek announced their integration of LLMs, we are still very much at the beginning of this process. Yes, it’s moving quickly. Yes, it’s a lot to take in and and yes, we need to keep up with it. But we’ve only just begun to understand some of the ways these tools can be used. There is a long way to go still and I believe that the primary way we will interact with generative AI in the future has almost certainly not been implemented yet.
This technology potentially opens up computing to a much wider range of users in the same way that the Graphical User Interface (GUI/Windows) did in the 1990s. Anyone who can speak or type will be able to use a computer to carry out relatively complex tasks. (As soon as Amazon rolls out LLM integration to Alexa, I should finally have my Star Trek computer.) For those of us who are comfortable in a GUI environment, this tech means we will never again look at a blank screen wondering what we should write, or what our presentation should look like. The old adage that “writing is rewriting” has never been more true. The first draft of your novel may be entirely generated via AI, but the readability of that novel will still depend on how well you clean up and adjust the “statistically probable” characters and plot the computer handed to you. More relevant to many of us, you’ll never again spend 3 hours scouring the web to find just the right picture to make a specific point in your presentation, instead you’ll just ask the computer to generate an image of what you’re looking for. (See the “suited surfer at sunset” above.)
Creativity just got really weird, but also super-powered, and the most creative people are the ones who will benefit most from this technology. From everyone else, we are about to see a whole lot of really badly conceived but well executed art, fiction, blog posts, articles, press releases, spam emails, etc.
2. “This technology is going to be everywhere.”
This is undoubtedly true. We are already seeing it implemented in a number of the tools Sente uses on a regular basis. Coda, Canva, and Bubble have all held webinars in recent weeks to show off what they’ve done with LLMs and how it will revolutionize using their tools. At LegalWeek, we started to get a taste of what’s coming in the legal market and I’m sure we will have a slew of vendor webinars over the next few months that will show incredible new features built into technology you have already licensed.
We know that a few legal tech vendors have already released some of their LLM tools. If you have those platforms and you can get access to the new features without paying, I highly recommend you try them out. If they insist on a substantial increase in licensing fees to get access, tell them to come back to you in a month. I’d bet my left arm, their price will have come down by then.
We know that Microsoft has preferential access to the OpenAI technology as a large investor and they are incorporating it into MS Office, the Power Suite and (of course) Bing.
We should start to see some form of generative AI in Word in the very near future. Someday soon you will be able to ask Bing to create a form that asks a user a dozen questions, generates a document incorporating those answers, sends the document to 3 different parties for signatures and drops the resulting, fully executed contract into your DMS, all without opening MS Forms, PowerAutomate, DocuSign, or iManage. Realistically, such a tool will probably create all of the correct components and tie them together, but you’ll still need to open each of them and do a lot of “tweaking” to get exactly what you want. Still, this LLM technology is going to help create the “first draft” of whatever you’re working on.
3. “This is moving so fast that if you’re not on top of it now you will be left behind.” (as mentioned above)
This depends on a couple of factors:
What does your company do? How do you define “on top of it”?
If you are a technology company who has spent years (or decades) building out specialized solutions that could potentially be rebuilt or replaced in a matter of hours with this new technology, then yes, whatever “on top of it” means, you need to be “on top of it” immediately or else you will be facing strong new competition in the coming weeks. That said, just as accomplished artists and authors are the people most likely to create better art or write better stories using generative AI, the companies who have spent years or decades building up specialized knowledge are the companies that are most likely to put this new tech to good use too.
Law Firm KM/Innovation
This is where the surfing metaphor comes in. I’m not an avid surfer, but I know enough to know that being in front of the wave means you’re just as likely to miss the ride as being behind the wave. Metaphors and analogies only get you so far, because none of us are really ahead of this wave, but my point is that “doing something” just for the sake of saying you’re “doing something” isn’t going to help you ride the LLM wave.
If you’re reading this article, you’re already scanning the horizon and have seen the wave coming toward you. As it gets bigger and bigger, you’re starting to question whether you want to ride this one or let it pass and catch the next one. It’s time to catch the big one.
You need to turn now and start paddling toward shore.
Be aware of LLMs and become knowledgeable about what they do and don’t do well Keep an eye on the market and the new tools that are popping up daily Play with the tools you can get access to for little or no money
You do not want to stand up on the board yet, and you don’t want to start paddling toward the wave at this point.
You don’t need to train large language models on your own documents or data You don’t need to purchase new software for hundreds of thousands of dollars You don’t need to unduly worry about falling behind everyone else
This technology is coming to the tools we all already use. You will probably not gain any competitive advantage by being an early adopter of any particular tool. However, you will benefit from exploring the technology, understanding how your firm is likely to use it, and talking to your current vendors about how they intend to implement it.
You will be bombarded with information about this technology over the next several months. Making sense of it will be difficult and a lot of people will offer to make sense of it for you. Some of them will be selling you something, their goals should be obvious and their claims should be taken with the requisite grain of salt. Less obvious are the people that appear to be disinterested 3rd parties, maybe giving you good information, but also stoking your fear of missing out (FOMO) and driving you to act rashly, at a time when the technology will never be WORSE or MORE EXPENSIVE than it is now. Make sure your FOMO doesn’t drive your decision making and ensure that you know how anyone stoking your FOMO is getting paid.
I don’t know where LLM technology is going, but it is undeniably cool and exciting and fun. It will also likely be a core component of the work we do on a daily basis in the near future. However, if I had to bet, LLMs will be embedded in the tools we already use and in all of the new technologies that come along in the future. They will make certain aspects of our work easier and more enjoyable, while making other aspects more difficult and frustrating. So...you know...LLMs are like any other technology.
Assuming LLMs are as truly transformational as they seem, then you don’t need to be an early adopter of any particular tool. The tools are going to come at you fast and they’re going to be ubiquitous. You just need to be in position, already moving the right direction, and prepared to ride the wave as it gets to you.