Demystifying AI: Core Concepts for Legal Professionals [Webinar]

 

AMY LEVINE:

I know everybody here, they’re busy lawyers, so why don’t we get started. And first and foremost, welcome to our first part of our series on a look at artificial intelligence’s legal revolution. My name is Amy Levine. I am the Director of Programs and Volunteers here at LCL, and will be your host today.

For this series, we’re going to be offering a comprehensive exploration of AI and its impact on the legal world, which includes AI’s underlying principles and how to leverage this tool. We’ll also focus on specific AI applications and ethical considerations. More to come, so please check on our website for the future presentations. And today’s webinar is called “Demystifying AI: Core Concepts for Legal Professionals,” and this is presented by our very own Laura Keeler.

Laura is a graduate of Middlebury College and is one of our law practice advisors who works with lawyers on all areas of law practice management, including technology, operations, marketing, finance and more. She also writes articles and speaks on topics relevant to practice management, and I encourage all of you to read her bio for more details.

If you are joining us for the first time, Lawyers Concerned for Lawyers offers consultations on mental health and substance use disorders, educational programming, support groups and law practice management for the Massachusetts legal community. Our services are confidential and free. At the end of the webinar, Laura will be answering questions, and we encourage you to use the chat or the Q&A functions. And without further ado, here is Laura Keeler.

 

LAURA KEELER:

Thank you, Amy, and hello everyone. It is great to be with you. So, as you may know, AI has become a huge buzzword across the last two years, and I know that a lot of people are saying, ‘Oh, I’m not quite sure if I understand it.’ So, today, our goal is to demystify AI by learning about some core concepts and key terms.

So, as Amy noted, this is a kickoff. LCL is doing an artificial intelligence series. Today, we’re really going to focus on the foundations. Upcoming sessions, we’re going to have one in August on a workshop with prompt engineering and ChatGPT. Then we’re going to have one on ethics and AI in September, and AI use cases in law practice and more in depth will be in October, and we’ll have the dates on those from the last slide. But so, to set the framework, today is the kickoff foundation for LCL’s AI series.

So, first of all, why should legal professionals take heed of AI? Why is this everywhere? Why is it important? Well, a key thing is that AI is transforming the legal industry, but you really want to embrace it and have this curiosity, because AI is going to enhance legal practice. A lot of lawyers are scared that it’s going to replace lawyers. It’s not. It’s going to augment lawyer capabilities, not replace them. A lot of law school deans and lawyers have said lawyers who utilize AI will eclipse those lawyers who do not learn to utilize AI.

There are many sample uses that we’re going to go through to give you some ideas of how it can help. Let’s say you have a long document and summary. That’s something AI can help with. Let’s say you are looking for a first draft of correspondence, or you want to change the tone in your correspondence. Maybe you want to change the reading grade level that your client is. AI can help with that. Let’s say that you have a practice management system and you want to query within it, ‘When did the accident happen, or what is the children’s insurance?’ You can question the data sets to retrieve information. So, it’s a lot of uses. It’s going to enhance legal practice.

So, first of all, let’s talk about the scale of transformation. This is revolutionary. It has been compared to the printing press and the internet, and the comparison to the internet includes by no less than Bill Gates. This is really going to transform all industries, including the law. To let you know, artificial intelligence itself is not new. What is new is this generative artificial intelligence. This has developed at an exponential pace, faster than when you’ve seen anything else adapted so far.

So, as an example, ChatGPT, which many are familiar with, as an example, artificial intelligence program, it… had a million users within five days of its launch. In comparison, it took Facebook 10 months to reach a million viewers. So, we’re just talking exponential rate of growth.

To set the framework, we’re going to talk about some terms and concepts. So, first of all, what is Artificial Intelligence? Well, it’s technology that allows machines to perform tasks that require human intelligence. Human intelligence such as understanding, learning from data, making decisions.

How does it do that? Well, machine learning. Machines learn from the data. The validity of the underlying data matters. So, for instance, if you have underneath it – this is an issue right now with ChatGPT, that when they were asking about how to make some recipes, they didn’t realize that Reddit, which is one of the data sets that is scraped for ChatGPT, often has the most upvoted answers as satirical, and not necessarily accurate. So, you really want to pay attention to what the underlying data sets are. This is where we’ll talk about closed and open systems later. For generative AI though, you want to think about it that it uses machine learning to generate original content. This content is creative, but it’s not necessarily accurate, so you have to be careful on that, as we’re going to learn with hallucinations.

So, for hallucinations, this is when AI generates content that is incorrect or it’s nonsensical. A big area in the law has been some famous viral cases where it has fabricated case citations. And how that worked is people were inputting questions into ChatGPT, looking for cases, and the response generated what looked like a case citation, but in fact, she was all fabricated.

So, this is a big risk, as you know, we want to have candor and truth to tribunal. We want to make sure that everything is verified. So, it’s actually pretty easy to do this. All you need to do is, when you get AI generated content, you want to verify it’s true. You want to fact check. You want to see, ‘Are these real case citations? Are these cases on point?’ before you cite them in your pleadings.

I also want to have you understand about search techniques. AI is actually going to make it easier for you to search, because it’s really using more human style language. It used to be, for a lot of legal research programs and others that you had to learn Boolean searching, so you had to have the and/or/not operators, things within a certain number of words. It’s much more technical in how to search across databases.

Now, since AI uses natural language processing, it looks for content, for context, and intent. This is actually incredibly handy. So, for instance, if you practice in both Massachusetts and in New Hampshire, you may know that in Family Law, child custody is no longer a term in New Hampshire regulations and statutes, but it is in Massachusetts. Rather than having to search for parental regulations and responsibilities, if you search for child custody, it will understand that it’s a related context. So, this is very helpful.

We’re going to talk a little bit about prompt engineering and how you can craft inputs to get to your desired AI outputs. We’re going to do more of this, though, in the workshop next month, and talk about how what you input can really depend on what you get out.

So, as a simple example, the more specific you can be, the better. So, let’s say that you were using AI to say, ‘Where should I go on vacation?’ It may give you a very generic answer. But if you were to change the prompt to ‘I’m looking to go on vacation in March for 36 hours to somewhere warm with an amusement park for the kids and such and such,’ then it will narrow down based on the underlying data set to give you something more on target to your prompt.

And when we talk about prompts for Artificial Intelligence… and prompt engineering, that is just how you phrase your search. It depends on the system. Sometimes you’ll have a role and an action, but essentially, you want to think of the prompt as your search bar. What are you asking the machine for information on?

We’re going to talk a bit of a high level overview on some use cases, features and tools and ethical considerations. But I do want to note this is just a preview, because future sessions in the LCL AI series will go into these more in depth. So, first of all, for ethical considerations, we talked about hallucinations. We want to make sure that we’re verifying information, and we want to make sure it’s factually accurate.

Something else is you want to think about confidentiality, and you want to ensure that client data is protected. One way to do that is not to put confidential information into any system that is going to train on your data. You also don’t want to have anything personally identifiable about your clients. You want to be aware that there are biases in the AI generated outputs, and again, that really depends on the data that was scraped for what the model’s trained on. You want to understand that this is a very fast moving area, so most states right now don’t have too much guidance. There are a few that do. There are few, including Massachusetts, have put out opinions on this. But as it goes forward, you really want to stay compliant with ethical opinions and AI regulations, because this is going to be a fast changing area.

When we talk about applications for lawyers – and this is something that the October session is going to go into more depth on – but there are a huge range of applications. These are just some of them. So, it will help us with legal research. It will help you being able to search faster and more comprehensively. It will help with – I apologize to I had increased the text on these to try to make it easier to read, and didn’t realize that it would skew as it came out with this.

It can help you with document review and looking for, let’s say, the custodians, key terms, any sort of first drafts. It can help you with analyzing contracts and legal documents. If you were using predictive analytics, for example, on certain judges or certain case types, it can help you that. And it can help forecast outcomes, draft and review documents, as noted.

One [way] that a lot of lawyers have used it so far is for correspondence. If they’re trying to, let’s say, change a tone of something. Let’s say that you write something out and then you realize, ‘Oh, wait a minute, I wrote that, you know, in too angry of a tone to outside counsel, to my opposing counsel, I want to change the tone to be more professional.’ Or, ‘I wrote this to my client, and I wrote it as though they’re a friend, because we’re very close, but I realized I want the tone to change more professionally.’ AI can help you with drafting, to change the tone.

It can help you with proofreading. And one thing that – I’m really excited – as a newer use case, is it can help you with chronologies and timelines. As someone that used case map to have to generate and time map very in depth chronologies and timelines for cases, this will save you so much time.

So, as we’re thinking about this and how are we going to use AI, a lot of it will depend on the system type you’re using. So, to give you sort of a context, there are open systems and there are closed systems, and they each have pluses and pros and cons. Open Systems are ones like Google Gemini and Open AI’s ChatGPT, ones that are open access for everyone. They’re usually free to use. You know, they have broad data access. But there are privacy concerns. The main privacy concern is usually that the models will, by default if you don’t change settings, train on your data.

There are a lot of closed systems that are being developed. These are proprietary systems. These are ones such as by your Thomson Reuters for your Westlaw AI, Precision Law. These are ones that are developed by vLex Fastcase, by others, that they have closed data sets.

So, when we think open systems that is open access. Closed systems, it is a closed data set, and they’re usually proprietary, so you have often more control on knowing what data, underlying data, the model trained on. So, that is helpful. You can also usually have more secure handling of confidential information because being proprietary, and usually these are licensed to firms, then they’re often not trained on for your data at large. So, those are something to keep in mind that if you’re doing something in depth for your law firm, closed systems right now have more data security.

Something else that… has more security and accuracy is if you use a process tools that have a process embedded within called RAG, or retrieval-augmented generation. And what that is, is it retrieves information from certain sources for accuracy. So, it’s not just generating content. AI works by predicting the words that you want based on the prompt. This actually points to a source, so it has far more accuracy. You do still want to verify, make sure the information is correct and on point, but this is a way to limit hallucinations and to limit inaccuracies.

If you use tools that have the RAG capability, I will note that systems, it’s hard, their terms and conditions are changing all the time, and they all have different settings. But for most of these tools, if you go within the settings, you can switch to toggle off or switch off the option to have the system train on your data.

This is very important, because, again, we want to think about that confidentiality. We don’t want the system at large training on any data that is relevant to the law firm. And we also do want to input any confidential information, so that includes anything that could identify your clients, to specific case information. It’s one thing to ask AI questions such as, you know, generate 25 ideas for voir dire for a case type involving an accident case, you know, where the lawyers are presenting with the plaintiff, and these are the main you know, issues at hand. That’s one thing, that’s more generic, but you don’t want to say, ‘John Smith was involved in an accident on June 10 and such and such,’ because that is really confidential information that you want to keep and need to keep private about your clients.

So, as we talked about – so, these are some heads up to think on. So, you want to turn off the functionality for that large language model – or the AI, as you can think of it, the large, the large language model, LLM, is what the AI uses to train on your prompts. Helps protect confidentiality. We don’t want to enter that personally identifying information.

I also want to note the idea of a temperature scale. It used to be for a lot of these early systems that you had far more control when you were doing the prompt engineering. You could switch over the temperature scale and such and the scale, you can have it be leaned more to the spectrum of being completely predictive or more creative. Both of these have use cases. Most systems now are preset with where they’re going to be, and it often depends on the application. So, for instance, AI that is often used for marketing will be set to more creative. AI that relates to legal research will be set to more predictive. So, that’s something to keep in mind as to how creative or predictive your answers will be.

I want to note that this is an area a number of people may feel intimidated with, but we are all learning. This is something that, as I noted, is going to be revolutionized in the industry, and how you can stay ahead of this and stay competitive is you want to stay updated. You want to have hands-on practice. You want to go to series like this, where you can learn about the AI tools. We want to encourage you to in your off time, you know, maybe start with an hour a week if you can, just playing around with AI, getting familiar with it, learning about prompts, keeping aware of AI advancements. Because AI advancements are coming fast and furious, and there are more and more use cases every week with what can be done.

But you also want to be aware that just because a product has AI doesn’t mean that that means you have to go get it. In fact, if you’re thinking, ‘Oh my gosh, I think I need to update some of my processes and systems, I don’t know where to start. You know, do I need to add this or this?’ As Amy noted, LCL has free, confidential consultations as one of our areas. So, if you were Massachusetts legal professional, you can book a consult with us, and I am happy, more than happy to talk with you about legal tech and what sort of pain points you have, what sort of features would help with those pain points, and then what tools to look for that would fit with your budget and your existing tools. So, just be aware to keep up with AI advancements, but you don’t need to chase the shiny object in getting every new thing.

Now for lawyers, I also want to note AI and the bar exam has been a very, a very keen topic within legal education. So, this is another example of the exponential growth of machine learning and the development of AI. This is how fast it develops.

So, back at the end of 2022 when we had an earlier version of ChatGPT, the GPT 3.5, it scored a 50% on the multi-state bar exam. A few months later, in March, it passed all sections of the bar exam, and it also scored in 90th percentile of active test takers. So, this is how fast the machines are learning. Again, this is not a normal, you know, steady rate of increase. It’s really that exponential rate growth.

Already people are saying that current AI can do work that is offered to legal interns, mid-level law school. But it won’t be too long, we’re not sure if it’ll be months or years before it can perform the work of new associates. What we want to do, think of this like other legal tools. You want to delegate to the technology for first passes of things that it can do quickly and accurately. And then you can use the more advanced tools in your tool set, and also your human skills to really perform at the top of your game. This, like other tech, should help you with eliminating some of the administrative parts.

I will note too that while generative AI can pass the bar exam – and to note too, that was ChatGPT 4.0 that or 4 that passed the bar exam. We’re already on a version past that now, and there will be ChatGPT 5 that’ll be coming out later, which are even more advanced. This technology still needs supervision.

You can think of this like a tireless legal intern. So, AI can draft content for you. It can generate ideas. You can do prompts such as, you know, I am looking for 20 different ideas to what to call my blog post, or what to call this article, what to call this CLE, and if you give it context, such as, this is the topic, this is my article, you know, it can generate however many you asked for. And then you can do that repeatedly. You can write in, ‘Draft another 20 title ideas, another 20.’ It can do this within seconds, and it is tireless.

Similarly, for summaries… you know, think of a legal intern that they are going through 500 pages, you know, and they’re really trying to condense. What is this new regulation about? It can perform quick level summaries, but you as the attorney also need to not only be technologically competent, but also to supervise the technology, the people that use the technology, and you want to confirm the final product is accurate.

This is especially important, as we talked about, with hallucinations. So, that’s, hallucinations were the fictitious case sites. The reason that this happened is that lawyers were using ChatGPT for legal research. That is not a legal research tool. If they had been using an actual legal research tool, it would have extrapolated to legitimate case sites. There have been a couple instances where even legal research prompt cases have had issues with that. But it is far more accurate than using a generic tool like a ChatGPT.

And then you really want to think about too, using your judgment. Even if you use AI as a first pass, you still want to check the databases, read the cases, confirm citations and all that, before you include it in the pleading.

So, before we go to questions, I want to talk about some takeaways from today.

So, AI is transforming all industries and including the legal industry. This has been the hottest topic on the LawSites blog, which is the number one blog in the legal industry in North America for the past year. I expect that it will be this coming year as well. This is something that everyone is talking about and thinking about.

You want to recognize that AI can enhance lawyer capabilities. And those lawyers who use it are predicted to have a competitive edge over lawyers who do not use AI at all.

There are many practical applications. The use cases are expanding and accelerating all the time. Some of the most common ones are that it can summarize large documents for you, it can help with document review, with predicted analytics, with drafting. You also, I will note that if you… live in Microsoft 365 or previously called Microsoft Office, like so many law firms do, if you use a tool like Microsoft copilot, you can even use the AI within those tools to say, ask questions of your Excel documents, to help you create templates and first drafts for PowerPoint, to summarize Word documents. There are a lot of use cases depending on your needs and the tool you use, but main ones are generating ideas and summarizing content and then drafting.

And again, you want to think of AI as a tireless legal intern. It’s great for ideas and first drafts, but you need to supervise and verify.

So, we’ll get to questions in a moment. I want to note too, these are the upcoming sessions in the LCL AI series.

On August 21st that’s where we’re going to have the workshop, playing around in ChatGPT, gaining familiarity with some sample writing prompts, seeing what you can do with it. On September 30th, we’re going to talk more about AI and ethics. And October 16th, we’re going to talk more about specific AI future benefits, looking at some legal AI tools for law practice, [going] more in depth on those, and how you can use them.

I also want to note, as much as we completely encourage questions now, if you are hesitant to ask something in a group setting, or think of something for later, we have free consultations. I included a link below. We encourage you to reach out. We can help you with law practice management, with legal tech, and we are here to help and give independent advice. All right, so Amy, I will look in the chat and I will turn it over to you while we look at some of the questions here.

 

AMY:

Excellent. First of all, Laura, excellent presentation as always. I think that for me, one of the big takeaways again, AI is not replacing lawyers, but it is an extremely helpful tool to be more productive, get work done faster. Clearly, it’s not perfect. There still needs supervision, but again, this is, this is a tool that it’s not going away anytime soon, and it’s only going to get improved. And I think, again, sounds like become more useful for folks.

We do have questions, so I do want to definitely get started. First question, the first question we had: ‘Is there a good, free AI application for lawyers that has good training and sources available?’

 

LAURA:

One of the tools, which I am so excited about, that is free for lawyers, is a tool called Describe. Here, I can put this in the chat and then Amy, if you can send it out to everyone. It’s actually created by Massachusetts founders who are in my, part of my legal tech Boston group. And what it is, it summarizes a whole host of judicial opinions. It’s free to use. It uses that RAG or retrieval-augmented generation I talked about, so it really limits the hallucinations. And if you’re looking for starting legal research that is free, that’s something to try.

Another one that I will say I have actually found a lot of a lot of use for in my personal life is Vetted.ai. I learned about this at ABA Techshow. This is a tool where, if you were looking for product, you know, recommendations, rather than looking for, you know, two or three hours on what sort of suitcase should I buy, or what sort of baking pan for this specific purpose or whatnot, if you ask it about, you know, ‘I’m looking for this sort of product, you know, with a focus on, let’s say durability or affordability, or whatever it may be, tell me some of the best rated products,’ Vetted.ai, again, it’s free to use, and it’ll generate a short list for you. So, very, very helpful. So, those are just some off top the head.

I see there’s a question about what blog did I mention – I’ll put this in too. This is LawSites blog, and this is actually written by a Massachusetts Attorney called Bob Ambrogi. He is the editor for that site.

Let’s see, someone asked about whether the presentation will be emailed. Great, Amy, I see you answered, but I’m not sure if this went out to everybody. Just to, for those who may be listening but not seeing, the presentation will be sent out to everyone, yes, and it will also be posted on our digital library.

All right, someone [wrote], ‘Thanks for putting this together.’ Thank you so much. We’re really, really glad that you were enthused about this, as you noted anticipating our professional needs and delivering appropriate training. Thank you. We’re so glad that you were here. And let’s me go to the questions part as well. Okay.

Oh, ‘What opinion/regulations has MA published on AI?’ There was a case. I can pull it up in a minute and get that for you. Let me just look for the next questions in the meantime.

‘With respect to not allowing one’s data to be used to train the system one is using, can you foresee a day when the AI system owner will insist upon data sharing for training purposes as a prerequisite for using their AI system?’

Um, I mean it could be, it’s hard to say what the AI system owners will insist upon, because they, they all have their own needs. I remember too, meeting someone recently at a Massachusetts conference, and there was a vendor that I thought, ‘Oh, this is not so good,’ that when I was asking them about their product and how they collected data on people, they said that you didn’t have an option to opt out for cookies, which for me is always, you know, a positive thing when they do.

So, it’s hard to know what system owners will insist upon for data sharing, for training. I will say that there are so many AI options out there that you don’t have to think of you are limited to just one. When I first was looking for AI specific legal tools, there were about 19, and when I looked two weeks later, there was already up to 103, so this is something that is growing very fast. So, if you find one that you don’t like the the terms of service, that you don’t like that they ask you share data, you can just find another tool that meets your needs.

Our other question, ‘Thanks for great overview, given the pace of growth, how is it possible AI won’t replace hiring future associates?’

Well, that is something that law schools are thinking about, how do they retrain current law school students, so that they can be more effective as future associates. I don’t think it will replace the hiring of full future associates, because even if the AI is, say, able to do first year level associate work, the law firm model you’re going to still need, you know, second years, third years and such. And how does, how do you get those? Will you get those from first year associates? You know, you have to go up through the training process.

It may replace some of the specific duties that future associates would be doing, such as AI generated document review tools such as Relativity, Concordance, you know, helped to just streamline the amount of document review that was needed from first draft from associates. But I think associates will still be needed. And Jordan Furlong is a great legal futurist who’s talked a lot about this.

We can, in fact, Amy, maybe even have an additional session on talking on human needs and how lawyers can really embrace what they will be able to do that the AI can’t.

In the meantime, well, I see, if more questions come in, I will look up the case with the Massachusetts AI guidance.

Let’s see. Okay. So, this was in Massachusetts Lawyers Weekly in February [2024], a lawyer was sanctioned $2,000 by Superior Court Judge for submitting a legal memoranda that included citations to wholly fictitious cases. This is what we talked about with the legal hallucinations. And, let’s see. Oh, Lawyers Weekly wants me to sign in. Trying to just get to the article here…

Okay, Smith v. Farwell, I mean, I can type that into the chat and Amy, if you would send that out to everyone.

All right, so Smith v. Farwell, and I’m sure that Luz will go into more on that when we have the AI ethics on September 30th. As far as other states that have put forward AI guidance, Florida and California have really been leading the way on that.

Okay, this question, ‘I know for Massachusetts state employees, including prosecutors, there was guidance back in April.’ Oh, thank you so much, Edward.

Okay, so yes, it looks like there was a Mass.gov Policy Advisory on acceptable use development of generative AI. That’s right, yes, because there are some government agencies that have guidance if you are government. Well spotted. Thank you so much for putting that in. And, okay, great. And Amy shared out the name of the case and also what Ed had shared with us. Beautiful.

Okay, you have a question, ‘What works well to summarize medical records,?’ Oh, I do not have a specific program name off my head for that. If you’d like, I can do some research and get back to you, but I don’t have one off the top of the head. Amy, are there other questions in the chat or Q&A that we have not attended to you? And I will say too, if people are thinking about the question, if you haven’t written in, please do so.

 

AMY:

I think you really did hit most of them. ‘What opinions’ – I’m not quite sure if we got to this one – ‘What opinions, regulations, has Massachusetts published on AI?’ Did we get to that one? I know they were coming in fast and furious, so I want to make sure that we do, we don’t miss anything.

 

LAURA:

Thank you so much for highlighting that. Yes. So, the one court opinion that I’ve seen that Massachusetts has published on AI, that is the Smith v. Farwell, where they were fined – the lawyers fined – $2,000 for submitting legal memoranda that included fictitious case sites to the court.

And then one of our attendees also helpfully shared the Mass.gov policy guidance for state employees, which includes prosecutors, on how to use AI. So, that is for opinions and regulations MA has published on AI thus far.

I will note that if you are cross-licensed in other states (such as a lot of Massachusetts attorneys are cross-licensed in the other New England states), and in Florida, if you are, say, cross licensed in Florida, you’ll want to read their legal field blog for their ethics opinions, because they have much more guidance on what can be done with that.

All right, let’s see. ‘Can you please share your findings on AI to summarize reading on medical records with the group?’ Oh sure, yes. So, what we can do — good question. Amy, after I do some research on that, I can include it to be as a supplement when we put it up on the website for, in the Digital Library on the LCL site, include some information about what I find on AI tools, to help with medical records.

 

AMY:

That sounds great. And also, I just want to highlight that Luz Carrion — I’m going to send this to everyone — just posted from the BBOpublicmassbbo.org website, the Wild West of Artificial Intelligence. And she said this is the article from the Office of Bar Counsel on AI and ethics, and will be discussed during the ethics webinar. So, I am going to send that off right now so everybody can take a look at that.

 

LAURA:

Thank you. Thank you, Amy, and thank you, Luz.

 

AMY:

And I know there, I think we just had another question come in…

 

LAURA:

So, he shared with the hosting panelists, ‘If you can share with everyone.’ This is very helpful. So, he gave a short version of the Massachusetts State guidance. ‘If you’re a Massachusetts state employee [the guidance] is that you need to be aware of the risks of AI, do due diligence on reviewing and confirming any generated information, and disclose any use of generative AI, and not put in any PI/PII/PHI into the system,’ and that’s… personally identifiable information or personal health information into the system.

So, this relates back to what we’re talking about with confidentiality, doing your due diligence to verify the information that it generates out, understand the risks of AI. So, all of that, most of that, relates to lawyers as well. The only thing that would be separate is for the Massachusetts State guidance you need to disclose, disclose the use of generative AI. And lawyers at large do not need to do that so far. All right.

Oh, and we have also, Ed shared, okay, ‘And furthermore, for the Massachusetts guidance, if you’re a state employee, any tool should be procured through the Massachusetts CTO for approval and to evaluate any tools.’ Okay, because there’s ongoing review and evaluation needed for ensuring security. Okay, very good.

 

AMY:

So, Laura, we do have a question on, just going back here… ‘Have you ever heard of SmartDepo for summarizing depositions?’

 

LAURA:

I have not heard of that particular one, but I do know that that is an area that last time I looked, there had not been any AI tools in. I am very hopeful since Thomson Reuters had just recently announced that they were able to do more with chronologies that, from that, we’ll be able to be getting some more AI apps related for depositions, because, as we know, those can be very complex.

You know… you have information that’s going back and forth with, with different timelines and different levels of information for detail. So, no, I have not heard of SmartDepo yet, but I will write that down, and I can get back to you in that, Maureen. And I can also put, if you’d like, when we put this up for the Digital Library, some thoughts about AI for summarizing depositions. Great questions.

 

AMY:

They are great questions. I mean, it says to me, people are really paying attention and thinking about all of this. So, that’s a great, great sign.

We have a question. ‘Is there a program that can fill out an application or other form? Example, every year I hate putting together our worker’s comp audit, where another example will be filling out a financial statement for divorce.’

 

LAURA:

Yes, I can see how putting together information for audits would be extremely laborious and time intensive. I don’t know of a program that can fill out applications. I will say, though, if you use templates, that can help. If you use… document automation, that can help if you, you know, let’s say, pull things from your database.

So, it depends too if you have fillable forms, I’m not sure how you stored your information for your workers comp audit or financial statements for divorce. But if you have them, for example, in your practice management system, for the financial statement for divorce, you should be able to, from what I’m hearing on the vendors and what they’re working on in development to coming up, be able to query any of the documents that you have within your practice management system, you know, such as a Clio or MyCase or Actionstep, one of those, which would include your financial information that should help you with filling out financial information, financial statements for divorce.

I will say, if you do a lot of divorce work, I did a CLE not too long ago on client intake, and one that was really helpful for divorce lawyers was something called Divorce123, because they helped with having very specific forms and templates specific for divorce. If you do a lot of financial statements, not sure if it would help you with QDROs or such, but perhaps.

As far as a worker’s comp audit, I’m not sure, because I imagine that the audit would be pulling from a lot of different places, but if I find anything on that, I will definitely update you. But as I said, if you can try to use some document assembly, if you have, you know, templates to pull from, that can help.

If you’re unfamiliar with the idea of document assembly, some of you might be familiar with mail merge. This was a thing that was very popular some years ago. If you would have all of your names, clients and their addresses and such in the Excel document, and then you would have a form template letter. Then, rather than changing the name and the address on every letter, what you could do is you could do a mail merge from within your Microsoft Word document, link it back to the Excel document where you had stored the information, and then you would map the fields over such that it would fill in for the first name, the last name, the address, the state, the zip code and such to fill in the template, and it would just help assemble the document that way.

So, I would imagine, if you have some templates to work off for your workers comp audit, and if you have some underlying data that you can put to that is in mapped fields, such as an Excel document or something like that, you could do some document assembly to help with it. It might take some work to set up, though. The more you can try to have set fields and standardizing where you keep information, the easier it would be to use document assembly. And I’m happy to talk with you in a consultation more about this, if you want to go more in depth on your areas.

Okay, let’s see. Amy, do we have more questions?

 

AMY:

Well, I just, I wanted to point out somebody earlier made a comment, and I want to give you kudos, Laura and Luz, who’s out there. Somebody just said, ‘Thank you so much for putting together this and related seminars and workshops on AI. This is a perfect example of you anticipating our professional needs and then delivering appropriate training.’ So again, that’s a kudos to you and Luz on understanding the needs that are out there, and that AI can be frightening if you haven’t used it before.

So I do, there is another question, not sure if we got to this one. ‘Can you please share your findings on AI to summarize reading on medical records with the group?’ Did we get to that one?

 

LAURA: 

And that was the one I said that when I came across information, I will include it as a supplement to our presentation up on the Digital Library.

 

Amy Levine 

Okay. Okay, great.

 

Laura Keeler 

Oh, it looks like we have another question as well. ‘Is there a good free, free AI application for lawyers that has good training sources available?’ Yes. So, there are so many great resources for AI training. It sort of depends on where you want to start. So, there are, if you’d like, more like a course sort of content, there are things on program platforms like Coursera and Ed Annex.

If you use, you know, say you that your office lives within Google. You know, there’s a Google Cloud introduction to generative AI learning path. If you live within Microsoft Outlook, now called Microsoft 365, there are specific tools for learning Microsoft 365 and Copilot. I will note too that Copilot, they actually, even though it’s called Microsoft Copilot, there are a number of different versions of the tool, depending on the operating system you have and on the level of access. But I just want to note, so if you, you know, live in Microsoft or Google, they are, they’re a great place to start.

If you are looking for sort of a general one… IBM has ‘AI Foundations for Everyone,’ that is very well rated. There are a number of schools, if you’re looking for the more technical side, like University of California has a class on big data, artificial intelligence and ethics. I will just put as an aside, having started my learning on this area, having taken class on big data, when they say big data, the spreadsheets and the amount of data is just massive. So, just keep that in mind if you’re interested in that part.

So, yeah, I would say it really depends on what you want to learn. If you want to learn about… training resources for concepts, for training versus for specifics tools. There are also some, such as Anthropic’s Claude, that have available prompts within them. So, that’s a vendor tool, but it has listed on its page some past prompts that people have used and found helpful. So, if you have something specific, this goes out to not only the person that asked the question, but anyone in the audience, that you would like to have a training resource on specific to XYZ, please reach out. And I’m happy to figure out sources available for your particular needs.

 

AMY:

I don’t see any additional questions in our chat and Q&A, but I would like to ask this, and maybe you mentioned. So, because there are so many different areas or companies that are doing AI – like ChatGPT or Gemini – are they, are they scouring different sources of information so you can potentially get different information? So, does my question make sense? Where one might use Reddit or, and somebody might use a different… So, is, do you have a feeling of what might be a better, a better one to use? Again, Gemini versus ChatGPT just comes to mind because I tend to use both of those.

 

LAURA:

Yeah, that’s a great question. I will say the hard thing is, most of the AI companies have what is considered more like a black box that they don’t want to be too specific on where they get their data. [They] say too, the versions of the data depend on how long they scrape until. So, for example, on 3.5, GPT 3.5, that was scraped up until about December 2021 or 2022, and then the later versions will rely on more content. There is some more recent data that can also get into those, those data sets based on what the information trains on, but at large, what was scraped at large to put into the data sets — and when we’re talking scrape too, this is trillions of data points that is on. So, they’re scraping like, you know, Reddit at large, they’re scraping Google Search Worms or searching legal blogs or searching vacation blogs or searching travel sites, or searching every sort of main website that is on the internet…

I do understand that from what a lot of the companies have used, it does have bias towards more the Western world, because it generally has scraped from North America. This may be in part because Europe has a lot, tends to have more strict guidelines on privacy than we do, and since most of these tools are developed as English first, [this] may be why they had more North America than Asia, Africa or some of these other places.

But it is hard to know what exactly they’re trained on. I do know that there are some lawyers, Greg Siskind is one who comes to mind, he is working on an AI tool. And part of what he and his team are doing are, they are trying to, on a small scale, test when they do prompts across different AI tools, and then see, you know, does the output differ from this one to that? As far as I know there has not been a cohesive ‘such-and-such-tool is, at large, more accurate or has a better data set than another tool.’

I know some of the ones that are most popular are Microsoft Copilot, ChatGPT, Gemini, which used to be called Bard. This is a hard thing too, I’ll note with AI, is that they keep changing the names for these tools, so it’s a little hard as we go through not only different versions, but different names. And Claude which is made by Anthropic, is another very popular one too, if you’re looking for open-source systems.

 

AMY:

Excellent. This is this is great, Laura. You are so incredibly knowledgeable and a great communicator. Again, I do have another question, thinking about folks that are neurodiverse and using this as a tool, one of the things that occurs to me, especially specifically for folks with ADHD, and they’re on a Zoom meeting…

Do you know of some type of AI or software in general, that helps summarize the meeting and what the most salient points were? So that person, again, who has ADHD, lots of things going on in their mind, can really be able to pull out what the biggest topics were and maybe the plan of action that was discussed.

 

LAURA:

That’s a great question, Amy. And I like how you’re thinking about our neurodiverse audience as well. Yes. So, there are a number of tools now that have recently added on summarization within meeting tools. You have to first of all make sure that if you are on a work computer, your IT admin has allowed in the permissions and the settings for you be able to use certain features, because sometimes they’ll block it for everyone, and you also sometimes depend on what level plan you’re on.

But yes, Zoom has AI summary, Microsoft Teams can as well. A lot of the web conferencing platforms can. If you are looking for a tool, if you’re, say, using something like a Cisco Webex, which I don’t think has an AI summarization tool yet, you can use another translation summarize tool, like an Otter.ai to help with summarization. But if you’re using a tool like Zoom, yes, there is a feature within Zoom to help summarize and capture those key points.

I know some folks that work at Microsoft, actually, and they say that it is very helpful for them. They use the internal features within Microsoft to summarize their meetings, have the AI capture the next steps, who’s responsible for what. And they say it’s really helped team productivity and cohesive communications.

 

AMY:

That’s excellent, right? So, it’s clearly not just for folks that are neurodiverse, but also helping the teams work more efficiently. And that makes complete sense, because again, you know, you always have that person, maybe who is a notetaker, and maybe they didn’t get everything, and we know for sure that this AI will get everything. So, that is wonderful.

Again, so helpful. We have a couple of minutes left, so just want to open it up if anybody else has any questions. And do you have an opinion on and recommend any AI systems that can create images?

 

LAURA:

Oh, DALL·E, I’ll include these in the chat, and MidJourney are the ones that are generally most popular right now for creating images, and that you can do some remarkable things with those. In fact, they even already have some AI image competitions based on who can have the most you know, artistic images that come out of AI machines.

 

AMY:

That’s excellent. And I’m wondering, are you able to personally spot a fake image? I know that sometimes people talk about their extra fingers, or the fingers are particularly long in some of these pictures, AI generated pictures. Any other pieces that you know you can use to identify maybe an image that is that is AI generated and not a real photograph, or something like that?

 

LAURA:

It’s hard to generalize, but usually they’ll be more stylized, more vivid colors, sometimes have almost like an illustrative look, as though it was a person who was trying to draw a picture for a picture book, and to really make everything stand out, and it’s maybe overly crisp. Maybe there’s no background noise at all, and it’s probably something that’s had some AI manipulation.

And if you were, if you’re using that, say, like a camera app on your phone, when you want to use AI to delete background noise, so that you just focus on the person in the picture that you want to and not the crowd at large, that’s great. But it is a different thing if you’re going to be submitting these documents for something like the court filings, and that’s an open area about how do we make sure we don’t have manipulated images and evidence going forward.

 

AMY:

Yeah, yeah, no, I think that’s a whole other topic, right, that can be talked about.

So, we are at the one o’clock mark, and just want to say thank you to Laura for sharing your knowledge again, once again, great presentation, as always, and thank you all for joining us. As a reminder, you do get a feedback form. We’d really appreciate you filling that out so we can always improve our systems. Somebody just said, ‘Thank you very much.’ Again, thanks for joining us, and we will see you at the next session, which is a workshop, August 21, on playing with ChatGPT and gaining familiarity with writing prompts. All right, thank you all.

 

Laura Keeler 

Thank you so much, Amy. Thank you all. Take care, and we’ll see you on the next part of the series.

 

Amy Levine 

All right, thank you. Bye.