The Rise of Automation: Legal Profession Under Threat

0:28Welcome to the AI Chat Podcast. forward to this. This is a topic that's near and dear to my heart. I'm a lawyer, but my undergraduate degree is in computer science. I've been interested in technology for going on 30 years now. Matter of fact, when I was in college, I had one of the first email addresses, which was totally useless because there was nobody else to email. But the convergence of technology with a lot of the professions, including my profession, is something that really, really excites me. So thanks for having me on the show. Yeah, super excited. So that's actually one of the first things I wanted to ask you about.

1:04I saw that you have a degree in computer science. Talk to me a little. I feel like, you know, I know when you go into law, you can kind of get your basic degree, you know, pre going to law school and whatnot. And I have, you know, a number of friends that have gone on to do this. And very few, if any, have I ever heard doing a computer science degree before going on to law school. So it's kind of your motivation or kind of direction with that. Yeah, great question. The short story is when I was a senior in high school, this was in 1990, I basically hacked the computer. I got put, we had 30 connected, there was no internet back then, but the computers were connected literally by cords.

1:46I hacked in I changed everybody's grade to an A I thought it was the funniest thing ever the teacher the teacher did not think it was funny and put me on a standalone computer and that was kind of the start of my interest in computers I go to Texas A &M and I was a general studies major didn't really know what I wanted to do other than play basketball I was on a basketball team at Texas A &M that was really my number one interest but you know I had liked computers since since high school and I started taking some computer classes and it just kind of went from there. I didn't realize when I decided to be a computer scientist that I would have to take linear algebra and differential equations and four different calculus classes at matrix math and stuff like that.

2:30And it was kind of tough. I also did a, I also got a degree in philosophy because I was interested in both of those topics. But in any event, I tell people that I spent four years in college with a bunch of dorks in a computer lab, and I didn't want to do that my whole life. And so I went to law school, right? And now 20 years later, what do I do? I hang around behind a computer with a bunch of dorks. Like everybody does. So it's kind of a circuitous route, but I can tell you that knowing about technology and seeing the development of technology over the last 30 years has been an absolute bend in my law practice.

3:11It's helped me in more ways than I can count. And I'm glad, I'm really, really glad, not only that I have the computer science degree, but also that I have the philosophy degree because I think they're complementary degrees. So that's kind of the short story. Yeah, very cool. And I think that a lot of people are actually jealous of you right now because I feel like as we see with this big AI boom that's happening right now, You know, we're seeing ways that AI is literally proliferating into every single industry, every single niche, and it's automating so many tasks. And a lot of people, I think, feel like some of the more intense ways AI can help is kind of out of grasp.

3:51It's out of reach beyond just using ChaiJPT to help them with some things. And so I think kind of having that experience in your background probably is helping you really set your law firm apart and push it forward beyond what others are able to do because of your background. I'd be curious to, you know, we've had other people on the show that have talked a little bit about AI and law. So I'd be curious to pick your brain and kind of see what your thoughts are on that. Specifically, how do you see AI technologies impacting the future of legal research and case preparation? I'm assuming this is an area it's kind of hot right now.

4:23A lot of people are talking about it. There's been some high profile cases of this going wrong, of course, for lawyers using Batch of BT that cited, you know, cases that never existed and they get in trouble and stuff. How do you kind of see this in the future? Yeah, so I think before I answer that question, I think it's important to know what a, what chat TPT is. Like what, what is a large, what, what is a large language model? Like fundamentally what that is. And I've actually, I've actually, I've actually taken some courses about how to build a large language model from scratch, like a blank slate starting with nothing.

5:00what would you do to start to actually program a large language model? And to be, there's some complications to it. I don't want to get into the technical weeds, but essentially it's a statistics and probability machine. They take, for example, they'll take, they call them tokens. People think of tokens, tokens can be a word or tokens can be a series of letters or it can be one letter. It doesn't really matter. But when you hear the term token, that's what we're talking about. So large language models basically break down sentences and paragraphs and multiple paragraphs and longer than that into tokens.

5:42And they, so you, like, for example, let's have, let's use English. Let's say you've got a token that is typically three letters long. And so we'll use the letter the. What is the chance that the word cat will appear next? And then it calculates those probabilities and it creates basically a matrix that shows all the different words and what the probabilities are. And then it uses some recursive programming, which just means programming that kind of refers back to itself. And it puts it in context. So all large language models are doing is trying to predict the probability of words in the context in which they appear.

6:22So if you understand that, then the magic, because when you use them, they basically they appear like freaking magic. It's like talking to somebody on the other end. But if you know that all they're doing is probabilistic or statistical reasoning, then it kind of changes your relationship with it a little bit. So now let's move it to the law situation. I can tell you that I've been using ChatGPT and other kinds of chatbots and large language models in my practice for almost a year now. I've told my team, I've got a team of people that work for me. A while back I said, this is coming. It's absolutely a game changer.

7:03It's as big of a game changer as I've seen in 30 years. Let's get in front of it. And what I've discovered in the law practice is for research, it's good but not great. you have to check the work that they call it hallucinations they say these chatbots hallucinate you could just call it being wrong and and they're confidently wrong a lot of the times like when you when you say are you sure about yes i'm sure even though they're dead wrong that confident so what does that mean for humans as it relates to the professions and law in particular what it means is somebody has got to check the work. And it's going to get a lot better, but you're always, I don't want to say always, at least for the foreseeable future, you're going to have to have some sort of human eyes that judge the ultimate outcome.

7:54So in terms of legal research, it's incredible, but it's not perfect. There are some other applications in which I think it is absolutely incredible. So I'll give you a perfect example. when I'm a trial lawyer, and one of the things I do as a trial lawyer is I send written questions to the defendants in my case, to the people that have been negligent, answer these questions. And we used to have to generate these all out of our brain, type them up, figure them out. Now, I can say, I can literally explain to a chatbot, I can say, I'm a plaintiff sword. This is generally what the case is about. This is what the defendants are saying.

8:35I need you to generate a list of 20 questions that I can send to them that'll focus in on this issue. And boom, two seconds later, you've got 20 incredible questions. The other thing you can do is like, if somebody files a lawsuit and then you can amend that lawsuit, have a different lawsuit, you can say, hey, compare these two and tell me what the differences are. And it'll write out a paragraph. Now there's all there, you know, you could do that in word processors already would track changes and stuff like that but the output you get here is better and here here's a final kind of example of how i've used it which is really surprising to people so whatever your definition of creativity is you can use chatbots to come up with ideas to come up with creative ideas so i i was having a little bit of a problem i do a lot of cases for people that are hurt okay in the marine in the marine industry and one of the things that the insurance companies always say, I'm always representing somebody who's been hurt or somebody who's been killed.

9:37They always say it was your fault. You should have stopped the work. You have the right to, you shouldn't have done it if it was so unsafe, you should not have done the work. They say that in every single case. And so I was messing around with the chat bot a couple of weeks ago and I said, what's the best kind of, you know, did a nice prompt. what's the best counter argument to this and it spat out a five paragraph closing argument that was it was good i mean tons of work needed to be done but the but the kernels inside of it the ideas inside of it were awesome so i so here so so that's a very long way of saying it's a ai and chat gpt large language models are an absolute game changer not just for lawyers but for all professions that, you know, a profession is basically manipulating information.

10:33Any job that is fundamentally about manipulating information is going to benefit from these AIs and these chatbots. But for now, for the foreseeable future, you're going to have to have human overseers, and you're going to have to have people that know the subject in order to oversee them. And so, for example, you have to go to law school in order to understand how to read cases and research cases, understand how they fit in and stuff. You're still going to have to do that with LLMs and chatbots, because if you don't understand that, there's no way you can oversee it to begin with, if that makes sense.

11:10Yeah, for sure. 100%. And it's interesting, too, I think, right along the lines of what you're saying, I have heard from a number of people say, oh, like I tried chat GPT for X, Y, Z. I did like a garbage job. And it's like, at the end of the day, the better your input prompt is, the more you know what you're doing and the more you know what you're talking about, the higher quality the result of the output is going to be. So if I say, you know, like, write me a article about X, Y, and Z, it's just going to just kind of free flow. But if I'm like, and include these statistics and these numbers reference this and that, and you can put the whole thing together, the quality of the output is like 110 higher.

11:46So yeah, I imagine the more you know and the better of a lawyer you are you know the ways to ask the questions to prompt it to to get you a better result essentially i'm sure as well prompt prompts are what you're doing with a prompt is you're programming your program and now you're programming using the english language or whatever other language it is but if you look at it like writing a program in english it makes a lot more sense. I think where people get stuck a little bit, Jaden, is at least with the LLMs and the AIs right now, is they think it's just like a form of a search engine. It's just like a Google, hey, find me five recipes for scrambled eggs.

12:26What's the best recipe? All this stuff that you can already do on Google and Bing and the other search engines, if you look at it instead as like a collaborator somebody like a somebody you're collaborating with and you ask them questions and delving a little bit deeper and things like that it works far better so the prompts like i suggest to people in my law firm when they're doing it tell it what personality to adopt give it context of why you're asking the question uh what what you want like be specific about what you want the output to look like. So what I'll do sometimes is I'll say, I'm a planarist or I represent John Smith.

13:13John Smith got hit by an 18 wheeler and is paralyzed. The trucking company says it was his fault. They're saying it was his fault because he wasn't driving fast enough on the highway. He should have been driving faster. Give me five reasons that's a bad argument. Write me a two paragraph closing statement and suggest 10 questions I can ask that we'll figure out whether this defense is true. So that's one prompt, right? Yep. Whereas in a Google search, you just literally write one sentence or a few words. So I have that big prompt. I get the output. And then you start playing with the output like you're not done yet.

13:55So I say, all right, add five more questions. Take away this paragraph of the closing argument. replace it with this and you you kind of you kind of it's basically an interactive yeah uh process where you're you're asking questions you're you're modifying things you're sharpening things and so when you look at it through that framework it works phenomenally well if you look at it as just a souped up search engine it frankly does it doesn't work that right i mean i you know like for example i have i've like you said i have my own podcast and one of my podcast managers is uh he started me use ais to write my show notes yeah like the announcement notes and it's awful i mean it is so bad it sounds nothing like my i like to speak with small words simple language yeah not a lot of adjectives i have a particular style i like to speak and write with and i see these show notes and it's like brian beckham has an illuminating most fascinating interesting they talk and it's all these adjectives and it's so breath and the first time he did this i said did you write that with an ai he goes yeah and that'll work but what you got to do is you got to say okay take out the adjectives and then you got to say maybe feed at some of the things I've written before and say, this is Brian Beckham's style.

15:28Write the show notes in this style. And you got to kind of keep, in other words, you can't just tell it one time to do it and expect it to be perfect. So it's actually kind of funny to see how if you don't check it, it'll be like completely obvious, at least to me. I don't know whether it's to other people when an AI is writing something. And by the way, there's a famous, have you ever heard of the Turing test? Yes. Okay. Do you know generally who Alan Turing was and what the Turing test is? Yeah, but explain it for the audience. So Alan Turing was a British mathematician and computer scientist during World War II who broke the Nazi Enigma code, basically made the first computer.

16:13And he has this thing called the Turing test, which is once you can't tell the difference whether something on the other side of a screen is a human or not, then we've reached artificial general intelligence. A lot of people think we've passed the Turing test for some of these chatbots. Yeah, big debate. I can tell you it's absolutely crystal clear to me when I see these chatbots that we have not had a Turing test. I mean, maybe they can fool you for just a little bit, but after a while it's pretty obvious that you're not dealing with a human if you know what i mean yeah definitely but something that you said i really like and i just want to highlight it because uh i think this is really useful for the audience i feel like i for one am not creative enough when it comes to prompt engineering to like for example style right i do show notes for my podcast as well it's very similar right like i uh i find it's hard for me to be like say in like a professional yet sort of casual, like finding all the right keywords to get it to do my exact sales incredibly hard.

17:20So I actually ended up doing what you mentioned, which is whenever I want something to sound like me, I just give it like an article or an essay or like a big thing that I've written. And I'm like, just copy this. And I've done that in a lot of different ways where like I'll find maybe someone else's piece of content, like a YouTuber even or something, right oh like recently i was um helping someone make a uh sas product like video walkthrough demo kind of like video i needed a script for it ask chat gpt and it sounded like ridiculous it was like this yeah like you said this illuminating piece of technology will dazzle you and i'm like oh my gosh like this is so cheesy so i went yeah like i copied the transcript gave it to him said just copy like this style and it gave a great one you know yeah yeah yeah it's funny yeah dazzle I totally got, I was like, I haven't used the word.

18:10I forget what word it was, but it was a similar word. I was like, I haven't used that word 15 years, you know? And it's like every other sentence is like an exclamation mark. You know, I'm not that. But the point is, is that they're, they're, they are trainable. They're teachable. So you can, so I'll tell you what I did with my show notes. So Sam Harris is a podcast. He's got podcasts I've been listening to for a decade. I like the style of his show notes. or I just fed him a bunch of his show notes. I fed one of the guys a bunch of his show notes. I said, this is Sam Harris' show notes. Write my show notes in the same stock.

18:46The cool thing about it is, so in order to be, I'm stepping into some deep philosophical ground or I'm about to. In order to be conscious, one of the things that has to exist is memory. You have to have memory in order to be conscious. Same thing with computers in terms of being able to interact with them. They have to have memory of the previous, in this case, previous prompts and previous answers. And the amazing thing about some of these chatbots is they have a massive memory, like way more than our memory. It'll remember its mistakes. It'll remember stuff you fed it before. It'll remember the style you've talked about.

19:32And so it does, you know, large language model, it does learn over time using modeling human language. So it gets better, which is super cool, I think. Like the more you use it, the more fine tune you get. So, you know, I can actually, I've got this lawyer that I deal with all the time who writes literally like an 18th century English barrister. It's hilarious. I don't even know if he knows he's doing it. But now when he writes me letters, I tell my A.I., I say, respond to this letter. Here's what I want you to say, but I want you to write it in the same style. me thinks thou doth protest too much about you know and it's it's funny to and so you can really have a lot on that's funny yeah yeah that's hilarious you know i think something that's interesting and i think maybe we get close to the turn test with this because i tend to agree with you as far as the quality that it is at right now like not bad it's good it's trainable but it's you know not like automatically incredible um something that you know you see right now is there's a ton of different companies out there that are allowing like corporations to fine tune on top of open AI or other AI models even open AI just launched like a fine tuning thing for chat GPT um I think it would be incredible if every conversation you specifically have on your chat GPT account is maybe a third party that does this or someone else but it takes that it can tell you know when you ask it to redo something that you didn't like the first one you like the second one better it remembers like what you're asking about so all of It's like making data on your businesses or your cases or your style or the way you do things, putting that together.

21:11And then it all of a sudden has a memory about you, your tone, the stuff that you're working on, the things that you do, your work clothes and jobs. It can pretty much get all of that. And I think at that point, we'll come to a much closer place with that, which we're not quite at right now. But yeah, very interesting if we could get it to essentially replicate us in a way, become like our own personal. um that that that exists already and there are multiple companies that are doing that right now i had a guy on my podcast has created a personal ai is she basically inflection or is she doing the um personal.ai that thing i think it's called as i recall it's called perpetuity ai i'll i'll look that up and yeah but basically what it does is it reads all your data and then and you you can train it but like for instance if somebody texts you and says where do you want to go eat dinner tonight it'll if you wanted to it'll say well my favorite two restaurants are this restaurant and that restaurant what do you think about that it'll have an entire conversation with people that you don't even realize I mean it's going on without any input from you at all so that's already kind of here and to me that's just like the natural extension of I'm out of the office this week if you really need me you can contact this phone number call my secretary you know the auto yeah yeah yeah this is just a more robust or robust version of that but of course a lot of these issues and i this is it's pure luck that i have a philosophy and a computer science degree but these a lot of these issues are not technology issues per se jaden they're philosophical issues.

22:59They're issues about what it means to be human, what it means to interact with another potentially conscious entity, what gives us value. You hear a lot about, oh my gosh, these AIs are going to take all these jobs. What are people going to do? Where are they going to find meaning if they don't have a job? And the truth of the matter is, For the vast majority of human beings in their world, their jobs effing suck. They're terrible. We're lucky that we like our job. But for most people, it's terrible. And the idea that we would trap these people forever in these shitty jobs is immoral. And there's no rule, by the way, that says human beings have to get their personal meaning of fulfillment out of work.

23:55Like, where did that come from? That wasn't the case for 10 ,000 over all of human history until recently. And so what I think you're going to see, and there's a bunch of downsides with AI that we can talk about or not, but there's a bunch of positive aspects to it. And I think one of the biggest positive aspects to it is it's going to take away the drudgery type work, the work that nobody, nobody really wants to do. And that's going to put a lot of people out of jobs. But those those jobs aren't that good right now. And they're going to find other sources of better sources of meaning, more more more important sources of meaning.

24:35So I think there's a lot of really positive benefits. is but i where do you i'm so curious to pick your brain on this and uh you know having a philosophy and a computer science degree uh there's so many there's so many things i would love to pick your brain on but specifically right this is the big argument everyone's kind of uh debating it right now you sam all the ceo of open ai he says he really believes that ubi is going to become absolutely essential where do you what do you see right like for example i saw an article today makes perfect sense um gizmodo tech journal uh they had like a bunch of different language departments had their english language they had their spanish they had italian whatever they just laid off and fired everyone on the entire spanish gizmodo like that whole company is just shut down because they said we're just going to use an ai and it's going to rewrite the english one and now it's the spanish one they didn't want to use google translate i guess in the past because sometimes the translations were a little funky but when you get ai to rewrite it It understands the full context and everything.

25:31And it's really good, apparently. So anyways, the whole Spanish one shut down. I can see that happening in everything and everywhere, right? So many jobs. What do you think happens to those people? Do you think that AI is going to, you know, create a need for universal basic income? What do you think the solution is to that problem or that, you know, occurrence? It's a fantastic question. So like the core recorders and videographers in the legal industry are, I would not want to be in that, starting out in that field for the same reasons you talk about. UBI, universal basic income, is this idea that everybody, I think it was kind of popularized by presidential candidate Andrew Yang.

26:13yeah and it's it's got it so the the basically the idea is everybody gets a some sort of income just by virtue of being a member of whatever country it is you use the united states just because you're in the united states you get a you know you get basic health care and you get enough to provide for some basic needs the pushback on that is well then nobody will want to work. Everybody's going to stay home. Everybody's going to sit there and they're going to play video games all day long. Well, who do you hear that from? You hear that from the rich billionaires that require manual labor and have to have cheap labor.

26:56Of course, they're going to say that because they rely on, in many ways, taking advantage of people that need the money. I mean, the Amazon warehouses are a human rights fucking disaster. The fact they have timed restroom breaks. What in the hell are we talking about in 2023? Yeah, it's pretty. It's horrendous. It's horrendous. Like, it is awful. And the people that are complaining about people not going back to the office are the people that run those kind of companies. So, of course, they're going to say, who gets their meaning in having to shit boxes all day long? And as a 40-year-old adult, it has to ask permission to go number one.

27:47I mean, they're not, let me tell you something, people aren't getting their meaning from those kind of jobs. If they are, more power to them. But those kind of jobs, what I would say is it will be a good thing when robots and AI take over those jobs. And it will create so much additional time for people to spend time with their families, spend time with their kids, pursue their hobbies, go get physically fit, spend time in nature, go do something, do charity work for your fellow man. I mean, we ought to be rooting for this is my point. And if it takes something like universal basic income to make it happen, I would support that 100%.

28:32And I guess my point is I think this idea that UBI discourages people from working is a stupid argument. It's the argument that is being made. The people that are making the argument are the people that are exploiting people that have these low-level crappy jobs. So as a lawyer, one of the things I've kind of been trained to do is to look what the underlying motivations are when somebody does something. And I can tell you, this may not surprise you, 99 % of the time, follow the money. So when somebody's arguing, for example, that all these low-level workers, they need to come back to the office.

29:16Ask yourself why. Are they saying that because it's better for the people to come back to the office? Are they saying it because it's better for them? And I think you know the answer to that question most of the time. Yeah, 100%. So do you think then universal basic income would be the solution? Do you think enough people would be displaced or do you think that the jobs that AI creates will fill in the gaps there? I think it'll be a combination of both. I think what you'll find, and I'm already seeing this a little bit on the various social media sites. Basically, people are creating their own personal brands and their brand is their job in many ways.

30:00And so, you know, you can take some extreme examples that were celebrities. You know, they'll get paid millions of dollars to hold a certain kind of drink and drink. Right. But but there's people all across the country that are doing this. So I'll give you an example from my life. So I'm I'm I'm totally obsessed with Brazilian jujitsu. I practice jujitsu almost every day. And Instagram is where all the jujitsu people hang out. and people have personal brands. They make money from making training videotapes and stuff like that. I actually, I think the thing that we want, in my opinion, is we want to take robotics and AIs and get rid of all the jobs that suck.

30:45Right. All the jobs that are drudgery, all the jobs that nobody really wants to do. And that frees up time. Like, let me ask you this question. jaden if you didn't have to work to make money what do you think you'd do with your time oh yeah i have a lot of uh i have a lot of passions and things i'm interested in i do like this is fun for me so some some level of it may be the the same thing i do honestly on the podcast like i i find a lot of uh fun from it but yeah i mean definitely there's a lot of things i'd probably spend more time with my family um yeah i probably need to spend more time thinking about that question because it's not quite the option yet but I do I do find a lot of uh joy in what I do yeah so let me ask you this question if if you had the option to spend more time with your family and pursue more of your hobbies and stuff do you think you would have less meaning in your life or more meaning yeah I think I definitely would find hobbies that give me a lot more meaning in my life for sure yeah and I think I think most people are like that.

31:54Like you and I are fortunate in that I get a lot of fulfillment out of my job. I get to represent kids, families, husbands, and wives, and real people that are in very bad situations, and I can make a difference in their life. That gives me a lot of fulfillment. So I like my job, and I do give meaning for my job. But if that went away tomorrow, I would find meaning somewhere else. I would find meaning helping young kids, like teaching kids jujitsu, for example. I would find meaning I have three children, spending more time with my kids. I love to read. I would find meaning in reading more books.

32:35So I guess what I'm saying is what we want is we want AI to take care of all the stuff that nobody else wants to do. And that will free up time to do, whether it's your job or whether it's hobbies or whether it's something else for free of time to do stuff that's much more human i mean you know let's think about it for a little bit is it you know like using the amazon warehouse as an example is is it is it natural is it good for people to be inside all day long doing manual labor under very strict conditions dehydrated no water no restroom breaks is that is that the way and and you know i should give you an example from my life when i first started practicing law jade i had to show up at a giant steel structure that had 50 floors and i was on the 45th floor and i had to go to a little box and where i had to wear these these clothes that it was like i had a noose around my neck all the tie and i had to put it on every morning at eight o 'clock and I had to get in my car and I had to drive in traffic for 35 minutes.