Lead with AI: Transforming Cybersecurity with Artificial Intelligence

So Marcus Kerry, he is a cyber security veteran. He's an entrepreneur. He is an AI researcher and he's got uh decades of experience. Um he's also uh in the forefront of network defense innovation. Um and how many you think here AI is going to kind of just clear us all out of the way? No. So that's why Marcus is here. He is in his talk, he's going to kind of let us know that uh really AI is going to really help us leverage the things that we do. He's going to help us uh see how to transform, enhance, and amplify in whatever area of cyber security that we're in. So with further ado, uh here's

uh Marcus.

[Applause] Hey, how y'all doing today? >> Hey, thanks for having me. Every time he tells that story about how I was the first talk that he ever saw, it makes me feel really old. And so, um, so this is pretty cool. So, usually I I've been doing I'm I I'm aspiring to be a stand-up comedian. So, I'm keep on doing this road work and I'm I'm ready to go. Right. So, uh I'm Marcus Kerry. I'll tell you a little bit about by about my background and uh I usually don't do slides lately for presentations, but I do have slides here and I'm waiting for the slides to boot up, so it's fine. All right, cool.

Sweet. And I think I have a clicker here. I'm not too good with computers, so I hopefully I can figure the clicker out. So, uh, in Tampa, Tampa is kind of like a second home. Uh, I work for Reliquest. Uh, and I've been working for Reloquest almost the last 6 years. I'm always in Tampa and, um, always uh, always loving Tampa. Love the Tampa people. Uh, and we got people from from all over the place. We got some Texans here. Hey, can can we Texans? Can Texans uh, raise your hand? Texans. Texans. Y'all y'all shy? I heard a lot of people. Who's from Austin? Somebody's from Austin. >> Yeah. >> Do you know me? Do we know each other?

>> Yeah. All right, cool. So, we got people from all over the place. We had barbecue yesterday at the speakers dinner. Uh, and I always if if I'm out of state, uh, I always scrutinize the barbecue. Funny enough, I I did a talk in CA in Canada at Sector a couple years ago and they took me out to a speaker dinner and they took me to a Texas barbecue place in Canada. I was like, "Man, I didn't come all the way to Canada for barbecue. Give me some poutine or something." All right. So, uh this this uh I've been fascinated by AI since I was probably like five years old. And I'm not kidding you. I'm going to show y'all the

evidence. uh five. Well, it's kind of like circumstantial evidence in in court, but I'm going to show y'all. So, in in third grade, I saw war games. I'm going to talk about that a little bit. War games kind of like inspired me. One, I was like hooked in. I wanted to work with computers. So, today we're going to talk about AI and uh and AI is two things to me. Uh it's not just artificial intelligence, it's actual intelligence too. We have to have we have to know how we learn, how we train, ho, how we how do we do all these things. So, what I'm going to do is I'm going to give you I'm going to

give you a lot of homework today. This is going to suck. We we're at a college though, so it's totally appropriate. I'm going to I'm going to give y'all a lot of movies to watch. I'm going to give you a lot of books to check out because Tampa traffic is is terrible. So, if you have to drive, you can listen to these on audiobook. I mean, I was like, man, Tampa traffic is terrible. It took me like 30 minutes yesterday for my Uber to come, and it took me like 40 minutes to get to where I was going, and I thought Austin traffic was bad. But this is way better than Los Angeles. I'll take this over

Los Angeles any day. All right, cool. So, we're going to talk about actual intelligence. I'm I'm going to game you guys up. For the last about three or four almost four years at Reliquest, I've been doing AI research and development. And uh we have a pretty cool team there called Octo. We have a machine learning team that's been there for about six years now. So uh Reliquest is kneedeep into AI stuff. This is not going to be a it's not going to be a Reliquest commercial. It's going to be a Marcus Kerry commercial though. All right. Anybody know who this guy is? I think I heard somebody whispered his name. Yeah, it's a Shukan guy. This is my boy

uh Bruce Potter. Bruce Potter, he I I'm gonna I'm stealing his line, so I'm going to show his face. Every time he does a talk, he says, "Don't believe anything I'm about to say." That's what he said. He tells the crowd every time, "Don't believe anything I'm about to say. I'm going to give you resources and you can go look stuff up yourself and you can make your own decisions because I'm going to talk about bias far as AI. Everybody's biased." And and if you if you say you're not biased, that's a lie. All right. And before I start, I love I love conference conference people. He started Schmukcon. Um he's a he's a serial

entrepreneur done really well for himself. He's a really good guy. And I just want to say shout out to all the conference organizers here. Let's give a round of applause for the people that put this on. [Applause] Again, don't believe anything I'm about to say. This is pretty cool. This is every all the images in this this uh thing are it's only one image that's not AI generated and that's going to be an image I I'll tell you when it's not AI generated. This is totally AI generated. And uh I am all in a lot of people hate AI. I let me get let me get a a a hand raised here so I can kind of feel the

temperature of of what's going on here. How many people love AI? That's that's about I said that's about 40%. How many people hate AI? How many people using AI dayto-day at their job? That's about That's about half. That's about half. So, we got some we got some adoption here. We got some haters. Hopefully, I can hopefully I can uh you know adjust the haters temperature on on AI because I'm I'm I'm an AI fanboy. All right. So, I got this picture up here. One of my one of the best AI things I've heard, you've probably heard all kind of pe people why people hate AI and all that stuff. So, one of my favorite uh

things is like Amazon was able to predict if a woman was pregnant before the woman was pregnant, before she knew she was pregnant based on her purchases. Like, that's crazy to me. And apparently it kind of tipped some people off that people were pregnant apparently because of their order ordering habits. That's crazy to me, right? That's that's that's awesome. So I'm like, "Wow, that's a crazy story." And you probably heard about mortgages, insurance, all this stuff. Everything is AI. So that's why a lot of people hate AI. I'm going to go over a lot of reason why people hate AI. And I'm going to go over a lot of reasons why I love AI. And the more you know, how many people

know that logo? I I'm dating myself. So, I I totally believe the more you know. I'm 100% into knowledge. Also, one of the things that I love is this mindset and the hunger for knowledge and to just do better. So, I'm all about the more you know. Again, you're going to have a lot of homework here, so hopefully you can take notes and all that stuff. This is where I'm from. This is a real photo. Marlin, Texas. I'm from a small town in Texas called Marlin, Texas. And the reason why this is funny because this is actually a building that's still standing in my hometown, right? And so the reason why I like to talk about uh where I'm from, I'm I'm

the black Forest Gump. I am a country boy. I'm the country you can get. I I grew up listening to George Straight number eight. Y'all might not y'all might not know nothing about that over here in Florida. That's a that's some real Texas country music though. This new country, I can't stand it. It ain't country. All right. So, grew up in this place. This is like the only building standing for what? Only building standing in a in a whole downtown little area. The town that I'm in, the town that I live in, they haven't had clean water for the last three weeks. Their water plant totally shut down. So, I'm from a very poor area, right? But thank goodness to

technology, I've actually been able to do pretty well for myself and and for my family. And and this is why I love technology so much. I think technology is like a really cool uh equalizer as far as anything. And and I'm going to quote a book later. It's called The Fourth Age. I think that I encourage everybody to embrace technology. And I I I that's what in my life that's what's makes the difference. Right. So, how did I get from Marlin to to this? Life imitates art for more than uh it imitates life. So, throughout my life, I've got all these inspirations. You're going to see some of my favorite movies. And every time I enter a different

space, whether that's cyber security, if I would enter cyber security, I would just go watch all the cyber security hacker movies. Now, since I've been doing AI, one of the first things I did is to understand the culture, the language, all the things that you hear about AI. I watched a bunch of AI movies, funny enough, because the truth is that a lot of times they have technical consultants in these movies that are AI experts, and you're going to see a lot of the same things. I'm going to show you books 10 years old. I'm going to show you movies like over 10 years old. And some of the movies are like almost 50 years old that

I'm going to show you. And all these and many of these movies talked about the same exact things that we hear everybody we see all the news headlines about AI this and how AI is going to take over. This has been going on for like 50 some years. And actually AI has had about three or four boom and bust periods throughout time. So it's important to to go forward. We have to understand history. So, I'm going to talk a little bit about the history. I'm going talk about some really good books. And what's really cool is this should be a nice little baseline. So, whatever talk you do, whatever you do professionally, whatever y'all do, I think you can you

can put a little bit of, you know, you like down where I'm from, we like to make gumbo, right? So, you can put whatever you want to in the gumbo and you can put some AI in there and it's going to be lit. This is what I say. Life imitates art and art imitates life, right? Because now we have it depends on what where you fall is like is is AI stuff is that generating art, right? And then humans are humans and machines and the things that we build are building each others. Somebody got their GPS on. What's going on? See, AI is everywhere. I'm telling you. So, I got my start in US Navy, the

world's finest Navy for all y'all. I So, so 18 years old, I I I I began my my career do doing cryptography in the Navy. And and I and the reason why I'm talking about where I I'm going to tell you like this is how Marcus Kerry was trained. We talk about how machines were trained. We talk about how AI is trained all the time. I'm going to tell you how I was training. And everybody in here is trained in a particular way. And also we need to learn how to learn too. And that's one of my big things. So how how was trained? US Navy. I I was lucky enough to score pretty good on my ASVAB

test. And I and I got into cryptography. So as 18 I was coming from uh I was coming from the hood. You saw where I'm from. Friendly corner, right? and and I got I got assigned to work at NSA in NSA subcommands. So when I was 18 years old, I was straight out the hood and they made me a spy. And this was pretty cool, right? I felt like the black Jason Bourne. You're going to listen to this talk. I'm a lot of black something. Like so the so the Navy taught me all this cool stuff. I'm all over the world spying on stuff, doing sigant elint, all kind of different intelligence stuff from 18 years old. And uh like I was

enlisted so I didn't have any college education at all. Right. But the but the Navy they will train the heck out of you and and you have to do it. I mean if you're in the middle of the sea you have to learn a dang job. That's what I love about the military. Ain't nobody coming to save you if you're in the military. You better figure it out. So the whole time I was in I did cryptography. So I always worked for NSA the whole time. Eight and a half years worked for NSA. And um a couple of things. It's funny because like I think like when sometimes somebody say they work for a company or I do I work for

Facebook or I work for Google you kind of think something about them and uh and if you say oh I work the NSA right I'm kind of like a mythbuster not everybody working NSA is smart but it sounds good though don't it oh man he worked at NSA okay so uh working there again a great opportunity to learn uh unlimited budget I I got like over $100,000 worth of technical training myself. I was in training doing work, in training doing work, in training doing work. So they taught me a lot of stuff. They taught me internet networking, cyber security. They taught me how to code. They taught me all this stuff. So I'm pretty lucky,

right? And thank goodness one of my favorite movies is Boys in the Hood. Anybody like like that movie? Yeah. But in that movie, they said the military is no place for a black man in a movie. Thank goodness I didn't listen to that advice. So eventually I started my own company called ThreatCare and I I ran that company for six years. I did the whole venture capital route and uh and it was pretty cool because I ended up selling my company to to uh selling my company to Reliquest. I also wrote these this book series called I I created this book series called Tribe of Hackers probably heard of. And then I finally landed every quest.

So I usually don't do slides. So this slide thing is kind of new to me. All right. So at Reliquest, uh what's cool is I I built uh the first breaching attack simulation platform on the market and Reliquest purchase purchased my company. Got to rely and I helped integrated into the gray matter platform in Riyquest. But what's cool, ever since then, I've been doing nothing but R&D. I build all kind of internal products and do all kind of really cool stuff. Uh I got the best job. Uh, y'all ever seen Q on James James Bond? That's kind of like me and my crew at at at work. We just build build stuff all the time. So, my

job, Joe Parlo is our uh uh CTO there, and he says he wants my team, we want to be two years ahead of whatever's going on. And that's that's what pushes us and drives us. We're we're so so we're like I I legitimately believe we're at least a year or so above most people in this AI thing. So, this is like chat roulette for This is like slide roulette. Y'all remember that at Defcon? I don't know what's coming up next. Oh, this is cool. Watch this. This actually This is kind of small for some of y'all, but this is the first This is one of my first conversations with Chat GPT. There was this movie that I was

interested in that I was like, I want to find this AI movie and it told me the AI movie. So I said, "Hey, I saw this movie in the 80s when I was a kid. I don't remember the name. This is what it was about." And it told me the movie name. And so the name of the movie is called Silent Running. Anybody seen that movie? One, two. Man, that's not a lot of people seen that movie. I'm We're dating ourselves. I'm not going to ask any more uh boomer type questions. Like Like So what's interesting about that movie? That movie was about a robot and there was a a guy that was like a

botist in space and and I remember this movie vividly because I was like three or four or something when I saw it and it was the first movie I cried at, right? I cried the older I get I cried a lot of movies now so it's not a big deal anymore. I used to like man I feel like it's one movie I remember I can't I can't remember the movie but it was a character that's like you know telling a a tear to crawl back up in there. I try to I but what's cool about this is like man this actually convinced me this is crazy man I cannot believe it knew this movie because you know how you ask

friends what the movie is and I'm what's that movie this so this actually converted me you see that was uh GTP 3.5 so that was a long time ago so that convinced me wow that's pretty crazy this is a crazy nobody knows this movie and out of this whole crowd this is probably you know a couple maybe a couple hundred people in here. I think three hands were raised. That's obscure information. And that's what's cool about this whole AI thing. That's the movie poster. I went to Amazon, downloaded it, watched it again. So, this is my favorite movie of all time. How many War War Games fans are here? Hey, this is definitely the goat hacker

movie. I don't care what nobody say. Like, this is the goat. So, this movie made me want to do stuff. And what's cool is this movie had some AI in it, right? You know, it had Whopper in there and he was playing, you know, games with Whopper. Whopper was that's the that was the first Beastboat AI I ever seen in Hollywood. Again, this was an inspiration for me to get into this AI stuff, right? And I I'm I'm serious. I think we're kind of like living in this era of what in my mind, you know, I'm I I turned 50 a couple of months ago. I mean, about a couple of weeks ago, I turned 50, right? And uh

I'm like, man, I've been waiting for since since the third grade. Shall we play a game? Quotable. Quotable. So, sometimes when I'm talking to Chad, it depends on what agent I'm talking to. That kind of reminds me of war games. That's the That's the feeling I get. How many can relate to that? All right. So, this guy is super dang smart. I bet you nobody know who this guy is. What the what the important part is this is Jeff Hawkins. Has anybody read this book? I'm going to give you three books. Now, this book is absolutely amazing. I'm going to give you my takeaway from this book, but I think you should check these books out if you can.

So in this book, what Jeff Hawkins talks about is this guy is really smart. He he invented the Palm Pilot and he's also a neuroscientist. Now some of the young kids in here are like, "What the hell is a Palm Pilot? Like y'all making me feel old." So Jeff Hawkins, a computer scientist, inventor, and he's also I think neuroscientist. This book is all about how he compares the brain to AI and machine learning. It's a fascinating read because you're going to learn neuroscience in this book, how the brain works, and you're going to learn how how machine learning models work as well. It's a fascinating read. And throughout the book, he's talking about are we

going to get to AGI, right? AGI is the big thing everybody's scared of, right? So, is it is are machines going to take over the world? Whatever, right? I don't believe that's going to happen. So, this book right here absolutely destroyed most AI stuff. It's also a really good history of AI as well. So, he goes through he goes through like three or four AI booms and busts. And why that's important to know is because I believe that we may be history may be repeating itself in some sense right now. So this is a great read. Again, it talks about the the crocodile brain and all this this neuroscience stuff and and how the synapses work and how your brain folds

and all that stuff. It makes a brother like me from the hood feel pretty smart when I read that stuff. Excellent book. Check it out. The the big takeaway from this is he he doesn't believe that one machine will rule them all. He says that you can build an intelligent machine, but artificial intelligence is mostly rubbish to use a British term. This guy Byron Ree, this guy has this this book right here. It might be the best AI book I've read in my life. This is what I'm on. This is the kind of time I'm on right now. This book is a more recent book, The Fourth Age. It's talking about how we had these

different ages. We had stone ages. We had the we had we had the stone age, the iron age, thinking like the industrial age, and now he's like we're in the fourth age because all about robots and all that stuff. He talks a little bit about AI, but this book even predates Open AI and all that stuff and chat GPT. It didn't have this stuff in there, but this stuff is so good. is is super pertinent to what we where we're doing today. What he talks about is is mostly what most people are scared of in cyber security. Many people that are here, they say they're they're scared that that AI is going to take their job.

This book talks about that throughout history, technology in general has taken jobs. There used to be people on the busy streets of certain places holding torches, stuff like that. There there used to didn't be a light bulb, but now there's a light bulb. So So whoever was selling oil was out of luck, but then better jobs always emerge. Like there's going to be engineering jobs and stuff that emerge. And like even with AI, what we're seeing here is we're seeing a whole different field of people that are doing new pin testing. people are trying to essentially um hack hack into hack into systems by doing prompt engineering or prompt injection. Right now there's a whole different field. There's prompt

engineering jobs where people get paid to do prompt engineering to create prompts. That's a job and that that job can pay six figures, funny enough. It's crazy. So I believe that AI is going to take the crappy jobs away. And even in in where we where we work at, we're trying to eliminate tier one and tier tier two like from a incident response perspective and sock perspective, but it's allowing our analysts to do better work. They're going to get better. They're going to be able to do like get into reverse engineering and better better things. So, we're not we're not getting rid of anybody. They're just going to be doing cooler stuff. And as for any company and most of y'all

work for some kind of cyber security company or you're your own consultant or you're doing whatever, you're going to be able to do more work. So, as a company like Reliquest, we're going to be able to serve more customers and we're going to be able to serve more customers better. And that's awesome. That's a win. Highly recommend. This is probably the best book I've ever read on like technology and like the history of technology. And it it brings you all the way up to AI, but like I said, chat GPT and all this stuff didn't even exist when this book came out either. Amazing book. And this is my third recommendation. Again, I'm just I'm just

going to give y'all some training material. You know people talking about training the AI. This is train. This is AI. This actual intelligence right here. This dude Pedro Dominguez. Has anybody read any of these books? Some of them couple of them. This book the master algorithm. Amazing book. So my takeaway again don't believe anything I'm saying. I think y'all should read these resources for yourself. But this is a good uh primer pack for if you want to be a beast in AI. So this book the master algorithm why this book is important is he w he walks through the hypothesis I think he works at the University of Washington or something like that but he's like he's a beast at machine

learning this guy's a beast he walk you through and he explains even like old uh models like he brings you through through AI and machine learning and then what he he he proposes like you hear people all the time I'm going to build this agent this agent's going to do everything I'm gonna build this thing blah blah blah. People talk about that all the time, but he kind of dispels that notion as well. He kind of says that the master algorithm is actually going to be a combination of different things. And why that relates to now is people are trying to build AI agents and many agents and all kind of different agents. But those agents, I don't I don't

believe that it's going to be one agent to rule them all. I think it's going to be a combination of things and we see things like MCPs come out. How many people have played around MCPs yet? Couple of people. So, I believe that it's going to be a bunch of different agents to do a bunch of different things that's going to be able to be that master agent, per se. This book is super good. Like I said, for from a cyber security perspective, I like to do the research because if you're going to be doing web application testing, you're going to be building systems, you're going to be doing all that stuff. We need to be super aware of all the things

related to this stuff. We need to we need to be able to go into a room and speak intelligently on on these topics. And I'm going to give you you some resources in a little bit here so you can train your own models and things of that nature, too. I think we need to we need to have a little bit of primer and then we can talk about this in an intelligent manner instead of just freaking out and being afraid of it. I think this is the best thing that's happened in my my life so far. Don't tell my wife that. This is an amazing movie. How many people have seen this movie? It's kind

of newer. Have to If you haven't watched it, you got to check it out. like all of the social things that people talk about. All of the things they talk about, people talk about how AI is going to kill everybody, how how and it it actually uses real like machine learning terminology in here. I'm like, whoever's the consultant for this movie was, he got it right. You know, sometimes you see a movie like, ah, none of that stuff is right. But what this movie does well is it it it addresses all the ethical issues, a lot of ethical issues, and you're going to learn it. Great movie. Love this movie. Next movie, same thing. All the ethical issues.

And when this movie came out, it was like, "Oh, this is total crazy, right?" But you hear people falling in love with agents and stuff now. This is happening for real. That's crazy, ain't it? This movie came out like 10 years ago or something. The other one about 10 years ago, but now this stuff is like coming becoming like real. So watch these movies because it addresses a lot of these things. So a lot of the talking points I hear sometimes I say, "Hey man, you should just check that movie out, bro. Check that movie out." And it addresses many of those things. All right, so that's my three movie recommendations. I don't know what's

coming up yet, but I don't know what's coming up next. Let's go. All right, cool. Open AI. The reason I I I have respect for OpenAI, I know some people hate it open AI and there's all kind of different things, but what they did from a development perspective is they started the whole uh model as a service thing to me. Before this, there was a couple of um there was some image stuff that you could do on Amazon. Uh, you could upload images, you could you could train some images, but you kind of had to be a little bit technical to do it. What I appreciate about OpenAI is they made they made machine learning and AI

available to the masses. Totally crushed it. Anybody can do it. Even like pick, you know, have a little cell phone, you could talk to chat GTP with it. They also made the APIs super easy. And matter of fact, how you interact with most of these LLMs now, these large language bottles, is by the standard that OpenAI created. Everybody's copying them. I I I love what I love what they're doing as far as that goes. And they're pushing the the limits. Some people hate it. I'm a technologist. I love it. I can't lie. And anthropic to me is the second big player in the space. They're approaching it a little bit different. How many people use Claude in here?

Couple of cloud users. How many people use OpenAI in here? So Claude, check out Claude if you haven't checked it out. It does it it operates on a different it's it's different, but I believe and I'm going to show you later. I'm going to make y'all laugh because I try every model. And that's something we do at at Reliquest. Uh, we built an internal system that I'll show you and we we love playing with every model at RQ and I'll show you some of the stuff that we're doing in in our mindset there. All right. So, definitely I I love this image. I didn't know that they had these lab coats on here, but I got I got a little This is AI

generated. a little AI generated brother looking like a looking like a young good-look me in the lab. So I encourage everybody like what I'm going to encourage you now is to like Okay, cool. You use chat, you use Claude or whatever. I want y'all to experiment by using the models to build your own applications because anything you do, you're going to be able to put that on steroids essentially with AI. I'll tell you my one of my NSA stories, right? Um when I was at NSA, I've always in my mind I thought about automating stuff. I used to I used to do Pearl back in the day. How many Pearl mongers in here? And and they they said

if they test too they like Pearl, right? So Pearl was a a it's kind of like the predecessor for Python. And Pearl, you could whip out some scripts, you could parse some logs. It was crazy. It was really good. And so at NSA, I was on uh what at the I was on the IP team at the time. It was it was the it was the global network engineering team at NSA. And I was an enlisted uh Navy guy, but I was just a nerd. So I got my CCNP back in the day. And so with that with that what I did is I started taking all the the configurations and I would use Pearl to do all the configurations for

me. And um we had we would have to cut over legacy systems. But I said, "Oh, we can take the legacy configs and then we can convert them into new to to the new IOS commands." For all y'all Cisco nerds, people know what I'm preaching to some people in here. Some somebody heard me. So, so my pro script was able to migrate all these new devices over, but there was people that work with me that didn't want to use the scripts. And they was like, "Oh, the script might be wrong." I'm like, "Bro, you have a better chance to be wrong than the script." And that's 100% certain with many things that we do daytoday. I actually trust

machines and AI more than I trust humans. And I I mean that that's probably a hot take, but that's the way I feel about it. I trust coding and humans more. So I want to encourage everybody to experiment. You're going to break some stuff. Some some stuff's going to happen, but that's that's life. You could break stuff anyway. That's that's how I feel it. So we shouldn't be scared to move forward and break stuff. We also have to work, we have to also think about security and the security implementations of it. But did y'all know, have y'all looked at the OAS uh guide on LLMs? Have y'all looked at that? Man, I got like two or three people up

in this congregation with that. Yeah. So, check out check out uh OASP guidance on LLMs. They got a top 10 for LLMs now, and it's pretty good. But the big the big news is we don't have to change what we do so much. We kind of like freak out many times when new things come out. We like freak out and and we don't know what to do. So, if you're a consultant, you're doing web app testing or you're doing some kind of compliance, just refer to refer to the dang OAS LLM uh guidance because another reason why I I have all these I I I say all these things, but that's why I bring in the authors, the experts, and

all that stuff because it's just not me. It's like people that are way smarter than me have said these things. Hallucinations. This is crazy. What if you woke up and saw something like that in your face? That'd be that'd be crazy. You probably own something. You probably hallucinating. So, this is one of the big things that we we we that's talked about all the time. Uh the the newer the models are, the less they hallucinate. But when I when I tell people all the time, I say, "Look, man, how many times has somebody told you something that wasn't true?" or you saw a pin test report, but that did not exist in your environment. It

could have been a typo, but I mean, is that a hallucination? Right? So, what I what what we found is the the the more you limit the AI on on giving a specific task. Remember, don't try to build a master algorithm or master agent. Build an agent for a particular task. Give it tons of context and you can avoid hallucinations. Another thing you could also do is you can build agents to check the other agents. This is one of the ways that essentially what OpenAI and all these other people doing. They're building AIs to proof other other things. They're building agents to check other agents. And that's something we need y'all y'all could implement in your your uh environment.

Also, the more context the better for sure. And that's going to prevent hallucinations or from happening as much. Context is king. So context the context is king. So what we do uh we build these agents and like say a ticket comes in right it comes into service now. Well we have tooling to go to service now and pull the ticket from service now. if there's an IP address there, it can automatically go out and it can pull the context from virus total or whatever or anything like that. So that's what we're doing. We're building tooling around our our LLMs or around this technology to go out and grab all the important contexts and what that

what that allows you to do and I'm going to tell you I'm going to show you some tools and some code stuff that you can play with like lang chain. How many lang chain people in here? couple check y'all got to check out lang chain because something like lang chain allows you to build tools you can build a virus total tool again uh IP domain lookups so if there is an API you can build a tool for that and and pull in more context now in some cases we had we we we've we've shortened down the response time where it would take somebody 15 20 an hour to do all the lookups AI does that

in a of I mean in without a problem because it's just doing the AI calls and then doing the the things and sometimes you kind of kind of hack the system. How we initially hack the system is my team uh that that I'm I'm with. We have a a guy that y'all probably know named Jonathan Etchavaria. He's a Tampa guy. He's probably talked here a couple times. So Jonathan and I, we built a browser plugin that actually did DOM DOM DOM manipulation. I'm a Texan. That's a big word for Texan. So DOM manipulation to pull data from the DOM and then inject that into an LLM and then do reporting. We did that probably like three months

after 3 point uh uh chat GTP I mean well the 3.5 model dropped and we knew instantly like oh man if we just pull data from the APIs with this same technique we're going to be able to get better reporting and that's going to free up time for our analysts. So, it's all about bringing in context. Some people uh you probably heard of retrieval augmented generation rag. That's the terminology that that people use for it. But rag could be a number of things. And I'm going to show you how I built I built a a digital version of myself uh with using some of these techniques. So, rag could be API or rag could be you you're giving it data in

some other way. It could just be looking it could be pulling information from a file or something. There is no silver bullet. Again, if anybody says that this one thing does everything, they probably lying. They are probably lying. So, a couple of things. What what makes this these new models so good? The new models actually improved because of human feedback. I don't think it's a AI versus human thing or these these machines versus the human. I think in every step the the the humans actually make the AI better. And I need to do time check. How long am I supposed to be talking? Somebody tell me cuz I can go forever. Y'all can probably tell.

I'm going to I'm going to save some time for some questions. So hum human feedback. What happened is the the AI would give give some some text and then what you would do is you would the human would say the text look good or not? Then the human would say they would give some more text. Does that look good? Does this look good? And then so what happened over time it actually made the machine better? And then over time they built agents to actually imitate the human thing to upvote. So that's how we got these really awesome models now. Bias. Everybody has bias. So I mean I know I'm I'm from Texas. I'm biased by

Texas barbecue and brisket. So my thing is I want everybody to start somewhere. Now, when I say start somewhere, you can't just idly sit by and wait for wait for it not to happen because this train is moving and AI is here to stay. I got a couple more resources that I want y'all to check out here that's coming up here shortly. So, here's some tools. Has anybody used Olama in here? We got a couple people. So, Olama is pretty awesome because Olama you can actually run your own models locally. So you're not submitting your data to OpenAI or any of these other things. You can actually run some pretty cool models um with Olama. So I I highly recommend

you check that out. Anybody heard of Kaggle? Anybody use Kaggle? Couple of people. This is the best way at home in my opinion to learn how to create your own models. Now those models are going to be something like uh one of the couple of the introductory models that you'll create. There's like a manifest of the Titanic on there. And with that manifest, you do you do a machine and the manifest has their ticket, how much they paid for the ticket or whether they're a woman, the age, all that stuff. And it and it help and you write a model to predict if they would survive the Titanic uh crash. That's an awesome model, right? Because

what's crazy about it, you learn some so it's like a social experiment, too. The people in first class were more likely to survive. A woman was more likely to survive. But what's cool about it is you get to build a model in Kaggle. It's one of the first models you'll build and you get to you get to build that model. Building a model is important for you to understand how these models work and how the models are trained. Another one is like a a ice cream experiment is one of the first machine learning things you'll do is you'll predict how hot it is and how much ice cream the ice cream man is going to sell. Those are two pretty cool

things and Kaggle is free and you can learn machine learning there. But I think that you don't need to you don't need to do Kaggle to understand how chat GTP I mean you don't to use chat but I think the more you understand about the underlines of how these models work and building your own models you're going to pick up it's going to be beneficial to you and then when you're talking to ML team if you're a cyber security person you're going to understand what's popping hug face all right cool now I'm good I'm in the zone now I have four minutes period Three. You got to love it. Hugging face. Check this out. Hugging

face has all these open source models. How many people have hugging face accounts? You got to check it out. Lang chain. You have to check this out. This is pretty cool. There's a couple other competitors, but lang chain allows you to build tooling to build really complicated uh applications and all kind of different agents. Cursor. Anybody use cursor? I love me some cursor. I mean, if people if you know, you know. Cursor allows you to essentially what they call vibe code. Now, that's what people call vibe coding. But but it actually you can give you can say, "Hey, cursor, I want to build this application and I want it to do all this stuff." And you like you

chat to it and it's like and I I always do it like this. I'll tell you my workflow. Cursor. I want to build an application that does this, this, and third. It's a AI application. I want to use open AI and I what I want you to do is I want it to be this that and the third. So you always give it a complete outcome of what you want done. And then you say make a plan cursor. You're chatting to this thing. It's a it's a it's like VS code. It will make a plan and then like if you like the plan you like code it and it will code up. It will it's it's witchcraft.

It's crazy. Like it's nuts. This is like taking one engineer and turn it into like 20 engineers. 100%. I could not recommend this more. So, if you probably use a co-pilot, co-pilot is cool, too. And C-pilot is actually trying to copy off cursor cuz cursor is cursor is ridiculous. You got to check this out. MCPS. MCPS are a way to build an agent that advertises itself to the to the model. And this is actually maybe from a movie, too. Anybody know that reference? Yeah, that's my that's what's that's my second favorite movie. Even though I'm a Tron Legacy fan, I can't lie. That's that's blasphemy. So, this is funny. So, I want y'all to experiment thing. I got like one minute.

All right. So, anybodyo? You can make music AI music with I made a whole YouTube channel with AI. Me and my wife have a YouTube channel. We make music and we do videos for kids. I created a I created a group NCD. Check out the music on Spotify. It's all AI. I created a book spot the spot the wolf AI generated uh Where's Waldo? I created all my writings, all the all the speaks, all the talks I ever did into an agent and I can just ask the agent, hey, write me a talk.

>> All right, I'm almost done. So, internally, we built our own chat GTP thing internally at RQ. We also built a pipeline that you can actually test and AB test all the different models. You don't have to know it all. Don't wait to start. No one is going to No one is going to anoint you. Nobody's going to make you an expert. You got to go out there and get it. Leave it all on the field. When I die, I done did it all. I'm leaving on the field. Focus on outcomes. We're almost done. GI Joe. Now you know. Now you know. And let's go. That's it. [Applause] Heat. Heat.