Good morning, good afternoon, good evening and welcome to the program. I'm Paul Kiesel. And I'm Jeff Koncius. And Jeff and I are partners here in a Los Angeles law firm. We each have our own unique practice areas and we're going to talk about them a little bit today. But I'll let Jeff give you a little bit of an overview of what we're going to be doing for the next hour. What we're going to be doing is giving you an introduction to the use of artificial intelligence, specifically Chatgpt and Bard, which is Google's version of it. We're going to run through the benefits of using artificial intelligence in your practice. We're going to go through some scenarios in terms of legal research. We're going to talk about specifically some ethical considerations that you have to keep in mind. We're going to touch briefly on the use of CHATGPT to program for purposes of the legal practice and perhaps give a few more examples about how we use it in our practice and then conclude. Perfect. So Jeff and I have a practice, as they say here in Los Angeles, pretty much a statewide if not a nationwide practice to some extent. My practice areas tend to be personal injury, basic personal injury claims, and I also do a lot of mass actions. So we have a lot of cases where you have multiple plaintiffs involved in individual claims, but they're all combined together in a single action. A couple of examples would be the Juul electronic cigarette litigation, where our firm has about 1200 individual Juul users around the United States, but their cases are filed here in California. And I am co-lead counsel of that Juul litigation that just settled. But again, that's an individual mass action. I also do some environmental work. So I've got a case right now with 15,000 individual clients, not class clients, but individual clients, all with claims here in Los Angeles for an environmental exposure claim that they have. And then on the other side of the equation, I've just have individual injury cases. I literally this morning answered, ready for a jury trial in a admitted liability claim where my client was struck by, ironically, a Wolfgang Puck truck and suffered a cervical spine herniation that required a a fusion surgery. So we kind of have the mix and match of personal injury cases on my side. And I'll tell you how I'm using the technology in that world. And now, Jeff, a little bit about what you do. So back in the day, I used to do work similar to Paul. I stopped doing that. I now focus exclusively on consumer class actions throughout the United States, whether it's suing a manufacturer of a product such as a car manufacturer over a certain defect in their cars or somewhat related to this. Right now, we're suing Facebook over the use of their pixel tracking tool on medical websites. And these are nationwide class actions on behalf of everybody whose data has been hoovered up by Facebook. And I also do work in the financial services space against brokers in FINRA or the NFA. And I also use this in my practice as well to an extent which you'll hear about. So we have Jeff and I have done sort of a number of programs about using artificial intelligence in your practice. It's obviously an evolving area of of use, obviously an evolving area of law because there is going to be a whole series of, I'm sure, legal decisions based upon the use of AI in your practice in a whole host of areas. But Jeff and I have done a series of programs for the Los Angeles Daily Journal, which is our California legal communication newspaper. But we've done it electronically and done videos introducing different subjects that I talks about. So to start out the project, a month or so ago, Jeff and I started to think, I wonder what artificial intelligence would tell us to do if we wanted to create an agenda for lawyers to do a program on artificial intelligence. It's kind of fun to ask AI to describe what AI is and how to use it. So, Jeff, what do we do? Well, so what we did is we put into the chat and we asked and I'm going to share my screen. But before you do that, Jeff, I'll have you start with this. It might be worthwhile just for the participants for us to talk about the two different types of AI that we're using in our practice right now. Right? There's two different distinct models that we're working with. So let's sort of set the map of what we're doing in terms of AI. So the first is of course Chatgpt, and we're using Chatgpt for which is a paid service and also what else? Jeffrey We're also using Google Bard. And when it comes to the difference between the two, there are a couple that are significant. They're both what are called large language models. So they've been trained on an immense amount of data and they pull from different places. So, for example, Google Bard takes from where Google products primarily reside, and importantly for lawyers, they have under Google Scholar all legal opinions, and it's also current through today, Chatgpt, as you're going to see in some of the examples that we run, is only current through 2021, which is something that lawyers definitely have to keep in mind when running requests through them. You'll also see perhaps a little bit ChatGPT perhaps can handle some nuance in terms of how it writes better than Google Bard. And those are the primary differentiations between the two products. But we're going to be using both of them today, sort of point counterpoint so you can see how each of the AI engines create sort of different answers to similar questions and identical questions, different answers. So that kind of lets you know at the first instance you got to be mindful of the responses you're going to give to clients or to the court, since the answers aren't the same between both of the platforms and so Chatgpt, which we'll just call it GPT, likely we're using GPT four in the office right now and GPT four is their most sophisticated, theoretically the most sophisticated AI engine that they operate were paying $20. Shocking, like $20 a month is what we pay to use GPT four But you'll learn that you can only ask 25 questions in three hours, and after 25 questions, it shuts itself down. Which is a good thing for Jeff because he can compulsively get into a cycle where he's just asking question after question. He'd go all night long if he had unlimited answers, right? Yeah, yeah, yeah. To be fair, Jeff does that when he's using it to program, which I haven't done in 50 years. But it was very interesting and I was very happy on a Friday night when it cut me off around midnight and it said, You have to wait three more hours because it forced me to go to bed. So yes, Paul is right. The 25 question limit is important and significant and something to keep in mind as well, especially if you're researching something and going down a rabbit hole and you hit that limit, then you're stuck for three hours. But Bard does not have a limit. Bard has a much larger language model that's taking data from. But truth be told, Chatgpt is a much better engine to write stuff. You'll see examples of written documents when we ask it to write letters. Gpt has got really good prose. Bard not so much so you can pick and choose where you want to go. So. So let me share my screen for a second. And what we'll do is we'll just ask Chatgpt And what we're going to do is I'll quickly type in that we are two lawyers. Giving a seminar on the use. Of artificial intelligence. Set forth an agenda for us. As you can see, it comes up with a title, the date, the venue. Okay, the duration, eight hours. Well. Okay. So obviously, we don't want to subject you to eight hours of this. So we're going to do is to tell it to stop generating and say our seminar will last one hour. So now it's going to take are you are you really embracing the future? Exactly. So there we are, the welcome and introduction, the welcome remarks and overview. We did that. The landscape in the legal sector for 15 minutes, probably not 15 minutes for ethical considerations, perhaps because that is something we need to address and use cases. And it will go on. And for purposes of not using up the 25 questions for today, we'll just let it finish as it did. But we could do is we could say this is only going to last for five minutes. What would you do? It'll redraft it. You know, let's do this. Why don't we share screen? Let me let me block copy what you just did. So Jeff and I have our own subscriptions to GPT. So I'm going to open up my screen. I'll share screen. Let me go up to share screen. Um. Gotten their share screens down here and pop on here and go to here and let me do new chat and. Uh, we are two lawyers. Lawyers come on in California doing. A five minute program on AI for lawyers. Please create. An agenda before you hit it and hit change it to 4.0. Paul at the top. Oh, thank you very much. Good, Good catch. That's something to keep in mind. You can run things in 3.5 and you can do as many as you want, but that's not the latest and greatest. You may have heard that GPT four passed the bar exam in the top 10%. Gpt 3.5 is not able to do that. So Paul is on four. He's going to submit the question. The title is a five minute introduction, so it's a different title than the last one. And be interested to compare Jeffrey to see if it's just given us five minutes to do the exact same things as it did for an hour, but less time. Because that's a lot of stuff to get in five minutes. But it's. I can't even read this in five minutes to the group. So let's move on. But you get the idea. And it's a generative process or iterative, I should say. So if you write a request, you don't have to retype the whole request in the second part, if you want to just refine it, as I did, I stopped it and I said, Wait, we only want an hour. And then it automatically started redrafting the whole thing and you'll see that come into play for some of the requests that we're going to do during this seminar. So let's talk about some of the benefits of AI in the practice, Jeff, And what do you sort of see those benefits right now? Think. I think the way that we've been using it the most is that if you use it as a starting point for your legal research, it's very effective in that way. Keeping in mind that it is not always correct, and just because you see quotation marks in a case citation does not mean that that quote actually lives in the case. And it certainly does have some flights of fancy. It will just make stuff up, put it in quotation marks. And that is the real danger which we're going to keep. Bang on throughout this thing. We really want you to understand this. This could be a starting point. And that is where the time saving benefits could be, such as? We're going to give examples about how to write a write me an opening statement, give me the jury instructions, put it in a chart. Things like that do save time, but it's just the beginning of the process for you as the lawyer. Yeah, cannot emphasize that enough. And that gets to the ethical considerations we'll talk about later on. But when you have the AI engine actually drafting a proposed motion and puts in quotes language that is theoretically taken from the case itself, which it would be if we, in normal circumstances wanted to quote from the case, it will use quotation marks on a sentence which does not exist in the decision. Hard to believe, but that's currently the level of sophistication. Yeah. So that would be an epic fail. I mean, even a first year lawyer wouldn't do something like that. And the problem is, is because it is so well written and you get it in a in a in a document that looks so complete, you think that you can trust it and rely on it. And once again, you should not. But there is certainly a time saving element to it. Cost effectiveness. You're going to be spending much less time with that initial draft of something while keeping in mind there'll be a lot of revision to it and it's very accessible. You can use it on the phone, you could use it on your iPad, you can use it wherever you have a handheld device or a desktop. So let's give an example. I received a few weeks ago a reply to a motion. So we file a motion, the defense files an opposition, we respond to that, and then they do a reply. And I thought, well, what is the reply? Can you do a reply? And so we put into Bard and chat. Let's use Bard for a second. What is the reply? And so you want to do that? Jeff You want me to? I can do it. Okay. So we basically went into Bard and we asked Bard to describe for us your California case and what is a reply? Or when is the short reply appropriate? So Jeff is putting our California lawyer. So we give the venue for the question and asking when does this reply appropriate? And Bard is thinking. And tells us. Now you're going to see there in the first sentence, it's a reply as a brief filed by a party in response to a reply brief, obviously, because we just received one. And then it says, Sir, replies are generally not allowed in California, but there are some exceptions and it goes on. If you notice, there is no citations to anything in this response. So you can just type in as a follow up. What are your citations for your answer? Perfect. Jeff. Exactly. So it becomes iterative. It takes the last response. And the question is what are the citations? And this is a perfect model to do it. And there you go. It provides the California rules of court and provides a couple of cases where cert replies were permitted or not and gives you actual citations. Just by asking the question. Give me citations. So if we take that exact same question, literally block and copy it and we stick it into ChatGPT, you'll see that the answer is not going to be exact. So it's also known as a rebuttal for some people who practice back east. For example, I'm admitted in New Jersey. There are some replies commonly used all the time. And then, sir, Sir. Replies. And Sir, Sir, Sir replies, It gets ridiculous. But out here in California it doesn't. Doesn't really happen. What's interesting again about this answer, it's slightly different, but notice there are no citations whatsoever in this response. So once again, we're going to say to it. Set forth the citations for your response. So. Here's the first time it points out that it's limited to up to September 21st. So as an AI language model, I cannot provide specific citations in real time as my knowledge is limited to the training data up until September 2021. That said, it starts to give responses. And what's interesting is, is so far no cases have popped up. And. Nor nor the rule of court. So if we went to Bard, Bard gave it the and correctly identified the rule of court that applies here. That is the right rule of court. This simply gives you a hypertext link to the rules of court. So I would say that in the fail button chat it fails on this one and Bard takes the takes the win. Exactly right. I agree with you. And that's the thing. When you have those two and you play them off each other, sometimes you'll get a better response. You can literally block and copy it over to the other and say, Well, what about this? And so you can actually use the power of of two I's to get there. So we also had an issue where Paul wanted to write a letter to a personal injury client about the taxability of a settlement of his claim. So, Paul, you want to do that one? Sure. So basically, let's start with the question in. I'm going to go to Bard first because Bard is not as good a writer. And so my question is, I am a California attorney and a personal injury. Client has asked. For a letter. On whether. Or not, whether it was improperly whether or not. Personal injury recovery. Is taxable. Uh, can you. Whoops. Can you write a letter? Answering. That question. And so. Here is the letter. And it's not a very comprehensive letter. Uh. It does correctly say that personal injury recovery is not taxable. But let's ask it to provide citations. Please provide. I have to be polite even to an engine. Citations or your. Letter. And so it identifies the IRS code correctly. And it provides a couple of other sources of information. It's not an overly comprehensive letter, although it does tell the client to consult a tax advisor, which of course is important that you do. And it identifies in the letter itself that punitive damages are. Does it say? Yeah. That they are taxable? Gets that correct. Punitive damages are taxable. But what it doesn't do in the letter here is remind the lawyer, I should say, the client, that if the injury is a non physical injury, that non physical injuries may well be taxable. And so let's take the same question. Before you do that, let me just make a suggestion. Ask it to redraft. Tell Bard to insert the citations into the letter you have drafted. And let's see what it does. Sometimes this works, sometimes it doesn't. Okay. It worked. There we go. And again, this is demonstrative of the iterative process, so you don't have to ask it to redraft the letter. It just knows you're working in this environment and then drafts the letter accordingly. Uh, let me see if I can. Avoid having to retype. Let me go back up to the original question. Uh. And that was here. And so I'm going to take the exact same question and go into GPT and. But the question out there. And see what Gpg's answer is. Starts drafting the letter. And identifies what the scope of the request is, and it's very pleasant. Hope this letter finds you well for your quest. I'm writing to provide you with an overview of the tax implications related to your personal injury recovery. Please note that this letter should not be considered tax advice. It's just a really good letter. I would be if I were drafting the letter. This is what I'd be writing. And it properly indicates that non-taxable where there's physical injuries. Now taxable recovery, emotional distress or mental anguish not resulting from physical injury or illness? That's correct. Under most circumstances, it got that one right. It has lost wages, profits or punitive damages. These types of recovers are generally considered taxable income and must be reported to your tax return. That's actually not correct. If you have loss of income, that's a result of a personal injury. I believe that personal and I'll probably get letters or emails from folks on this one, but I believe that loss of income due to a personal injury that causes that loss is not a taxable loss, but that we're. Not giving taxable we're not giving advice on taxes, and neither should you rely on AI for that either or. Me Exactly. But but the reality is, is a much better letter than the last letter was as a as a starting point. And if you can compare the difference between Bard and GPT, this is a much more robust document than the prior one. So that's an example of a letter I really didn't was asked by a client to draft and we used this AI to begin the process of drafting that letter. Okay, let's talk about something that I know is on everybody's minds, which is Easter Island. So my wife and I are going to be traveling to Easter Island. It's one of her bucket list destinations, not easy to get to. It's LA to Santiago because Easter Island is a territory of Chile, even though it's 5.5 hours off of the coast of Chile, much like Hawaii is to the US. So we actually went to Chile a few years ago and there are very specific COVID travel restrictions traveling to Chile. So I wanted to know what travel restrictions were in place going to Easter Island just to get a sense of how the language I would describe tell me what I needed to do. So let's just go in again and open up. Bard and let's reset chat. We said, and I'm going to be saying I will be traveling to Easter Island in December. Uh, December of 2023. What? Covid? Travel restrictions. Restraint restrictions are in place. Paul, when you hit enter, we're waiting for it to generate when you clear your chat. Interestingly, Google will remember the questions, but as of right now, you can't retrieve the responses to your questions, which is a bit odd. Supposedly they're stored by Google. There's a way to turn it off in your privacy settings if you don't want it to store your questions. But if you do in the future, at some point you'll be able to retrieve the answers as well. So this tells me as of May 2nd, which is today when I'm doing this. As of May 2nd, 2023, these restrictions are in place. You must be vaccinated. You can read these. But this telling you what the restrictions are today for May 2nd, 2023. Not telling me in December, but letting me at least know as of today what those restrictions are going to take. The same question. And give that to Chatgpt. And prepare to be disappointed. And so it tells us, is it? He's only current as to September 2021. And therefore really can't provide any current details on what we should be doing, but gives us links to the State Department, to the Chilean government. It only gives us resources but does not give us an answer. And as you can see from the prior response with Bard, that gave us up to the moment responses to what the current travel restrictions are, which is fantastic. Now we have to assume that the restrictions identified for us are correct. But this obviously the GPT model does not get you there. So that was one real life example of how you can use the technology, not in the practice of law, but in your day to day life. And it does highlight the difference between using a Google search to find information. You could have stuck into Google that exact same query, but what you would have gotten back was a number of links to websites, and then you would have had to piece that information together. And in terms of the time saving aspect of AI, that's where it shines, where it kind of aggregates the data and then says here, it's still up to you to figure out whether it's correct or not, but at least it puts it all in one place, as opposed to Paul having had to click through three or 4 or 5 or more websites trying to find the answer to his question. So those are sort of real life examples between the reply and how that applies in California. Drafting a letter for a client and how that would vary between the two different engines and a personal side. If I just simply want to travel and want to know what the current restrictions are, now we're going to show you how really effective the engines can be in terms of actually helping you in practicing law and how you can use these resources to supplant a lot of work that otherwise would go to paralegals or to law clerks and really get you ready, for example, in a couple of different trial types. So we wanted to say, all right, I'm a personal injury lawyer. I am we have a personal injury case going to trial. It's a clear liability case. We want you to please draft for us an opening statement. So and I'm going to pop into Bard and let's say I am a personal injury attorney in California. Uh, I represent. A 50. 52 year old woman who was struck from behind. While her vehicle and suffered a disc. Herniation. In her cervical. Spine. She had. I'm going to use an initial and acdf. Surgery. And has lost one year of employment. Please. Create. An opening statement. To read actually for for for this jury trial. For this jury trial. So that's the hypothetical. And let's see how far does with that. So. My name is Paul Kiesel and I represent Betty Smith, the 50 year old woman who was severely injured in a car accident on March 31st, 2018. On that date, Ms.. Smith was driving home from work when she was rear ended by a driver who was distracted by his cell phone. Now, look, it made that fact up. I simply said that I was hit from behind, but it says was distracted by his cell phone. The impact of the collision caused Ms.. Smith to suffer a disc herniation in her cervical spine. She was immediately taken to the hospital. I didn't say that. Where she underwent surgery to repair the herniation Again, she did have the surgery, but it was not on the day of the incident. As a result of her injuries, Ms.. Smith has been unable to work for the past year. She's also suffered significant pain and suffering. The driver who caused the accident was clearly negligent. He was not paying attention to the road and was operating his vehicle in a reckless manner. Okay. We're here today to ask for justice for Ms.. Smith. We believe that she is entitled to compensation for her medical expenses, lost wages, pain and suffering and other damages. Thank you for your time. In addition to the above, you may want to include the following in your opening statement. So it's giving me some suggestions of what we might want to put in. I'm going to ask why didn't you put it in the first place? Let me say redraft. With your. Suggested. Additions. Taking more time to think. And. I don't really see that it has incorporated a number of the elements that it wanted me to address. Uh, and so an A stands for an anterior cervical. A disc. Foraminotomy. Anterior cervical discectomy and foraminotomy. There you go again. I said it correctly. So let's take the same question from Bard and let's put that into GPT and see how that varies. And it thanks the jury for being there. Much like that, your presence in time and attention are essential in delivering a fair and just decision. I love this. What a pleasant approach. This sudden impact caused significant trauma to the cervical spine, resulting in painful and debilitating disc herniation. As it as the evidence will show. So it recognizes that you need to rely on the evidence. So it's actually reiterating that as the evidence will show, the negligence is what caused the injury. Uh, son of a bitch. It got anterior cervical discectomy and fusion. It corrected me. Uh, so I didn't say what an was and GPT knew it was an anterior cervical discectomy and fusion, which is what she had. A fusion anterior cervical discectomy and fusion. Amazing. So this is a pretty good opening statement. Yeah, but it's a it's a point, though. You don't want to correct the senior partner like that. You want to do it in a polite way. And then all kidding aside, it is important to see the nuance. And what Paul pointed out the evidence will show is a key phrase you want to use in your opening so as not to draw an objection. And that's something that GPT came up with that barred. It's just nowhere even in sight. And that's just a little nuance that unless you're an experienced lawyer and have gone to trial, you wouldn't know that thing. Right? Because remember, an opening statement is simply what the evidence is going to show during the course of the trial. It's not an argument. So you obviously have to continually say this is what the evidence is going to show in order not to get an objection and have it sustained from the trial judge since it has to be what the evidence is going to be. So now let's take this question and ask, uh, GPT to create the jury instructions for this case. So I'm going to say create the jury instructions that I oops, that I would be using in this case. And after the disclaimer. It's going to start. Spitting out for us. Interesting what it's doing right now. I haven't seen this before. So it's giving us examples of what instructions we want to have, but not actually identifying what we call they're called Casey instructions. So I would stop I. Would stop it from generating Paul and then say, give me the Casey jury instructions that. I need for this case. And KC stands for California Civil Instructions. And and it's correcting accurately. That's what Casey is, but it's not yet giving us the instructions. Here we go. So it's actually now identifying the different cases and the case. Casey number and I'm going to let it play itself out so you can see all of the cases that it's giving us. And when this is done. We're going to actually ask it to then put this in a column format, which is what we need to present to the judge at the time of trial. So you can see it's identifying each of the cases that would apply. And look, this is a fantastic starting point. If you're trying your first personal injury case and you don't know what instructions are appropriate for this kind of case, you can use this AI engine at a minimum as a starting point for making sure you don't miss any of the cases that you should definitely have. Again, you need to check to make sure it got it right. And there are probably some that's missing, but this is an excellent starting point. So now I'm going to say put this in a. Column format. And that's going to take the exact same instructions. I would revise that slightly. I would say stop generating and I would say make a chart with the instruction number in one column and the title and the other. Title in another. And this is brilliant because you can then take this block, copy it and paste it into a word document and you will have your formatted instructions to present to the judge. So that is how GPT handles it. Let's go back to the beginning and I'm going to try a slightly different question in this one. I'm going to ask, uh, let me go back to where the question was here. Copy that. And. Copy. Go to Bard and. Let me. I'm personally a lawyer representing. We did this. We want to ask it to Casey. Instructions. Yeah. Create. The jury instructions. Oops. Great. The jury instructions I need. Interesting. Uh. Please. Identify. The KC. Uh, numbers. All right. And then. Put this in a. Column format with the. He a number and. Uh. Description. Now I'm going to spell that and see what it does. What did I miss? Jeffrey wanted to have it. Put it in a column format. It kind of looks good, but not. Yeah, I mean, it just depends sometimes on the verbiage. If you said put that in a chart with the with a number in one column and the title in another chart, that's a good thing about Bard. You have unlimited questions. So you could just keep trying putting in different iterations or sometimes Bard just wants to do what it wants to do and that's as good as you're going to get. And it's just showing off the differences between Chatgpt and Bard. But forgot to have the opening statement. So I put here the basic details of the case here, but I didn't ask it to do an opening statement. So now I'm going to say create an opening statement. So I'm not going to repeat the query. I already put that above. It's giving me the jury instructions. Now I'm going to ask you to do an opening statement. And there's the opening statement. But you'll notice here is that it is a far less comprehensive opening statement than GPT did for me. Does not really comport with what I need to communicate to the jury in an opening statement. And although it does draw upon my original request because I hadn't had to retype the information, it is nowhere near as effective a an engine for creating an opening as in this case GPT was. So that's an example with both Barred and GPT on how you could create an opening statement, jury instructions, and then create a chart to use in that context. So that's how it works in the personal injury world. Jeffrey. So we have the same example. And the only reason why I'm going to show it to you is because it's kind of interesting to see again how you can pause it to get to whatever answer you're looking for, but not necessarily in that in that order. And Paul did it already where he said, give me the jury instructions. And then he said, give me the opening statement, but you can put in anything in there and just the scenario that you're operating under. So I am an attorney in California. And represent a car dealership. And a consumer who took out. A car loan stopped paying. I am going to trial and need the jury instructions for California state court. What are they? So first it's telling me that jury instructions are crucial. Okay, good. Uh, then they. It talks about KC by itself. So. And then it's going to hopefully start listing them for purposes of what we want to see. So number one, contract formation and interpretation. And then it gives a range. Kc instructions 300 to 349 breach of contract, again, 300 to 349. But then it's calling out specifically 303 and 305. So we could let it run through. It'll do the same thing. But then what you could say is, I need a special verdict form. For the jury. Please draft one. And why we're always polite to the to the I. I'm not sure why. Yeah, I just think we're because we're polite. So hopefully it explains what a special verdict form is. But let's say you're a first year lawyer. The judge says, I want to see your special verdict form. You think, oh, my gosh, I've never seen one of these things before. And you stick it in and it's going to give you a basic one. Is this correct? Well, we'll see. Probably not. It interestingly, it formats it nicely with little checkboxes, but let's see what it does on the signature. There's a little twist in California. You don't need a unanimous jury. If you have 12, you can go nine out of 12. And sometimes on a special verdict form, Bard and Casey will have a signature block that says we, the jurors, find by unanimous verdict or something like that. Again, though, if you were an attorney and you had no idea what a special verdict form looked like. Sure, you could Google it, find images, but this is specific to your case and it should give you some ideas about what you're looking for. And this is great. You can literally block copy this into a word document and have it to present to the to the judge for the jury. Correct. So let's see. Here's the drum roll to the best of our ability. Jury foreperson. Please consult with a legal professional so it doesn't say Right. I'm sorry. It got that right. Yeah, it did get it right. So it didn't say anything about it being unanimous. And maybe that's because I corrected it the last time we did this. So those are some examples in the practice about how you can put in practical questions, not just questions that you would be Googling around, whether it's about a vacation or otherwise. Um. So maybe now would be a good time to kind of advance to what are the ethical implications of using this software. What dangers do you have as an attorney in relying on this software and where do we go from here? Jeffrey Yeah, so this is interesting. So there are a number of ethical considerations that you have to keep in mind. And I actually did my own research before I asked Chatgpt and Bard for what ethical considerations it thought we had to be mindful of. So let's just run through those. You have a duty of competence, obviously. There's ABA model rule 1.1 that basically says you should know what you're doing when you're a lawyer. So you can't just start practicing out of your area of expertise. I don't do trusts and estates. Theoretically, I could stick something into Chatgpt or Bard and ask it a trust and estates question. Draft A will draft something and then if I wasn't really thinking, I could then print it out and present it to a client. That would definitely get me into trouble because I don't have any competence in that area and that's not the way you do it. So you have to keep that in mind. The duty of competence under that rule. There's also an ABA comment that reminds lawyers that you have to keep up with changes in technology and reasonably understand its, quote, benefits and risks. And that's something that we've been highlighting throughout this presentation. There's certainly benefits, but you have to be mindful of the risks, which are the limitations of AI. The next one is the duty of confidentiality, which obviously hopefully all attorneys keep top of mind. You have a couple of rules, ABA 1.6 that you can't reveal information relating to the representation of a client unless the client consents. So obviously that brings up issues. You can't just start sticking your client's name and your client's personal information into one of these chat bots to try to figure out an answer to a question before you read the Privacy Policy. And by the way, the. Privacy policy specifically says we have humans that review search criterion. We have humans that identify personally identifying information. So if you put your client's name in there, you're actually beginning to provide data to a third party about a case that you might have to think twice about ever providing that kind of information to the defendant, because that's ahead of myself there. But to anyone, because it's going to be forever in history. And that is a potential problem, right, Jeff? Correct. However, there might be a slight twist to this. If you're working at a firm that has an individual license to a specific AI product, it may actually be in their coverage that it doesn't store the information or do anything with the information. So if your law firm has that kind of a product or you have that product, you may be okay. But before you start sticking personal information into bar ChatGPT or anywhere, that is something you should definitely keep in mind. The next one that comes to mind is the duty of supervision model Rule 5.3. Under the ABA rules, if you don't know this, lawyers can face ethical violations for the conduct of non attorneys at their firms or people who act at their behest. So you can't simply just send work out the door. You put in a question block and copy, send it to a client like that tax letter. That is certainly not the way to do this. And you're obviously not supervising Chatgpt or Bard in the way that it's working. What's interesting is that the comment three two, that ABA Rule talks about outside services that a law firm may use and again, having to supervise them. And it talks about printers and private investigators, things like that, cloud storage. It does not mention AI in particular, but it certainly would fall under that. So you have a duty to supervise. Those are the three that came to my mind after doing research. But what I thought I would do is ask Chatgpt What other ethical considerations should lawyers keep in mind so that you don't have to see me typing? I'll just tell you what it said. It said, Hey, unauthorized practice of law, which was kind of interesting. I hadn't thought about that. You want to share your screen so we can at least see if it's possible what the response was to the question. Sure. So let's do this. We'll stick it in. So let's ask it. Uh, what ethical considerations should an attorney keep in mind? Wow. That's okay. A turn. He. When using. Chatgpt or artificial intelligence. Site specific. Model roles. So it says we're going to look at the ABA model rules of professional conduct. Keep in mind, they might vary by state. 1.1 competence. Okay, we hit that. Talks about a comment. Scope of representation and allocation of authority between client and lawyer. Well, that's interesting. Not one that would necessarily come to my mind. 1.3 diligence. Okay. Communication. This is one thing that I likes to talk about a lot, which is you need to tell the client you're using Chat, GPT or AI. Paul and I discussed this. If you use Westlaw or Lexis to do your searches, I wouldn't think that that was something that you would necessarily discuss with the client. Um, Paul, what was your thinking about this, about the communication aspect? Well, I think that that's the point. In other words, your client, if you're going to charge a client, for example, for the use of AI, as you might for Lexis or Nexis or Westlaw, where you're charging the client for your legal research, I was wondering whether or not you could charge the client for the cost of doing the research using this engine, which 20 bucks a month is not a substantial cost, but whether or not the appropriate to charge the client for your use of this service and my conclusion, it probably is not accurate to to be able to charge a client for this or not appropriate. But that's certainly an evolving area. Yeah. And so it brings up 1.6. And number five, confidentiality of information. 5.3 regarding oversight and supervision, which we already talked about. This is a different answer than when I got previously. It raised bias and fairness, which is ABA Rule 8.4. And that's just a reminder that obviously we have a duty to be unbiased and to be fair in the way that we treat our clients and deal with the law. And Chatgpt was pointing out that there's an old expression, garbage in, garbage out. Chatgpt and Bard were trained on a set of data, and if there was a certain selection bias or a selection in terms of that data, obviously the response you're going to get is going to be biased to the data that was put into it. So that's something to keep in mind. Transparency Rule 1.4. It brought up again letting the client know and that's it. Interestingly, at the end of it I said to Chatgpt and let's see what it does now. I said, Draft a rule of professional conduct. So we're asking AI to draft a rule of professional conduct for the use of artificial intelligence. What rules would you apply? So we're asking them to regulate themselves at the moment. Yeah. And it's going to say. Proposed rule of professional conduct for the use of artificial intelligence in the practice of law and its rule. Responsible use of artificial intelligence. A and using artificial intelligence in the course of providing a lawyer shall. And then, of course, like any good lawyer, it starts making numbered lists. Take reasonable steps to understand the capabilities. One exercise due diligence to supervise the AI tools work three reasonable measures to identify and mitigate any inherent biases or five maintain a reasonable level of control. Six Reasonable steps seven Inform the client and on it goes. Great. And it's brilliant that it actually is creating this massive rule on how we should be using AI and what the ethical considerations are. We ought to draft this, take this and send it to the American Bar Association. Let them adopt it. We'll see what happens. So there we go. And then, of course, it just mentions it's a starting point. So those are the ethical considerations that you should keep in mind. There are certainly many of them. And no, by no means is that exhaustive, but that's something that should definitely be kept top of mind when using it. So I was going to say, well, let's just go from the ethical, which is a very serious side to the somewhat more much less serious side, which is something we touched on earlier. It is possible to use Chatgpt to program for you. We're not going to bore you with real time screens of that. That would make your eyes glaze over. But it's something to keep in mind if you have a project that you want, that there's some sort of automation that you can think of, such as I'm interested in looking for all of the websites that sell this model of a product. Can you write me some code? Chatgpt will actually start responding to you and saying, Oh, I think I can do. I think I can do that and it'll give you something. And it says, Here are some python code and then you can type back. I don't know what the heck python is. What do you want me to do with that? And then it'll. Jeffrey we actually have like 15 minutes. Okay, So maybe we have some time. Why don't you just show how brilliantly this can be done? If you're thinking intuitively as you're going through it? Okay, so let's just see. And while Jeff is getting this loaded, I want to point this out. All of us have been trained using sort of Google like searches where they're Boolean. They're not natural language searches. But what this AI has created is really the process of doing natural language searches that get the results you're looking for. And so we need to get our brains away from the Google based search mechanism to a much more natural collaborative approach to getting your answers. Buying time for Jeffrey to type in his. Well, let's see if this works. I said I'm interested in seeing what websites discuss. A water leak and BMW sunroofs, which was, let's just say, an issue that came up here in the office. Can you write me python Python code to look for that? So it's excited now certainly exclamation point you can use the and then it talks about the specific prompt in Python library to perform a Google search. This is what you do this is how you can search for it. The search will do this and it goes on and on. And then you so then you're like, okay, well now that I have that, what am I supposed to do with that? So you can just say simply, how do I run Python code on my PC? And then it says to run Python code, you need to install Python and set up a suitable development environment. Here's a step by step guide to get you started. Install Python and it tells you whether it gives you a link to download it. Run the installer, open a text editor. Like notepad. To write your code and how to save the file. And then it gets further down into the weeds. And what you can do is you can follow the code that it wrote above. So, for example, up here in the black boxes, it started to write a simple code. And what you can do is you can say you can you can run that code and inevitably it may or may not find the answer. Most likely not. And you can copy back in the chat box and say, Hey, I tried that. It didn't quite work because and it'll say, I'm sorry, how about this? And then you could try that. Or if you get an error message, you can then stick it into the message box and say, Hey, I got this error message and I'll say, Let me fix the code. And this is where the three hour time limit and the 25 questions really comes into play. And it's really very helpful is that if you have my personality, you get all sorts of excited and you sit there and you forget to eat and you're sitting in front of the computer and you start going through this process and then eventually you run out of questions and it forces you to go do something. Then you can come back and you can keep it, keep at it. The use of ChatGPT as a coder is does have an application in the legal profession. And what you have to do is to think about if you are going to manually look for something across the web in a bunch of places, maybe you could ask Chatgpt to draft some code for you to help you do that by itself. So, Paul, we had some examples about some studies that we wanted to use in the office about your dual litigation, for example, about the cost to schools. Is that something that you wanted to, to show? Yeah, although I think what we can do is if we go back to the auto case and have my client who was hit from behind and the question was, when you have a rear end collision or any collision at all, there is a change of velocity which affects the human body. And that change of velocity is called a Delta V, And Delta V is very important in calculating how the body moves within the structures of a motor vehicle. And I had I've got this case coming up for trial and the question was, can you find out the Delta V from a vehicle, the change of velocity from a vehicle using the vehicle's black box or the electronic control module, The ECM Does the ECM store the data of the change of velocity that a vehicle experiences in a collision? And if so, how long does that data stay in place? So I did a little bit of research. I'll open up because this is something that might be an issue that you haven't in one of your cases. So. Uh. Let's say. And in this case, it was a Tesla model three. So does a Tesla model three, uh, show the change of velocity in a collision, uh, in the data? That's captured. In the. Tcm. Let's see what it says. So yes, a Tesla model three does show the change of velocity in a collision in the data that's captured in the TCM. The TCM or Telematics Control module is a computer that records data about the car's performance, including its speed, acceleration and braking in the event of a collision. The TCM will record the car speed and acceleration just before and after the collision. This data can be used to calculate the change in velocity and to determine the severity of the collision. The TCM data can be used by insurance companies to assess the damages caused by a collision and to determine who is at fault. Not sure I can do that. It can also be used by lawyers to build a case for personal injury. So there it found exactly what I'm here looking for, how I'm going to be able to use. So let's say how long is the TCM data stored in the vehicle? Because remember, my incident occurred in March of 2018. And I'm wondering maybe is the date is still there? And it's going to tell us. After it checks its sources. The data is stored in the car's computer for 30 days. The third. After 30 days, the data is overwritten. If you need to access the data after 30 days, you'll need to contact Tesla and request a copy of the data. Very interesting. You can also request a copy of your data at any time by being by going to your Tesla account and clicking on vehicle data. Didn't even know that. Wow. But let's say let's say this was a Land Rover defender. Does it have a TCM? Well, what's the difference between a TCM and an event data recorder? Do you know? And if not, we should ask it. So it tells us the Land Rover does not have a TCM, it has a power control module, a PCM. The PCM is responsible for controlling the engine. So you said, what's the difference between. Tcm and an event data recorder? Between A. A TKM and an event. Data. Recorder. So it tells us a TCM or telematics control module is a computer that records data about the car's performance. An event data recorder is a device that records data about a vehicle's performance in the event of a collision. So that's the EDR. Only looks like the only triggers if there's a collision versus the TCM, which is just reporting the data. And yesterday we were talking about does the airbag have to deploy in order for the EDR to record data? Right. Yeah. And one is an airbag. Uh, control module. Call. I can't help it. I like that it punted. Every once in a while you'll get this prompt from Bard where it just says, I'm just an AI language model, so I can't help you with that. That doesn't necessarily mean that you should give up at that point. Sometimes you can just enter the question again or add a little bit more color to it. Yeah, let's add a little bit more color. Let's say, uh, what information is. Information is collected, I'm going to keep that spelled wrong. When an airbag. Airbag deploys. Come on. Well, there you go. It gave us a much more robust answer, even though I spelled airbag incorrectly. And it identifies it. It's an EDR, an event data recorder. And this is the information that's obtained when an airbag deploys. And it gives us lots of details which are helpful to a personal injury lawyer trying to research what information so we can say, well, I think we've probably gone down that rabbit hole long enough. But but this is a really good example of how you can ask a questions in the Google model. I mean, the advantage to to Bard is Bard is free. So I can do this as long as I want to. Even though people at Google are going, who is Castle? What is Castle looking for? Right. Because they probably know my IP address or they're wondering what what we're doing here. Um, and so that's one way you can do that if you want to talk about our case against the defendant in this litigation and what you're doing. Uh, not not in terms of the use of chat GPT or Bard, but you're talking about pixel in particular. Yeah, sure. So right now what we have, we have a nationwide class action against Facebook because it has literally a piece of code that is one dot that displays on a person's browser when you sign into a website. And that one dot allows a direct connection to be made to Facebook. And it then offloads the data that is being exchanged between you and the website. So for example, you sign on to a hospital website. As soon as it makes that connection, Facebook makes a direct connection to your computer and starts listening in and getting directly communications between you and the website. And so that nationwide class action is pending in the Northern District of California. And we're leading the charge relating to that because a consumer simply doesn't know. Yeah. And you want of course, you need to be careful when we're doing searches, not to call out the kind of searches that we're doing because there's a high probability the defendants are going to get access to the information just by the fact that we're searching information about them in the platform that they might have access to. So again, that's one of the ethical considerations to be really mindful of that This is a publicly disclosable search. Exactly. And also the an earlier idea we said, which is something to build upon. The reason why it's good to put things into both Bard and Chatgpt is because you'll get differing answers such as that air bag answer. We put that answer into Chatgpt. It's going to give you a slightly new, a slightly different answer. You can put all that information together and then you can have a somewhat intelligent conversation with your expert about what data is collected. And of course, you're going to be relying on your expert for the actual truth of the situation, but you won't come in as a total neophyte. And by having used Bard and Chatgpt together, that will give you a much better understanding. Outstanding. Well, I think we're just about to approach our one hour mark. This has been a fantastic discussion. I appreciate your you're paying attention to us. Hopefully, as we've gone through this and recognize this is an evolving area of research. It's an evolving area of interaction and an evolving area of ethical consideration. So you got to stay on top of this, be aware of this, incorporate this into your practice to the extent that you can. And it's an exciting time to be alive and have this available as a resource. But the caution always is trust but verify. Verification is critically important since it doesn't get it right all the time. And in fact it gets it wrong directly some of the time as well. Exactly. This is a tool like any other and not one to be solely relied on. So be well, prosper and enjoy.