 Thank you for coming to today's session on getting started with AI. I'm Miles, your AI-generated emcee from Murph.ai, and I am just one example of the many things that AI can create for you. Now, here's your speaker, Shelly Reid of the Legal Services National Technology Assistance Project. Miles, as Miles said, my name is Shelly Reid and I will be introducing you to the wonderful world of generative artificial intelligence. Today's session is going to be a brief introduction to GenAI, and I'll be using several GenAI tools. To prevent technical difficulties, I've created recordings of many of my interactions with the tools, and we'll be seeing those throughout the presentation. The slides and the embedded recordings will be available along with the recordings of this webinar on our YouTube channel shortly. All right, let's get started. First, I want to take a poll on see how many people have used AI before. If you've never used Slido before, you can just take your phone and scan the QR code, or you can go to slido.com and enter that seven-digit number there below. Let's see, how many people have ever used AI? So far, the results are higher than I expected. Perfect. So we have about 63 percent currently saying they have. We're about 22 percent saying no, others are not sure. So that gives us a picture of where we are as we get started. So I find myself using Microsoft's being GenAI chat a lot, and one nice feature is the voice input option. This triggers an audio response. So it's really helpful if you're busy working on something, you can have it responding in the background. So let's see it in action. Here we're on Bing. Searching for everyday tech that uses AI. Hi, this is Bing. I'm happy to help you learn more about artificial intelligence, AI. AI is the branch of computer science that deals with creating machines or software that can perform tasks that normally require human intelligence, such as recognizing faces, understanding speech, or making decisions. AI is becoming more and more integrated into our everyday lives, and you may not even notice how many things you use or interact with that rely on AI. Here are some examples of everyday tech that uses AI. Voice assistance, devices like Amazon Echo, Google Home, or Apple Siri use AI to understand your voice commands and respond to your queries or requests. They can also learn from your preferences and habits to provide personalized suggestions or reminders. Image recognition, many smartphones use AI to unlock your phone with your face or to enhance your photos with filters or effects. AI can also help you organize your photos by detecting faces, objects, or scenes in them. Financial fraud detection, banks, and credit card companies use AI to monitor your transactions and flag any suspicious or unusual activity. AI can also help you manage your finances by providing insights or recommendations based on your spending patterns. Recommendation systems, platforms like Netflix, YouTube, or Spotify use AI to analyze your viewing or listening history and suggest content that you might like or enjoy. AI can also help you discover new products or services based on your preferences or needs. So whether you realized it or not, AI has been changing the way we do things for quite some time. In late 2022, generative AI technology became available to the public through chat-based interfaces such as ChatGPT, followed by Bing, Bard, and thousands of tools that are now available using those interfaces. So the first thing we need to do is really learn a bit of terminology. Artificial intelligence is the capability of computer systems to mimic human cognitive tasks like problem solving and learning. It's how we describe a computer system's thinking. Machine learning is an application of AI that allows machines to learn these capabilities by extracting knowledge from data and capturing it in a mathematical model. Generative AI is a class of machine learning models that, given a prompt, can generate outputs such as text, images, code, video, music, and lots of other things. A large language model is generally, though not necessarily, a gen AI model that are trained on huge amounts of data such as the entire public internet. And they contain billions of parameters that they use to respond to our queries. GPT is a large language model that was trained by the company OpenAI and they were the first really that released it to the masses. And chat GPT is a chat bot that was trained on a special chat transcript data and built on OpenAI's GPT model. And it's really, it's a new way of doing things where it has a conversational interface and can be thought of as one application of an LLM. So we also need to understand what AI is not. AI is not a search engine. It's not going out and summarizing the materials that it finds, even though many systems seem to function like a search engine. AI is not a research tool. It uses algorithms to create text based on predictions. So I guess I should clarify that. It's not a research tool in the way we think of research in the past. AI is not trustworthy, even if it's very convincing and it is really good at sounding correct. So you always have to remember to verify the information that AI finds for you. And AI is not infallibly consistent. You can ask the same question to the same generative AI model and not get the same results. So it's, it generates text, it generates other images based on what the algorithm predicts at that time. And one thing that you can do is many of the systems have what they call temperature settings and they allow you to set how much creativity or how much leeway you give it in its responses. Some people have called this the bullshitter meter and the higher number is more random and more likely made up. Since last year, hundreds of companies have introduced AI tools that specialize in everything from transcription to poetry. And on your screen you can see just some of the tools that are available. And we're gonna focus today on more of the chat and search tools. We'll mention some of the art tools and some of the speech or text to speech tools but there are so many more available. So AI is being used in the legal industry for tasks such as document review, contract analysis, litigation prediction and legal research. AI is really good at first draft creation for certain things like client emails in letters and with good prompts you can produce material that is a good start but it always needs the human touch to come after. AI is really good at rewording things or making them more readable to the average user and it can also almost be thought of as a turbocharged thesaurus. AI is good at, it seems really good at research. So again, we mentioned earlier that it's not a research tool per se in the way that we think of research but it is able to go out and predict things and give answers to queries. And there's a really fun survey that Sam Hardin did and he can find it at Sam Hardin's LinkedIn page and he did a survey of legal advice versus legal information. And I think as we're moving forward in using gen AI for legal work this is gonna become really where the key is determining whether generative AI responses are going to be considered legal advice or legal information. And it was really interesting to me that when the generative AI did a disclaimer that it was not providing legal information more people found that, the disclaimer was that it was not producing legal advice, I apologize. More people found that the answers were legal information and not advice. So AI can be really beneficial in the practice of law. The things that it can do, it can be really efficient and AI has been used for document review for years. Chatbots and virtual assistants are where we're starting to see some action in the legal world and it's really great at streamlined client interactions because clients want an immediate response. So with a chatbot who is trained on a closed environment and not the whole internet they can give good advice or good information, not advice, excuse me and respond to queries and common questions and office hours and things like that that clients want an immediate response to. AI chatbots are also cost effective and they can also be used to predict legal outcomes based on historical data and so they can be used to shape case strategy. I mean, like Lexis has been using AI to predict what a judge will decide on a case and so we're gonna see more and more of that as we move forward. Some of the potential benefits that I think we'll see with AI are that it's going to reduce the cost of service. It's going to allow us to provide faster service, provide better service and serve more people. It's also going to improve access to legal services and access to justice by simplifying court processes, providing self-help and information, connecting self-represented litigants with providers and other resources and it's going to assist in language and literacy assistance. Right now it's really good at doing plain language editing and it's going to help open that field up for organizations that don't have bilingual staff, for example. Some of the things that we need to think about in the legal field are the ABA model rules and those rules require us to be competent in relevant technology. So just by attending this webinar, you're learning about AI so that you're meeting those obligations. The other thing we need to think about is the privacy concerns that come along with generative AI. Each of, when we put material into a generative AI system, that material can be used to train the system. So as an attorney in working in the legal field, we do not want to put personally identifiable information for our clients into those systems. It's recommended that we use closed systems that are designed and built for legal work. Some of the other considerations for using AI in the legal field are we need to be concerned about the accuracy of the generative AI system. AI systems are known to hallucinate. That means that they make up things when they don't have an answer. We also need to be concerned about bias that's built into these systems. We all know that our country has, excuse me, has a bias issue and when those biases are in the training data, that also is reflected in the results. We also need to be considered of the ownership. Who owns the data? Who owns the system? Where is the training data being housed? And also if you're creating something with AI, the courts so far have ruled that AI created materials cannot be copyrighted. So you can, if say you're creating content for your websites and it's completely AI generated, you can't copyright that information. However, if you edit that material enough, then you can copyright materials and you should be editing because you want to make sure that the material that the AI is producing is accurate and true, you want to make sure that it didn't hallucinate for you on you. So by now, almost everyone has heard of the case in New York, MATA versus Avianca, that ended with attorneys being sanctioned for relying on cases generated by chat GBT. The real issue in this case was that the attorneys did not validate those cases. They did not take them to a legal research site and verify that the cases that chat GBT produced were real cases. And then they buckled down and when the court asked them about the cases, they went back to AI and asked AI if the cases were real and AI said, of course, they're real. And so instead of actually spending the time and verifying those cases themselves, they used a system that is prone to hallucinating. The other thing that we've seen in the news is the do not pay controversy where Josh Browder offered to pay someone to let AI feed them the words to argue before the Supreme Court. Obviously we're not at the point that AI is capable of creating a true legal argument. We've also seen many judges in the US district courts responding to the Avianca case by requiring disclosure if AI is used in preparation of briefs. So all of these things are building up in regulation but there have been many opinions that the current regulations that requiring technology competency are enough. But I think we can also create some guidelines that will help us as we use generative AI. So what we're beginning to see in the legal field are closed systems that build on larger language models to create tools that address the concerns of the legal field. And on your screen you see some of the products that are being released. And these closed systems mean that they are trained on legal related documents whether it's the Harvard Law Project where it has all the cases or Lexis database of cases. So it's not going out to a larger pool of information but it's pulling from legal vetted information and it has less chance of hallucinating when it's using these closed systems. And now we're gonna move into what you've all been waiting for. Let's get started using some of these generative AI systems. So we're gonna try out Microsoft Bing first and we're gonna learn about some of its features. Bing works best when you use the edge internet browser to get started, simply log into Bing or speak your query as I will do in this demo. Hey Bing, give a very brief description of how to get started using you. And as you can see it takes just a moment. Hello, this is Bing. I'm a chat mode of Microsoft Bing that can help you with various tasks and queries. You can ask me anything you want such as facts, trivia, jokes, poems, stories, code, essays, songs, celebrity parodies and more. I can also create graphical artworks based on your prompts. To get started, just type your message in the chat box and press enter. I will try to reply as soon as possible. If you wanna change my chat settings, you can use the toggle button on the top right corner of the chat window. You can choose between three modes, balanced, creative and precise. Balanced mode is the default mode where I try to balance between creativity and accuracy. Creative mode is where I try to be more imaginative and innovative, but sometimes less factual. Precise mode is where I try to be more factual and rigorous, but sometimes less imaginative. You can switch between these modes anytime you want, but it will start a new conversation with me. I hope you enjoy chatting with me. So I find that, you know, the sometimes less factual. So it readily admits that it's going. So chat GPT has received so much attention. Let's watch chat GPT in action. Chat GPT and other generators are really good at summarizing information. So I thought we would do a demo of something that, you know, we have to worry about in my office and that's the image sizes for posting to different social media platforms. So let's look at what is the best image size for posting on social media and see what chat GPT tells us. That's pretty good. Instead of having to go to each site and figure out, you know, where to look and where to find the information. Within seconds, chat GPT was able to give me a list of all the different sizes for the different platforms, but it's not really what I wanted. I was hoping to get one image that would work for many. So let's see if we can go a little farther. And the prompt there, if you couldn't see it was, is there one image that would work for all platforms? Chat GPT is telling me that no, there's not one specific, but they might be, we might be able to use a 1080 by 10. So can you think about things where you need to summarize lots of information and how this feature might be usable for you? I'm hoping that you're starting to see the benefits. So have you ever been in a Zoom meeting and you noticed an attendee named OtterAI or Fireflies AI in the meeting? What you're seeing there are AI transcription services and these services are definitely handy. However, there's some difference of opinion on the legality of recordings without notification. Also, there are definite privacy concerns when the conversation is with a client. You may have noticed that Zoom notifies you when the host records a meeting. So somewhere Zoom's lawyers are cheering because they won that battle, but it's something that we need to think about as attorneys. If using a transcription service for telephone calls, for example, be sure to consider the laws of your state. I was in a meeting with lawyers recently and I was really surprised that the enthusiastic response when asked how many of the attorneys were using these services. And it didn't seem that they were even considering the laws of the state where they were practicing, but it is something that we have to be mindful of if you live in a state where recording without notification is illegal, it's certainly something that you don't want to be doing. So all of these systems require us to tell them what we want and how we tell them is what is called a prompt. And when you're creating prompts, you want to make sure that you follow the guidelines that you see on the screen. You want to be clear, you want to give context. For example, you might say that I am in a attorney and I need to give a speech on creating a PowerPoint presentation. And so you tell the generative AI system what you need. Use complete sentences. For some reason, the AI systems seem to perform better when complete sentences are used. Ask open-ended questions, just like you would with a child. If you don't want to know for an answer, don't ask a question that it can limit the response. Ask multiple questions versus one complex question. So break your questions into multiple sentences. And then tell the system what kind of format you want. Do you want a table? Do you want a paragraph? Do you want to limit the number of words in the response? If you find that you're having trouble getting the expected result, there are some AI tools that can help you create better prompt. And we're going to use Fido as an example. We're going to take a quick look at how it can help you with your prompts. So let's go take a look. Here, we're in Fido and we're going to enter the prompt, craft a client letter saying, I can't take their legal case and hit the generate button. And the system creates a prompt that can be copied and then pasted into chat GPT. So these prompts, there are prompt generators for texts, there are prompt generators for image creators. So if you have problems, maybe try using the prompt assist generators for a little while and then that will help you be able to learn how to craft better prompts on your own. So a fun way to bring AI into your world is to use one of the many AI headshot generators. So using photos from your phone selfies and other things, AI can create a number of options for you. I will warn you that AI has problems with glasses just like it has problems with hands. So as you can see, all of these images, all the photos that I uploaded had glasses because I wear glasses all the time, but you can see the results are mixed. And it seems to me when I used photos with glasses, the results looked less like me because the system was having to try to generate something without the glasses. But you can, if you do a lot of speaking and you want different headshots, this is a fun way to build your portfolio of headshots without having to go to a photographer every time. And what we're seeing in the AI headshot generators, now they're using DNA to predict facial features. So in the future, you might not even have to upload a selfie. So that's actually being used in the criminal world, you know, criminal legal world now to create mugshots of who they think might have done something based on DNA evidence. We won't get into the problem, addict aspects of that though. So, but think, if you're going to use one of these headshot generators, one of the things that you wanna think about, review their privacy policy and opt out of having your photos used for training. Many of them also give you the option of selecting to have your photos deleted once your images are created. And I would recommend doing that, then you know that it's not being used for the training data. So you heard Miles introduce me at the beginning of the presentation. I wanna show you how you can create Miles using Murph AI. And this is fun if you have to do a presentation to have someone or have someone introduce you or if you're gonna have a podcast to creating segment for the podcast. Well, here at the murph.ai homepage, we're gonna create a new project we're gonna select audio and start blank and create a project. Then I'm gonna paste in the texts that I want to have read. From here, we can divide up the text into pieces. We can add pauses. And this is very similar to what I did to create the introduction. If we want Miles to pause, we can add different lengths of pauses. We can change the tone, whether it's going to be conversational or inspirational. We can change the pitch, the speed. We can also change the speaker. However, there are different charges for the different speakers. So let's get rid of the pro so that we have just the free voices available and we'll generate the audio. Thank you for coming to today's session on getting started with AI. I'm Miles, your AI-generated emcee from murph.ai. And that's how it was done. So one of the things that GenAI does well is summarizing long documents. Let's go to Claude and watch. We're an anthropics Claude now and we're gonna use Claude to summarize the US Constitution. So we simply click on the attachment icon and find the file that you wanna summarize. Let it pull in and then tell Claude what you wanted to do. Summarize this document. And Claude gets to work and will produce a summary for us. Some pages down to a few lines. This would have been great in law school trying to summarize all of those case files. So also, this would be really great for summarizing long documents. When I was in law school, I worked at an organization and we were doing social security disability cases and the many, many, many reams of documents. It would have been great to have been able to use a system like this to create a timeline and to create a list of all the different cases and claims and all the different pieces of that case. GNI tools are really good at cleaning up text and making it more readable. Let's take a look at how Bard handles this task. I've preloaded Google's Bard with the instructions, edit the following text to be concise and easier to understand. Now I'm going to paste in the Gettysburg address. Let's see what it comes up with. Here we can see that 87 years ago is certainly more understandable to today's speech than four score and seven years ago. I really like Bard because it shows exactly what it did to make the changes. We can also go over here to see other options. Let's take a look at draft two and that is certainly more concise. And draft three, again, it's provided all the directions or all of what it did to change the document. So again, using like this as we try to make our websites more readable for the clients and more usable and making sure that we're using language that's understandable to all, say, you can tell the systems to use a seventh grade reading level in rewriting for readability. So another way you can use AI is to create images. And this is really nice because you can create images specifically for your needs and you don't have to worry about whether or not there are copyrights on those images. So we're gonna go to bing.com and use, sorry, bing.com slash create. And the best thing about this is Bing's image creator is free and it works great. So because of the heavy usage, Bing limits the fastest response times through a point system and each week users are given 15 points that allow images to be created faster than those without that boost. So this just helps level the playing field so the occasional user is not stuck in the queue behind others who are making repeated images. And if you do any creating of images, you'll quickly find that AI art has some quirks. I mentioned earlier that it's not good at creating hands and glasses are problematic, but it is, I mean, all of these images on the screen were created. And you can see here the prompt that was used when you go to Bing's image creator, if you hover over the image, you'll be able to see the prompt that people use to create that image. And that's one way of learning how to create prompts yourself. And you may not think of Canva, the graphic design tool as an AI tool, but Canva's had AI built in for quite some time now and they recently released their Magic Studio series, which is multiple tools that use AI. And the new Magic Studio incorporates AI into everything from PowerPoint presentations to recordings, to images. And the price is right because they have a very generous nonprofit discount. It was free for nonprofit organizations that may have changed in the last few weeks. They may have increased the price slightly. I know they've limited the number of team members that could be in, but their nonprofit pricing is certainly worth asking about. So AI has the potential to lift some of the burdens of our everyday tasks, but there are some serious concerns that we have to consider and we've mentioned those throughout the presentation. You know, I mentioned earlier that I worked on a disability case while I was in law school. And AI would have been great at making that list of dates in the summary of clients condition, but a summary would not have provided me the full picture of the client struggles and it can't be an acceptable substitute for knowing the details of a case. It sure would have helped comb through those stacks of documents to help counter the government's objections, but nothing replaces actual knowledge of a client in your matter. So AI can be a tool, but it is not the only thing that we need to use. So now I wanted to just kind of see where we are and see if any of these tools that we've mentioned so far have you consider using them? So what generative AI tool are you considering using now? We've talked about Bard, we've talked about Bing and chat GPT and Canva as this piqued your interest in using some of them. I won't give just a minute for people to respond. And the nice thing about the word clouds is getting to see things. And also not just the tools, if you have a job that you think that generative AI can help you with, are there tasks that this has got you thinking that it will help you do? Because sometimes all we need is, you know, seeing what someone else is doing to give us an idea for ourselves. Not very good. All right, so we're gonna move the poll will still be open. So if you want to go in and look later, you can see any other responses. So what can we expect from AI in the future? Right now they're working on AI that can smell. Research is underway to develop AI systems that can diagnose illness based on someone's breath. They can detect gas leaks, they can create perfume. Using sensors and cameras, touch AI can detect whether fruit is ripe and place it into a basket for someone. So at some point in the future, grocery bots will be able to select bananas, for example, with the ripeness level you prefer. AI is they're working on hearing capabilities. And, you know, for example, one of the tests that they've done is an AI that can analyze the sound of a gunshot and then it can alert authorities. I mean, haven't we all heard fireworks and wondered if it was a gunshot? So AI will be able to do that for us. So I apologize, this is not the slide that was supposed to be here. I'm not sure why we don't have it. So if you want to keep up to date with what is on the horizon, well, maybe it is. Google Labs does have an AI test kitchen that gives sneak peeks of things that they are working on. Right now they're working on a gen AI tool that can turn text into music. And I think that's gonna be fun. So we had text to speech. And one of the things I've done is in the past is create text and ask AI to create a song. So now we'll be able to, you know, in the future we'll be able to put text in and it will create a song and create the music for it. If this has got you thinking about the ethics, I know we've just barely skimmed that portion on AI. We do have an ethics of AI webinar coming up on December 19th. You can always go to our website, to the events page to see what's coming up. But if you want to scan this QR code, it will take you directly to the registration page. And we also are always trying to come up with resources that are helpful. So we've created an AI information resource and it's available on our website. It's searchable. You can look at all of the cases. I have alert set up. I'm gonna put a link right there so that you can go right to it. And you can see cases, you can see judicial opinions. You can see state bar opinions and state bar information on AI. So this, we're hoping that it will become, you know, the web, the resource for everyone for knowing what's happening in the world of AI and legal. So I wanted to leave plenty of time for questions. So I'm going to leave my contact information up here. And I'm going to, I've got to get so I can see where people are. If you have questions, I would love to answer the questions that you have. And let's talk about AI and how it can help you in your work. Maybe raise your hand using the control panel and or just unmute if you have a question. Molly, I have a question. Sure. This is Molly from Quality. Hey, hi everyone. Hey, I was just wondering what, so I was thinking about AI in terms of just the evolution of tech and how we started with just, you know, regular search engines. Now we have AI. And there was a movement and I'm not sure whether it happens or not. Maybe someone knows on this. Maybe you know. But if, so the way that search engines, I'm not sure that they work like this, but I was just thinking in terms of, you know, making money using AI if I were like a vendor or whatever. And I said, what is the best seafood restaurant in Denver? AI will generate a response based on, you know, previously entered data. But what if the oyster house pays a company to elevate their restaurant to the top of an AI response? Is that gonna happen like it does in search engines? Do you think? I think that it certainly could. Based on the training data and also just like search engines, you know, they, I think it will happen. I was just curious if you thought that might be the evolution of making it into more of a commodity than a free, you know, sort of a, well, it is biased. I mean, you read all the articles, but just if that was the potential of the future, the use of AI could sort of look that way. Well, certainly. And for example, GPT-3 is based on information up until 2021, I believe. So it can't give you anything that wasn't on the internet before 2021. So if you're using the free version of your chat GPT, then your information is already old. So there's lots of problems with that because there's not, well, it does, it sometimes will tell you that in its response that my information is based on information up until it's where its training data ended. However, at some point, just like, you know, Google and Yahoo and all the other search engines, they have to monetize. So what that will look like, I don't know. You know, there certainly seems to be a movement, you know, with the White House's statement on AI, you know, and the seemingly commitment to ethical AI use that the developers are paying lip service to anyway, you know, but at the end of the day, they're still selling a product and how they monetize it, we have yet to see. Okay, thank you. That's where closed systems, I think, is where the legal field is going to find the most use because we can choose the data that is used for training and those closed systems will be safer for putting client information because right now we ethically can't put client information into one of these large language models that is not a closed system. Right, right. We did, we met with case, I think it was case, GPT, case text, and it was pretty cost-prohibitive at this point for a nonprofit and no nonprofit discount. I think it will evolve into, you know, a broader use and audience at some point, but right now it was really cost-prohibitive. There are a few AI systems in development, we're working with an organization that want to or say they want to provide their service to legal aid organizations as, you know, a benevolent kind of thing as kind of a side job to their profit-making service. So we'll, you know, we'll keep an eye on that and certainly if it's something that we're able to share with the community in the future, we certainly will. So I think that, you know, legal aid and nonprofit market will have options in the future at an affordable price. It's just not going to be anyone's top priority. I don't think any of the companies that are creating. Yeah, they weren't really interested in following up with us or anything at that point, but okay, thank you. So what other questions, comments, thoughts? Well, I don't want to keep people if there are no questions and certainly we'll let everyone go. I will stay around for a few minutes if you didn't want to ask questions on the recording. Actually, we'll stop the recording now. I'm going to do a short closure and then we'll stop the recording. And if you have questions that you want to ask off the record, we'll be available to do so. So thank you everyone for coming to the webinar today. I hope that you learned some things and gained an understanding of what generative AI can do and are a little bit curious and are going to go out and try some things. Have a good day.