The term ‘revolution’ continues to be thrown around when talking about AI, but what does this revolution mean for the world of family law, and do we need to be concerned? Returning to the show today to help us make sense of AI in law is Managing Shareholder at STG Divorce Law, Raiford Palmer.
In this episode, Raif offers us a brief description of AI, its current state in the world, and how it is impacting various industries, as well as how he and his team are using it at STG. We then get a sense of how AI is already changing the family law landscape and why there will always be a need for a human element, and we also look at some of the potential dangers of AI in law and the limitations that exist within current AI models.
As this intriguing conversation comes to a close, you’ll hear about the ethics of AI, why humans will always rely on tools, and why AI is nowhere close to being a flash-in-the-pan phenomenon. Love it or hate it, AI is here to stay!
Key Points From This Episode:
What is Divorce at Altitude?
Ryan Kalamaya and Amy Goscha provide tips and recommendations on issues related to divorce, separation, and co-parenting in Colorado. Ryan and Amy are the founding partners of an innovative and ambitious law firm, Kalamaya | Goscha, that pushes the boundaries to discover new frontiers in family law, personal injuries, and criminal defense in Colorado.
To subscribe to Divorce at Altitude, click here and select your favorite podcast player. To subscribe to Kalamaya | Goscha's YouTube channel where many of the episodes will be posted as videos, click here. If you have additional questions or would like to speak to one of our attorneys, give us a call at 970-429-5784 or email us at firstname.lastname@example.org.
DISCLAIMER: THE COMMENTARY AND OPINIONS ON THIS PODCAST IS FOR ENTERTAINMENT AND INFORMATIONAL PURPOSES AND NOT FOR THE PURPOSE OF PROVIDING LEGAL ADVICE. CONTACT AN ATTORNEY IN YOUR STATE OR AREA TO OBTAIN LEGAL ADVICE ON ANY OF THESE ISSUES.
Ryan Kalamaya (3s):
Hey everyone. I'm Ryan Kalamaya
Amy Goscha (6s):
And. I'm Amy Goscha.
Ryan Kalamaya (8s):
Welcome to the Divorce at Altitude A Podcast on Colorado. Family Law.
Amy Goscha (13s):
Divorce is not easy. It really sucks. Trust me I. know Besides. being an experienced divorce attorney, I'm also a Divorce client.
Ryan Kalamaya (20s):
Whether, you are someone considering divorce or a fellow family law attorney. listen in for weekly tips and insight into topics related to Divorce co parenting and Separation in Colorado. Welcome back to another episode of Divorce Altitude. I am Ryan Kalamaya. This week we're gonna be talking about AI and there has been talk about an AI Revolution and we're gonna talk about how AI is going to impact divorces. And we're joined by a former guest, Raiford Palmer from STG Divorce Law in Chicago, Raiford.
Ryan Kalamaya (1m 2s):
How's it going?
Raiford Palmer (1m 2s):
Great, thanks Ryan. I appreciate it. Thanks for having me on the show.
Ryan Kalamaya (1m 6s):
Yeah, welcome back. We'll talk about when people hear you talk about AI, I think they'll realize why we're having you come back on. Previously we had talked about your book, but AI, it's kind like Terminator. We've been hearing about AI and there have been movies like The Terminator about AI, but it feels like it's happening right now and so Can you for Listeners that may not know what chat G p t or what AI is, Can, you give our our Listeners maybe like a baseline understanding of what's happening right now as we're recording it and how it's impacting every industry, at least when we're recording this in, in early 2023.
Raiford Palmer (1m 49s):
Sure. Ryan. So a couple big things have happened recently and research in artificial intelligence and the recent manifestations of that, which people will say are technically not artificial intelligence but they really look like it to the average person. There are large language models like chat, G P T and image models such as doll, E D A L L E. Both projects have open AI started famously by Elon Musk. Now it's a private company. Microsoft recently spent a whole huge sum of money to purchase an interest in OpenAI and then acquired access to chat e p t to incorporate into Bing.
Raiford Palmer (2m 31s):
So the short story is some brilliant engineers have figured out how to digest a large volume of data. So in the case of chat e p t to take a huge pool of text information, so imagine a vast sea of books webpages, tremendous amount of text information and digest it into a computer model and then basically teach the model how to think in plain English by referencing all of that text information. So if you imagine it all that information stuffed into a mixmaster and chopped up into pieces with a model, then looking at that in a way that we might think is almost intelligent and then when you ask it questions in English, it can respond in a way that seems intelligent and in plain English actually very good English responses.
Raiford Palmer (3m 25s):
This stuff has been researched now for years and everyday people are now seeing the fruits of that research and they're using chat E P T and other similar tools to do things nobody really thought possible before with these artificial intelligence models. And similarly, doll E is an image model that allows you to create a, a painting in the style of the impressionists or line art drawing or charcoal of any image you can think of. And the artificial intelligence engine in that case, the model is essentially digested a tremendous amount of artwork and then it's been, if you will, taught rules about it so that it understands what impressionism is, it understands what a charcoal is and can then render images based on human English language prompts.
Raiford Palmer (4m 17s):
So it's pretty incredible when you play with these things to see what they can do. And I've been fascinated by technology ever since I was a kid and follow this stuff all the time, both personal interest and for the business. So it's been a special interest of mine.
Ryan Kalamaya (4m 33s):
Likewise, I mean I think Listeners to Divorce at Altitude know that I like to geek out on the text off. I am certainly admittedly a victim of the shiny object syndrome. I think you know, Raif, you are too. So okay, we can talk about AI and how cool it is and and for Listeners, if you have tried it out, Chachi p t or any sort of version of ai And I would submit that anyone listening to this has experienced AI in some capacity. But how, just as an example, we came up with the outline and the title to this podcast episode with AI.
Ryan Kalamaya (5m 15s):
It's spit out, we put in AI and Divorce and it spit out the title that people have clicked on, presumably if they're listening to this, it was intriguing and it knew that this is kind of a sexy or or an interesting title and then it came up with a an outline which Listeners can't hear, but what are the examples? You're in Chicago, you run a Family Law law firm, STG Divorce Law, how are you really using it right now? And then we'll talk about where we see things going and maybe Listeners will have their eyes open to where at least this could be going when it comes to, to law lawyering and specifically with Divorce.
Raiford Palmer (6m 2s):
Sure. So we'll get to why these models aren't quite ready for full throttle application in the legal field, but there's a lot of usefulness right now with a law practice, really any business. And so right now I'm using chat PT and jasper.ai, which is sort of a, it's a version of chat PT that's tuned for writing, really for writing things like marketing pieces. And I've written extensive blog post about parenting using the large language model using Jasper and Chat G P T. That article might have taken me three hours in the past and it took about 20 minutes.
Raiford Palmer (6m 45s):
The reasoning is it came up with a very good outline that I edited for the article similar to the way that we generated an outline for this show. And then I fleshed it out, made some changes, rearranged it the way I liked and then I asked it to flesh out those outline items and it gave me nice bullet points underneath those And I added some more of my own and, and then used that and asked it to elaborate on the outline, which it did, and created basically each paragraph of the blog post with a pretty good introduction in ending paragraph conclusion paragraph And I did some, some editing, which the editing took most of the time.
Raiford Palmer (7m 26s):
And then I had a nice like 800 word blog post that I then published on our firm website. So, and it's, it is valuable content for our readers to know about, you know, things to look out for with parenting with children and it was a tremendous asset for me to speed up the process. And so that's an example. Another example is writing a description for hiring a paralegal. so we wanted to post a, a job post for paralegal and for an associate attorney and chat G P T came up with a very good post that needed some editing to customizing it for our PURPOSES, but the baseline material was all there. And that's some of the power that you can unleash with these tools right now because the tools have digested so much information probably from articles and business books and blog posts that you've got access to tremendous power under the hood in these tools.
Ryan Kalamaya (8m 23s):
Yeah, a couple things that come to mind for me, Raif, well first of all we're using it in a similar capacity where we send out to various clients emails explaining the Divorce process and those emails, they've been optimized so they've been streamlined. There's a a feature within the program that we use where it summarizes something into five bullet points and so it makes things a little bit clearer but similar to you, we we're using it within blog posts and and other marketing, but I don't think we are that far away from, and, and you hear these stories for example about chat G B T passing the bar exam and if you think about it, I mean there are so many different applications.
Ryan Kalamaya (9m 10s):
And, I there I was recently solicited by a vendor who has a program where they put in bank information and then the program looks for unusual spending in essence does a forensic accounting and it's specifically for financial fraud or hey Divorce lawyers, we deal with that. And you go through that whereas you file a motion to modify parenting or some other motion. I think we are gonna be very quickly talking about the AI bot or the AI program, drafting that motion, pulling the law and then analyzing who the judges and knowing that judges preferences in terms of do they like the petitioner versus respondent as a referral to the parties versus husband and wife, mother and father.
Ryan Kalamaya (10m 3s):
There's gonna be data, there's so much data out there, And, I, think those are just a couple examples of how AI is gonna change the landscape when it comes to Divorce and how it's being done.
Raiford Palmer (10m 16s):
Yeah, one of the early entries in the field is called spell book And. I did a demo recently for a week. Spell book runs alongside in Word as a plugin and it uses ai, I think they're using the G P T or the open AI api. So they're accessing open AI's model and they've added a huge database of contracts known as Edgar, which for computer people I've described as like GitHub for contracts. So the government requires certain contracts are posted in our public, so it's a big repository of contracts written by some large law firms. So they incorporated those contracts into their, into a private instance or sort of a, a large language model.
Raiford Palmer (11m 4s):
So it's digested those contracts and it can intelligently look at contract language and suggest missing terms, show you where there are inconsistencies in the contract, show you things that you should be concerned about if you're looking at it from one party standpoint versus another. And you can even use it to draft language. So because of the sa, that power of that language model, as long as you feed it the right data, it theoretically can then use that data in a way that we think is intelligent to look for flaws and also suggest fresh content. So the power of that is, is you're right, is already here for people doing commercial contracts and the company was talking to me about creating a a s an instance for our firm where we would have it digest marital settlement agreements, parenting agreements so we could intelligently look at a document sent by another attorney to show us things that are missing, suggest definitions of terms that aren't defined and help us draft language.
Raiford Palmer (12m 4s):
So you're absolutely right, it's, I think it's a lot closer than most lawyers. Most lawyers think certainly and most lawyers are comfortable with for sure. so it's important that we stay ahead of these tools and that's Why you, And I. No, not only do we geek out, but we think it's good for our clients and for our business to be aware of this stuff.
Ryan Kalamaya (12m 24s):
This episode is brought to you by our law firm, Kalamaya Goscha Amy And I describe our law firm as an innovative and ambitious trial team that pushes the boundaries to discover new frontiers in family law, personal injuries and criminal defense in Colorado. We currently have offices in Aspen, Glenwood Springs, Edwards, Denver and Boulder. If you wanna find out more, visit our website Kalamaya dot law now back to the show L Yeah, my firm, you know we use text expander, which is kind of an expansion we'll put in various codes and, and the way that we have it is a parenting plan. For example, we have different language, for example, when alcohol is involved in parenting and there is different testing protocols.
Ryan Kalamaya (13m 12s):
So, and it depends on the severity of the problem. So if it's like you shouldn't drink with the kids, that's one thing is it, should you prohibit both parties from being drunk with the kids? Well what if you get into monitoring? so we have all these different options in our system and I'm sure you do too, Raif where you're selling the house, it depends on is your client keeping the house while it's being listed? And that might be a little bit more looser in terms of the language versus you represent in our kind of hypothetical Divorce clients, Erica, Melanie Wolf, if Melanie's living in the house and Eric's like, I want this house to be sold, there's gonna be a different kind of language that's gonna be pretty soon where we're gonna figure out what's the best solution.
Ryan Kalamaya (13m 55s):
And I don't think it's that far away where Eric and Melanie will be able to kind of answer a few questions or the AI program's gonna know who Eric and Melanie are and it's gonna draft their optimal settlement agreement. I I also, you know right now you And I, we have paralegals that work with clients to gather up their financial information, put together financial disclosures. We're gonna soon be able to plug in someone's username and password to their Wells fargo.com account and their chase credit cards and it's gonna run a spending analysis and create all that for the client to begin with. And I think people when they don't either lawyers or Eric Melanie, you know when they're going through a Divorce and they don't realize that there's a lot more efficient ways of doing business.
Ryan Kalamaya (14m 42s):
I think that that is something that we really want to kind of make sure and highlight for people to think about as they go through the process. Cuz it's changing every day.
Raiford Palmer (14m 51s):
Yeah, the, and we've got, you know, there are tools like Family, Law software that already allow us to model finances in terms of showing a balance sheet and divide division of assets much more easily than the old days using just a bear spreadsheet, right? And also modeling support, you know, child support alimony and the point you're saying is well taken and that is it's not going to be long when all you'll need to do is is feed the data to something like Family Law software with an AI component and it will just pop in all the relevant data where it needs to go. And also then based on some inputs, give you maybe three settlement scenarios, you know, whatever you wanna look at what 50 50, 55, 45, whatever it is favoring one party or another give us the most tax advantaged thing for party A or party B and it'll, it's going to eliminate a tremendous amount of grunt work.
Raiford Palmer (15m 45s):
I agree with you in this process both in the, the document analysis side and the document generation side. So doc review, doc analysis and then the document creation side, still leaving room for the lawyering but a lot of raw processing and grunt work squeezed out of the process.
Ryan Kalamaya (16m 5s):
And I think that there is always going to be a role for you and me to talk with Eric and Melanie about their fears, about their goals and really connect with them on a human experience to talk about their kids. I just, I have a hard time really conceptualizing as easy as it is to come up with an optimal settlement scenario where Eric gets the forks, Melanie gets the spoons and Melanie gets the stock portfolio that is balanced in an optimal manner. You can run that scenario for sure in a computer system. It's not going to really tell them about the emotional connection that Melanie may have to the house for example.
Ryan Kalamaya (16m 49s):
And if we talk about potential dangers associated with using ai, if you don't know you were talking about editing a blog post, there's certainly going to be role, a role for Attorneys to edit and to customize and to really understand how the tools that they have at their disposal apply to Eric and Melanie's situation. So what other potential dangers do you see that are associated with using AI in, in a Divorce or in any sort of legal context right now?
Raiford Palmer (17m 25s):
So at present the, there are substantial limitations with these models and it's important to know what they are. So for example, jet G P T is an iteration that they're calling I think 3.5 and they're saying 4.0 is coming up and the near future, the data set used for 3.5 goes somewhere through 2021. What's interesting is I've queried it to ask how Illinois child support is calculated and this gives an example of where, what its limitations are, it will confidently tell you the correct answer, which is the income shares model. And then if you ask it again, it'll confidently tell you the old system that existed before 2017 and tell you that a straight percentage of the payer's net income is the model or is the system.
Raiford Palmer (18m 14s):
And it'll tell you both, both things with equal confidence. That's just because this model hasn't been taught how to distinguish what is good law and also current data aren't in there like current cases, current legal information like current statutes. So, but it's only a matter of time like you point out where somebody spends the time to train the model to approach it like an attorney and understand what controlling law is, what is now, what is good law, what is now outdated, what cases are overruled and what is controlling. So it's only a matter of time before the model understands that and can distinguish that and it's been trained on the cases and the statutes and trained on how to understand all that.
Raiford Palmer (19m 1s):
Once that happens, I think you'll see that error rate drop dramatically. I'm sure the legal research companies like Westlaw, Lexus are working on incorporating AI model into their systems. They've gotta be scrambling to do that at this point.
Ryan Kalamaya (19m 15s):
So for sure I mean I got a message case text is our research software and it already has a component of AI it, if we upload for example a brief or a motion, it will then read it, it will tell us, hey did you think about using this case or that case? It seems like you're filing this kind of motion and it is absolutely on the radar for Lexus and, and Westlaw. I think Westlaw, one of 'em has some sort of AI program that they are already using. you know, for me I think there's some ethical issues, you know, that haven't really been ironed out.
Ryan Kalamaya (19m 55s):
We hear about the, just as an example, some of the, the autonomous driving, they haven't really figured out, for example, if a car is driving and a baby crosses the road and there's a elderly woman on the side of the road, is the car programmed to run over the old lady because you know, society values a baby more. Those sorts of ethical conundrums I don't think have really been processed, but I mean they've already come out with, you know, various programs that could listen to testimony and object or do those sorts of things. you know, I think reading a jury or reading a judge the the facial expressions, it's not ready for prime time but you know, Raif, you And I.
Ryan Kalamaya (20m 40s):
When we schedule things, we use AI to, to kind of schedule to show our availability. And I think that the legal administration is going on. We're certainly seeing, you know, some efficiencies. If someone is paying for that, a client is paying for that, you know, it's cuz it's baked into the lawyer's overhead. I think that that's certainly something that is, that needs to be considered. And then just some of the analysis and the visual you mentioned at the beginning of the dolly we used custody exchange for parenting schedules and really having that visual component to a calendar. I was talking to a friend of mine who's going through a Divorce and I'm too close to it so I'm not involved, but he was like, my lawyer doesn't even know what that is.
Ryan Kalamaya (21m 24s):
I listened to your podcast on that. And I think it's those tech tools, whether it be AI or in some of their capacity that I think is really the, the kind of main point of this episode.
Raiford Palmer (21m 36s):
Yeah, and one other word of caution about this, like recently lawyers have recently been playing with Jet E P T and asking it to give them like three cases on a particular topic and the app confidently gives them three cases that are completely fabricated, the names and the holdings and everything, but they look great, they look legit. Now when you go look 'em up, you find out they don't exist. But again, that's just because chap G P T hasn't been trained in looking at those cases. It hasn't digested that information and doesn't know how to think about that information, if you will. And that's only a matter of time. So you're, you're right, it's going to impact the legal field in numerous ways and the only thing we can do is inform the Listeners about it so they understand and they're asking the right questions and understand the power and limitations of it.
Raiford Palmer (22m 29s):
And, and hopefully we can stay ahead of this and understand how to best use it for the benefit of our clients and to handle, you know, run our businesses better.
Ryan Kalamaya (22m 37s):
There's certainly a risk in plagiarism because the way that that the model, it basically scrapes other people's information in gathering up its intelligence. But the one analogy that I thought was very helpful for me, I saw I think on LinkedIn and it was a clip of Steve Jobs and Steve Jobs explained the use of the computer and that if you looked at the efficiency of, you know, a human versus a bird in terms of, of movement, the bird was far more efficient in terms of getting from point to A to point B because it could fly. But when you put a human being on a bicycle, then the bicycle enabled the human to be way more efficient than the bird in its human beings is capabilities of using other tools.
Ryan Kalamaya (23m 27s):
That really is what we do as a species incredibly well. So, you know, people that just want to say, oh, ai, I'm not gonna get into that and, and the next thing I wanna talk about is whether or not we're really kind of talking about a bubble and there's some financial issues involved with AI because a lot of money is going into this industry, but if people just reject it and just I'm a Luddite And, I don't really want to embrace it, those people are the same people that just said, zoom, I'm not gonna enable zoom during the pandemic. And you know, they're either gonna get pushed out, they're gonna retire or it's incumbent on people. It's arguably unethical not to familiarize yourself with these things.
Ryan Kalamaya (24m 10s):
But the one thing I, you know, before we started recording this Raif, there was an article on the Wall Street Journal about whether or not this is a bubble for me. There's gonna be a lot of money going into this. These industries and these people that are involved in AI are gonna go through divorces and we're gonna have to know how are their businesses operating and, and just generally be aware of, you know, the, that industry in figuring out Valuations and really what these people are doing at work. Any thoughts on that particular issue?
Raiford Palmer (24m 41s):
I certainly think it's good for us to understand these tools, both for our clients and understanding how these businesses operate for the benefit of our law firm, staying ahead of the technology curve and embracing them to solve problems for clients. Like you said, a famous memory of mine was as a law clerk a long time ago, I had a boss who said I could buy anything for the law firm as long as didn't have a screen on it. He was not a fan of computers, that's how old I am. But computers were at the time, PCs were out there and he just wanted the staff to have typewriters, believe it or not, el electronic typewriters, but still typewriters.
Raiford Palmer (25m 22s):
So lawyers are, are notorious for being slow to embrace change period and certainly technological change. And because of our regulatory moat that protects our industry, you have to be a licensed attorney to, you know, be in, be in the business, at least in most states. That is insulated lawyers to some extent from needing to adopt tech tools. But so many lawyers have adopted the tools and have been successful that I think eventually it sweeps everyone along. Or like you said, they retire or go out of business. So I think those who fail to recognize the power of these tools and fail to figure out how to utilize them will be at a dramatic competitive disadvantage.
Raiford Palmer (26m 9s):
Eventually, as clients see the tools used in their own business and understand the power of them and the economies And, we will demand that their lawyers use the tools too. They'll say, well of course you're gonna use a, a large language model to help write the documents. In my case, why would you draft them the old fashioned way? And that's And. we saw that a generation ago, 20 years ago, clients started saying to us, well, I know you're just pulling a form out of word. You know, you've got a, you've got probably 50 marital settlement agreements there. I know it isn't taking you eight hours to draft the settlement agreement. Maybe it's taking you two. Well this is just that writ large, like another leap forward where I know that settlement agreement doesn't take you two hours.
Raiford Palmer (26m 54s):
It probably took you 10 minutes. And, and that's how it's gonna be. Like you said, the AI model will be able to absorb the client data, the financial data and all the information that we've put into the system and crank out a document better than a junior associate could.
Ryan Kalamaya (27m 9s):
One thing we've talked about in other capacities is what does that do to the industry? I mean competition is, is ultimately good for Eric and Melanie Wolf. If they're going through a Divorce and there's competitive forces at play that ultimately is a benefit for them. Here in Colorado, we just had a program approved by the Colorado Supreme Court where there's licensed legal professionals and they're essentially paralegals that can help with very simple, I shouldn't say very, but simple issues. And they'll be able to kind of leverage some of this technology and that is the reality that people are facing. And you can either embrace it or you can, you know, ignore it.
Ryan Kalamaya (27m 51s):
And, I, it's going to be coming up here I think, a lot faster than people imagine. But Raif as always really enjoy talking about these issues. It'll be interesting to see what happens when we come back in five years or 10 years to listen in And. we could be totally off base or we could be, you know, ahead of the, the game and and on point. So it'll be fascinating to just kind of see what happens as we go through this technological disrupting force. That is ai.
Raiford Palmer (28m 25s):
Thanks a lot for having me on the show. I really appreciate it. I agree a hundred percent.
Ryan Kalamaya (28m 29s):
Well, until next time, thanks for joining us on Divorce at Altitude. This is Ryan Kalamaya signing off. Hey everyone, this is Ryan again. Thank You for joining us on Divorce at Altitude. If you found our Kips insight or discussion helpful, please tell a friend about this podcast. For show notes, additional resources or links mentioned on today's episode, visit Divorce at Altitude dot com. Follow us on Apple Podcasts, Spotify, or wherever you listen in. Many of our episodes are also posted on YouTube. You can also find Amy and me at Kalamaya.Law or 970-315-2365.
Ryan Kalamaya (29m 9s):
That's K A L A M A Y A.law