I ❤️ AI Q&A: Responsible AI Practices Explained | How to Start with AI + What Is AI Red Teaming?

Hello and welcome! We are here. Happy Friday, Fri-yay, as it's known to me and my friends. Good to see you all. Welcome! I meant to do the whole lucky lip gloss before I started, but we'll fix it in post. Um, nice to see you all. Um, I, uh, really great to see you. Oh, anyway, nice to you. Thank you for being here. Welcome, welcome. We have so much cool stuff to talk about today. Um, most importantly, last week we talked a little bit about, um, some of the, uh, the fact that we've got a— I got my book that just came out and it's doing really well, climbing the ranks. I saw it kind of debut on the bestseller list, uh, about on like 92 or something, and then it went to 50 and now it's at 36. So I'm just like watching it do its little thing. Um, so, uh, if you haven't had a chance to grab it yet, not— no better time. Um, it is available everywhere books are sold, but my My biggest, most fun moment this week was heading over to Barnes Noble in my local community and finding it in the New and Notable section, or New and Noteworthy section, which was really cool. So great, so great. Like, I don't know if you can tell, I took a picture and I was like a little kid. And it was such a cool thing because I, um, I don't know how many of you know this, but I like, I graduated from the University of Barnes Noble. I mean, it's where I learned every skill. In my earliest part of my career. I would just go to the bookstore and actually full circle moment and kind of sad is when I went to Barnes Noble as a kid or as a, you know, young adult learning. I mean, the first book I picked up was TCP/IP to learn how to, you know, learn about internet networking. This was like '96 or something or '98. Anyway, the first one of the books, I think I still have it somewhere. Every time we move, I'm like, Don't throw that book away, very important. But one of the things, anyway, so I, I would get these books, pull them off the shelf and learn from them. But when I would go to the tech section back then, it was like a wall of technology. I mean, it was everything from like networking. Of course, cloud wasn't, you know, as cloud became a thing, it became networking, programming languages, cloud technology, you know, all the certification, security, like all it was just. It was like my home, you know? I was like, I love it here. So I would go every weekend and spend time in the book section. And I would eventually, as I got more and more successful in my career, I'd go to Barnes Noble and I would just, I wouldn't be able to leave without spending $500 on books. But mind you, these tech books were crazy. They're like $75, $140 for a book that only later did I find it became obsolete the second I bought it. Like all the screenshots would be old. Anyway, publishing and all that, it's tough. But, um, but yeah, it was really cool. So I go to my bookstore and I'm like, surely it's in the technology section. That was where my mind was at. I'm like, surely they would've put one, like stuck it in the corner of the technology section. And so I walk into the bookstore and I go to the technology section and it is not a wall. It is literally like one section of a bookcase, not even the whole bookcase, just one little section. It was. A little sad. So we need to write some more tech books, y'all. And I don't mean textbooks, I mean tech books. Books that would help someone like me who's not going to go and sit through a 6-month or 6-year program. Um, I mean, goodness, just step by step. I love these books. I don't know if I have one handy here. Um, I loved these books because they were like lab guides, so they were intentionally written in this kind of like do this, then this, like 1, 2, 3, 4, 5, and you would like check them off as you went. It was really, really good. Anyway, um, yeah, so, so it was cool that I went back and I went and looked at the tech section. I was kind of sad that it was such a small section, um, and then I was like, do you even have my book? They're like, oh, it's in the New and Notable section. And I was like, with, with all of these other tech greats, like the Nexus book and like You know, uh, Elon Musk's book is there anyway. Um, but it was very cool. It was very cool to be seated amongst— they didn't have a little placard for me that they had for some others, but it's early. It's early yet. I have some goals. I have some goals. So, uh, so thanks to everybody who's gotten one so far. I appreciate all the feedback. I've had a couple people who did, uh, end up going through the book and, and finding a lot of value and telling me how much they really appreciated my perspective. I caught my first typo. Maybe you all could be on the hunt for it. Um, don't tell me if you find typos. It breaks my heart because I work so hard. But, uh, but it does prove that it was written by a human because if anything an AI-generated, uh, AI writer will do is spell stuff right. Um, and have grammar correct. So there were certain sections, there was a section where I was like, oh my gosh, I forgot the word. Like I forgot one word in that sentence. When you read it fast, you don't realize it, but then when you read it slow, you're like, wait, there's a, There's a word missing. Anyway, I like the evidence of it being real. Um, oh good, ARD. Oh good, good, good. I, I can't wait. Um, there is such a— 100%. I have my book right here. Look, I like exactly what you said, like putting little, little dog ears in it. Speaking of which, why don't we do this right now? I'm gonna go through and see what we're gonna talk about today. Let's just pick a topic, right? Once everyone has one, we'll go through it. Word. Oh, maybe we should start at the beginning. Maybe we should. Last week I think we covered, um, something that was, you know, really important to me, which is this concept of lamplighter leadership. Y'all remember? Um, and, and I have a podcast which has a few episodes. We put new ones out all the time. I also, uh, since last time we spoke, revamped my existing podcast, Good Morning AI. Um, and so that is up and running. I haven't published my Greg Brockman or Cynthia Stoddard interviews yet, but those will go up this week. I did find them, um, and downloaded them, so I'll put those up this week. I also interviewed some really cool people last year during our AI challenge. I interviewed Erin Reddick, who's the— a cultural futurist, um, really concerned about kind of the cultural integrity, uh, that needs to be preserved as we build systems that amplify and automate stereotypes. So I really, really appreciate the work that she does. Um, and then I also interviewed, um, uh, Renee Cummings, who's one of my favorite people. She— if you don't know her, uh, you can look her up. Tell her I sent you. Um, she'll be like, "Don't send people!" Uh, but no, she's super cool. She's very, um— she's had a very, uh, dynamic life. Uh, she's been in broadcasting, she's been all in all sorts of areas, but where she found her passion was in criminology. Some of you know Renee. Um, She found her passion in criminology and, uh, then from there realized that criminology plus AI could be very dangerous. And she became what is known as an AI activist, which is the first time I had heard it when she had said it. And I was like, oh, it's so good. She also introduced me to the term that I personally have come to know and love, uh, called AI Jedi, which we will definitely be having an AI Jedi pro— like a, uh, a, a certification or badge or something where you get like lightsaber. I don't know, I'll, I'll make something work for that. But AI Jedi stands for Justice, Equity, Diversity, and Inclusion. But let's just face it, being a Jedi is pretty cool, so maybe people will still wanna be one when they grow up. But she introduced me to these concepts and how important it was to really be intentional in the development of these systems. So really, really like it. Yes, of course. Of course. I will definitely be talking about the, uh, tech AI Tech Fest. Uh, well, AI Skills Fest. Test that's happening right now and an opportunity that you have to win a certification. And I'm going to be extending my own kind of sweepstakes of sorts. So I will be letting you know a little bit about that. It can't— you're right, it is— it can be overwhelming. Oh yeah, of course I will definitely do that. Yes, AI Jedi, what? Oh, okay. I feel like my comments might be delayed. So if I say something and you said it a long time ago, just Work with me. I'm doing the best I can with the tech as it is. OK, cool. So let me see here. So some of the things that I wanted to share with you, obviously in the book I'm going to cover something. I got a chance just to give you a sense of my week. I got a chance to work to speak at the Privacy Pro Summit where last year the Privacy Pro Summit was for privacy professionals. Um, but those professionals, those privacy professionals last year focused really on the work that they were doing, like the type of work, how to do things differently, like the integrity of the work. This year they wanted to shift that focus to not just what they do, but who they were. Like, who are you as a privacy professional? What do you share in common with your peers? Like what makes you dedicated to the craft of privacy? And so I was honored to be part of that because they saw in me that I like to share why I do what I do and my personal story. So I got a chance to do that. I made about 100 new friends. So that was really fun. I'm really close to 40,000 followers on LinkedIn. Um, I love, I, I care about that number because I actually have met or at least spoken to every person, um, on that list for the most part. I'd say 90% of those people I actually know. And so the more people I know, the more people I feel like I can reach with the messages that are in the book, but also that I speak all the time about how important it is not only to take take advantage of artificial intelligence right now and the capabilities that it offers you, your business, your team, but also that it is our responsibility to do that safely and responsibly. That we are not gonna wait for government regulation. We're not gonna wait for policies to be defined. That building responsibly is, it's like thread woven into a fabric of a project. It is woven in. It is not something you sprinkle on at the end or get funding for at the beginning. Like it is, It is how we think, it's how we build, and we don't deploy AI systems without these guardrails and frameworks in place. So I have a couple really cool product announcements that have happened this week that are in conjunction with Microsoft's AI Skills Fest. So for those of you who are not aware, the Microsoft AI Skills Fest is alive and well and in progress. Many of us, uh, got a badge stating that we participated in a Guinness World Book Record attempt to have the most people learning something at one time online. There, there were a few opportunities. Um, and I saw a few members of the community post their badges. So congratulations for those of you who were part of those attempts. Um, and the Skills Fest, it's 50 days of content. And so what I decided to do, as I do, um, is I decided to take a few of those topics and repeat them. A little bit more slowly, intentionally, and give you an opportunity to actually learn with me, have like expert support, if you will. In the I Love AI community, we are going to just, you know, share our experience in that Skills Fest. It's free, anyone can go, we could talk about it. But in the AI Leadership Institute, in our premium membership, the upgrade, if you will, we're gonna actually spend time building and, and do some, I mean, some of these labs, I mean, they weren't really labs, they like were walkthroughs of people doing stuff. And I was like, wait, wait, but we, if I wanted to do that, how would, okay. Okay. Bye. And then it was done. So they were really, really fast. Exciting though. Very cool. But I am a believer in learning by doing and by putting your fingers on a keyboard and building it yourself. There is so much power and capability in that. And so we're gonna be doing some more of those sessions in the AI Leadership Institute. We'll offer them, pay by, you know, you can pay to play. But the reason why is because it's actually on our corporate tenant. You get access to that tenant and it requires resources and a team to help support you. So if you do that, obviously there is a cost to it, but it's negligible for most of you. One of the other things, we have two really interesting— let me think what I want to focus because I get all excited. So the Skills Fest, let's start there and then my product announcements. So I am very— of course I'm a Microsoft MVP. As many of you know, and I, uh, get access to working with the product teams at Microsoft, specifically in Microsoft AI and Trustworthy AI. I get to work with these teams, give them my feedback, give them my vision for what I wish these products would do, and they get to look at me and kind of pat me on the head and be like, okay, thank you. We'll see what we could do. Uh, I remember one of the product managers being like, ah, that's pretty ambitious. I'm like, hey, I don't know any other way. It's actually on my shirt. These are the Lamplighter principles, by the way. Today's focus— let's see if we could share— is resilience. So that's kind of cool. But every day I have 12 of them, of course, one for each principle, and every— I might have 24 of them, but depending on which one I grab from the thing, it has one of those principles highlighted, which is kind of nice. All right, let me— I might start sharing some links and resources with you, so I want to make sure that I am Putting in the right place because last time I put them all in the hosts and panelists and none of you were in there. One of the other things we are going to be doing in the AI Leadership Institute starting next week is that we are going to be starting to do small group networking. So it's actually not even small group, it's going to be one-on-one speed networking on the call so you can learn who is, you know, we have a lot of consultancies in this organization. We have a lot of people who are working in companies. We have I don't wanna call them headhunters, executive search firms that are part of, uh, our community. We have, um, certainly hiring managers. And so it's a chance to just, and I, and I was talking to someone, it was so important for us to be like face to face and get to know each other as real people. Like, it's one thing, obviously I'm in doing this this way, um, for a reason, but it's really nice when we can get together and get to know each other. So I wanna add that as a capability for us, as an opportunity for us to add additional value to each other's lives, that it's not just about like Noelle talking, though I do talk every Friday, but our community is really about sharing and growing together. And there are so many leaders in our community that are worth getting to know. Yeah, so my boot camp, I'm like, when did I do that? Oh, was that like my AI boot camp? That was awesome. So we, as a way to kick off the AI Skills Fest for Microsoft, we did a global AI boot camp. It was really, really fun. We had 100 and something, 120 people, and it was virtual. Virtual and in person. I gave t-shirts to people who came in person. Here is a fun fact for all of you. I, I don't know if you know, but we have a map feature in our community. So when you go to I Love AI, you can see there's an option called map, and you can see we are quite well represented across the globe. And so what I will be doing is as I go to a city, for example, next week I will be in Boston at MIT speaking at the— I some, some Idea AI session. I'll be publishing it in the school, but I'll be speaking at MIT on Tuesday. And so on Monday night or Monday afternoon, I'm going to invite anyone who's in Boston to come hang out with me at the local Barnes Noble and either get the book you have signed, get a new book and get it signed, or just hang out with me in the cafe, which I love that idea because I— it's really cool. I've been I've been in the Barnes Noble cafe for decades, so I'm really excited to, to now be there in a capacity where I can meet some amazing people from— what? What? Don, you are not in Boston. What? That's Monday. Monday, my friend. We are meeting in person, taking a picture. It's happening Monday. What? Okay, cool. I will ping you directly too, so you know. What? How come I didn't know you were in Boston? How did I not know that? Anyway, some people I've known for years, they tell me where they're from and I'm like, wait, where are you from? I didn't know that about you. And they're like, oh, you knew. It's just not, not a detail that stuck over the years. So, uh, so cool. I'm glad I said it then. So anyone else who's in Boston, please feel free to, uh, say hello. Um, so my philosophy on the AI Skills Fest, if you sign up and register, and I'll pull up the link as well, if you, uh, register for the AI Skills Fest, Um, it is a 50-day journey, which means it's a lot of content, a lot of sessions. Um, they're again attempting a Guinness World Book of Records attempt at multiple different, like, learning and massive amounts of people learning a topic. Um, I should find out what that record is and try to beat it myself, make that a mission that I'm on. Um, all right, hold on. I can't do both. I can't type and talk, type one set of words and talk a different set of words at the exact same time. So here we go. I'm gonna go ahead and it's so fun. So I'm gonna share my screen so you can see what I'm looking at here. There you go. So this is the AI Skills Fest. You can register now. I'm gonna go ahead and grab the link so that those of you who are not part of it yet. Oh dear, no. Let me go ahead and throw that in the chat. So come— oh yeah, that's long. Anyway, if you click it, it should work. But that is the Skills Fest. You can register. I'm already registered, but you can see some of these really cool sessions that are around. I've attended a few of them already. Some of them are scheduled— I mean, they're scheduled all around the day, which is beautiful because depending on where you live in the world, there are sessions in your time zone. So they're follow the sun sessions. Sessions. So as the sun goes around to your part of the world, you will find sessions there. There are tech professionals, business professionals, students, and public sector. So they've given different pathways for people to look and discover sessions that are appropriate for them. I, of course, am all about agents. The agent sessions have been very good, really cool, very good demonstrations. So, so it's just a chance to really uncover some of these technologies that you haven't seen before. But as part of that, yesterday or maybe the day before yesterday, the Responsible AI team, the Trustworthy AI team at Microsoft, launched a brand new capability called the AI Red Teaming Agent. And it sits inside of, and maybe I will show you. I'll do this because I want to see if I can get them to— oh, look at all these sponsored ads. Um, there we go. So the AI red teaming agent— oh dear, it took me into like— oh, that's— did I hit close? Oh, I don't know. Hold on, let me try that again. AI. Let me do this. That's so weird. Okay. So the AI at least shares with you that I, I obviously had looked this up already. But the AI red teaming agent, it's in preview right now. It is included as part of Azure Content Safety. You can do the same thing I did and Google it or Perplexity it or whatever you choose to do. These previews are really important because as you all know, AI doesn't get better unless people like us actually use it and provide good feedback to the team. So this agent, its job is to do, and perform some of the automated testing that the red team at Microsoft, AI red team at Microsoft was doing prior to that. Now, I really like, one of the other features. So there's a bunch of features in Azure Content Safety, and I've talked about them before, but I'll mention them just in case you haven't heard of them before. But go take a look at it because when you deploy a model, let's just say you're by yourself, you have a small team, or you just don't have resources to invest in building an AI red team. An AI red team. For those of you not familiar with the term, it is that inclusive engineering organization that serves multiple purposes. One, the red teaming that comes with cybersecurity, that's like adversarial attacking, but then also another 3 levels. So there's adversarial attacks, then there's accuracy. How do you make sure the model is accurate, is telling the truth, is grounded in truth? There's fairness. How do you make sure that the model is fairly delivering, uh, similar responses to everyone regardless of who they are? Or if the dataset changes, that it doesn't change the ethic— the, the ethics, uh, or the, um, aspects of truth that are in those answers, right? So accuracy, fairness, and finally security. How are we ensuring that the model is secure from a perspective of, uh, common attacks that happen to LLMs specifically, like the jailbreak attacks or prompt injection or cross-prompt injection, these types of things. So all of these, someone in the company has to be thinking about how could this hurt someone, right? That's the whole baby tiger mentality. The baby tiger trainers, right? Us, uh, we need to be the ones in an organization saying, how are we managing truth or accuracy? How are we managing fairness? How are we managing security and prompt, you know, violations or attacks? How are we doing benign and adversarial attack testing? So what the cool thing is, Microsoft has been doing this for quite some time. They've documented all of this. So that's another thing I would encourage you to do is just look up red teaming. I'll give you the link to the document, but red teaming Microsoft. I of course have looked this up. Across all, like Microsoft, Amazon, and Google, and they've all, all got different things to say about it. But this is really interesting because it allows you to see— I always say, you know, this is what I, I always say, build something once, do it once so you know what it takes before you choose to automate it or hire out to have it done. Red teaming is no different. So I think this is true for building a model. Everyone should know how to build a model before they outsource or scale a model. And it's the same is true for red teams. You wanna know what a red team does before you automate it. At least be aware of it, the reason that it exists and why it does what it does. So this, I'll just send this over to you all. Here is another link. So if you want these links, you could just join us in the I Love AI community, which is pretty easy to do. Just type in my name, noelle.ai, go to that website and you'll see the ability to join us. And that way you get access to all the links that I'm sharing and you're live in the Zoom with me, which we get new people every week. So thank you all for being here. Um, so this goes through what a red team is, what are the practices, what do you do before, what do you do during, what do you do after, model development, model evaluation, model selection, model deployment, model monitoring, right? And so this was— oh, it recently was updated, so that's cool. Uh, it did not come out in 3.27. I've used this article over the— I mean, you can go back and look at some of my content in like 2018, and I was talking about the, the documentation that Microsoft has done, but they documented what their team did kind of the old-fashioned way. So in lieu of you doing it the old-fashioned way, read a document like this that tells you kind of what they're trying to do. That way, when you get access to an easy button like Azure Content Safety that will provide you a set of capabilities that will replace the human doing this work, you will find value there because you'll know what it takes to do it. And you'll, you'll see why having an agent do it is so powerful. Now, a couple of the things that Azure Content Safety focuses on, one of my favorites is groundedness detection, right? The ability to have a model grounded, of course, in truth. But before an answer ever hits a user, like I ask a question, how tall is the tallest tree in the world? When the model goes in and tries to find that answer and comes back, if it generated the answer, meaning it's not grounded in truth, The groundedness detection feature on Azure Content Safety will evaluate every answer that leaves the model for how grounded it is. And then you, as an expert, as a responsible AI practitioner, you can set the threshold for what you're comfortable with passing through to a user. If you're in healthcare or manufacturing or finance, you might not necessarily want a 60% variability in truth. You might want to make sure it's 98%, right? So you can set a threshold for how comfortable are you with something that deviates from the truth. I mean, if you're a scriptwriter or a copywriter, you probably aren't, you know, maybe you're willing to see a little bit more creativity, but if you are trying to do something that needs a lot of precision, this is your opportunity to set these thresholds. Oh my gosh, absolutely. Uh, thank you. Yes. So one of the questions, um, that just came in, thanks, Don, uh, is, The where to start. And so my perspective is the best thing to do, and it's actually very good segue, the best thing to do when you're getting started learning AI, number one is to learn the fundamentals, the basics. So in I Love AI, there is a masterclass. It's like 5 sessions and it teaches you the basic, like what is AI? What are the different types of, uh, AI, like supervised learning, unsupervised learning, um, general adversarial networks, you know, convolutional, like all of the, what are the different types? What are the most What are the most common reasons people are using AI? What are the most reasons people fail? Like, you wanna have a good grounded sense of what, what we're talking about. Then the next thing, once you've got that under your belt, the next thing I encourage people to do is actually build a system end to end. And so we built what's called the AI Bot Builder course. Many of you who purchased the book also got the Bot Builder course that I recorded last year. And this is you building a bot, a robot, um, building a robot that does a certain task. Now, what I think is valuable when we learn it, and, and this is a good segue into my favorite topics in the Skills Fest right now, are around agentic AI. And of course, I recently got an award that said I was the number one thought leader in agentic AI. So that was awesome. I was like, yay. But the challenge with that is that what agentic means is it is you creating systems that will autonomously talk to each other. And if you didn't take that time, remember I mentioned do it first, then automate. If you didn't take the time to build a system, an AI system end-to-end that's responsible and safe that you implemented guardrails for and implemented audits for, if you have not done that work, if you start automating bots that do not have this groundwork laid, those bots are going to become very unpredictable. Dare I say dangerous or Skynet, right? Like it's pretty dangerous. So we wanna make sure that when we're building any AI system that we start with a very specific, it's what many are calling small language models or fine-tuned models, like a very specific model that does a very specific thing. We turn that into an agent because it's so good at what it does, we can trust it. To work autonomously. But we don't just by default let every model we build, every agent or every robot we create become an agent. I— this is Noelle Russell's belief, and it's what I talk about in the book. Like, my belief is that agentic is an evolution of systems thinking. It's— you've built this system to do this thing. You've now measured its success. You— it operates with excellence. You understand how it works. It is documented. It is safe. It is responsible. Now it can become an agent. If you start building agents without understanding the integrity that's required to build a good bot, a good robot, a good robotic process, you're gonna end up scaling massively something that's quite immature. And I think we'll see over the next 3 years as those baby tigers become big tigers, it's gonna cause a really detrimental impact on society, on customer relationships, on all sorts of things.. But we can stop, we could save ourselves right now by just learning to build responsibly first. That's why I use this analogy, right? I always walk around with a baby tiger wherever I go. This isn't Bruiser. Bruiser I think is upstairs still with my stuff. But this is Eloise, and Eloise is my daily reminder whenever I pick her up or look at her of the idea of baby tigers in companies that when we get an AI system, when we adopt an AI model, that oftentimes we're very enamored by its capabilities. Like enamored. Oh my gosh, it's so cool. You'll see, I'm gonna demonstrate 3 tools for you today and you will be like, that's pretty cool. So you'll be enamored by it. However, what we wanna make sure we do is we wanna ask questions about this cute little enamoring technology. Like, wow, like how big are you going to be? Or what are you gonna eat? Or where are you gonna live? Or how long am I gonna keep you? What happens when I don't want you anymore? I used to always tell people, um, about this analogy of, you know, that you're one acquisition away from the Death Star. And as a data scientist, that rings a little different because it means like, as you build a technology, a, a, a data-driven AI system, if you're one acquisition away from the Death Star, what do you, what can you do to protect that system from ill intentions if you don't own where it's going? Um, I think I mentioned this last week, but like, you know, my favorite little tool, I, I hope it's, it's in the same drawer as my, uh, Amazon Fire Phone, which is sad, where all good tech goes to die. Um, but like the Humane AI pin, right? Super cool. Love this pin. I was such a fan. Uh, and then they decided to sell for parts, which is unfortunate, but what about my data? What about all of the information that I ingested into this in order to test out their content? Anyone think of another example of an amazing, really cool technology organization that is now selling for parts on the open internet? Yes, you and I are this. Yes, 23andMe. Like, you would, you have to build responsible innovative, but responsible practices in— you have to weave it into the fabric of the product so that when it is acquired, which we are watching happen all the time, someone can't just pluck it out. That's why I always say, you know, responsible development, it is not a raisin in a bun or a, a chip, a chocolate chip in a muffin. It's like water in a wave. It's like a thread in a fabric. It's woven in so tight. In fact, you cannot take it out without ruining it. And that should be the idea. You cannot remove responsible practices without ruining the entire thing, because if you can, that thing can now be used for a purpose outside of its intention and maybe for a purpose that isn't good. So this is what I think about and I do every day is I kind of contemplate how do we weave responsible practices into the fabric of every project that we build? And I do mean every project from how are you triaging your emails to how are you sending reports to your customers to how are you acquiring customers, how are you managing your pipeline? Like all the things. Everything is an AI solution waiting to happen if it's not already impacted by AI. And my goal is when I see that solution opportunity that I weave in these practices. So for those of you not familiar, what are those practices? Like I said, uh, thanks Don for the question. We do have an AI, a free AI masterclass that's inside the I Love AI community. So you can go through that masterclass. Class and get a sense of kind of the general, the basics. The second class though is how to build, you know, an agent. It's an agent workshop that allows you to build a single agent that is a customer service agent. Um, but a single customer service agent and you're training it the old-fashioned way through system prompt engineering. And so you'll be prompt engineering through the system prompt to give it instructions. And at the end of that exercise, we have what's called a break your bot section. Session. And so everyone throws their URLs to their bots into the community and we all try to break each other's bot. And that gives you such insight into what it takes to build an AI system from the thoughtfulness that goes into the intention of that machine, um, who it's for, what it does, what data it uses to get in, in a good, uh, a good answer. We don't talk specifically about retrieval augmented generation. That's a more advanced course, but we do talk about how you can control the integrity of the dataset from inside the system prompt. And that as you evolve, you would add a, you would augment the retrieval of an LLM from retrieving from its training data, the internet, to retrieving from a source you define. But the only way to do that is to ensure that the source you've defined is really good. So, so these are like very important ideas that even in the moment when you're super excited about your baby Tiger and you're like, oh my gosh, did you see Sora did blah, blah, blah, blah, blah. It is really important to think about like, where's my data going? Who owns it? Can I get it back? Can I turn it off? Can I, you know, how big is this gonna be? How many people are gonna get access to this? These are questions we ask about ourselves, but those questions get even more elaborate when we ask it about our organization or about a system that's gonna serve our customers. I, I often encourage clients, so I have, a couple different types of businesses that I run. One is a marketing agency, and for those clients that I have that are, I'm helping them market their business, uh, leveraging AI to create content, leveraging AI to communicate and follow up with their prospects. Like when we do that, it's really important to think about, you know, like how brand voice, how, what is the human to do in the, collection of activities and be very intentional about the human and AI experience. And so also in the book, but also in our new book that some of the members of the AI Leadership Institute and I are writing, in our, in this new book, we are talking, we have a whole chapter dedicated to human AI experience and the power of building, uh, with what's called HAX or H-A-X. And there is a toolkit available if you are not familiar with it called the Hacks Toolkit, which I've searched so many times. And so I'll just give you a link. I actually think there's a very short link for this, like aka.ms/hacks-toolkit. But I'm gonna copy the big link anyway since it's here. There you go. Awesome. Yes, exactly. Oh my gosh. So the Hacks Toolkit is all about, and this is in my mind, stage 1. When I go into an organization, I want everyone everyone to number one, be educated on AI, have built a machine so they know what they're like, what they're in for. The benefit of our bot builder and agent builder workshops is that we don't require you to know how to code. You're using your voice to create technology, and that is the future. I know we all talked last week about vibe coding. That is the future. And so the Hacks Toolkit is a detailed approach and toolset that is provided to you for free by Microsoft to help you navigate the earliest waters of baby tiger adoption, right? The earliest moments of when you're trying to figure out what is, what is mine to do as the human and what is, what can I delegate? What should I delegate? And you own that decision. And it is in that decision that you have so much power. And I'm watching organizations and individuals just give that power away. And I, I'm sure it's because they don't realize. Miss the opportunity. But in a, in the world we are in now, it's so critical for you to have agency to choose where does AI help you and where do you hold on to the work that you want to do. And I think in either case, it's augmenting your capabilities. So the Hacks Toolkit is pretty cool. And like I said, the new feature inside of Azure Content Safety is groundedness detection is one of them that allows you to see that groundedness. The other one is that there's what are called prompt shields that allow you to create some level of like middleware that will shield your model from getting prompts that are not good. And what I mean by not good is that they might be injecting data that's gonna corrupt the dataset or the ground truth knowledge of your model. And most people are not thinking about this, but it's just like in the olden days when people used to do DDoS attacks and do a script injection, right? Where they'd put a bunch of plain text or a bunch of of white text on an HTML white page. And so you couldn't see it, but every time the transaction was sending hundreds of thousands of bytes over, or, you know, gigabytes of data in each transaction, causing a denial of service on the backend. And we are going to see the same thing happen, except of course in the form of tokens. So how, you know, if you remember, if you own the model, you're going to own tokens in and tokens out. That is a cost structure you will own. So imagine you don't have a shield that's going to protect you from someone injecting a gigabyte prompt in, you know, or, or not a gigabyte, obviously our prompt context window is not that big, but pushing the edge of your window on every single request just by filling it with blank text or, or gobbledygook or things that you can't see. So these types of prompt attacks are very common. It's also pretty easy to prevent them, but it's like cybersecurity. Most companies either don't know, but if they do know, they don't think think they have the resources to do it. So this is something I think I'm pretty passionate about, is the fact that you have an opportunity to implement an AI system that saves you time and saves you money. So please think to yourself before you've even started, how are you going to carve out 50% or 25% of the revenue or cost savings that comes out of your new gen AI tool? How are you gonna take that and apply it to your cybersecurity processes? Because that's what's gonna need to happen, especially in the world of quantum. Oh my gosh, so cool. What's up, Matt? Oh, yay, I'm so glad you're here. Thank you for joining us. Cool, some OGs in the house. Amazing. Yes, thank you for the onboarding class. Thank you so much. Cool. Thank you, Sherilyn, for those links for those of you who are here. Let's see. Oh my gosh. Yes, N8N. Yes, good one. Yes, yes, yes. Some other tools. Um, this is a great question. And so if anyone is building agents right now, kind of on their own, if you wanna throw the tool you're using to build your agents, I'll throw the one, I mean, I'll talk about the ones that I'm using, but it would be great for people in the room to know if you're doing this work, what, what tools you're using. Um, I'm personally, uh, using a few because that's how I roll. Uh, one I'm very passionate about is we have a relationship, the AI Leadership Institute, with chatbot builder. And, and the reason why is 'cause we've both been in this game for a long time. I a little bit longer, but still, they started teaching people how to use applied AI to create conversational systems for businesses. And I loved what they were doing. And over the last 2 years, they've evolved their capabilities to allow you to not only build agents, but also choreograph. It's like what we would've called in the past, orchestrating behaviors of these bots together where it's not necessarily using like MCP or any complexity in that sense. It just says, what if you want your bot to make a call to a system, right? Those actions, system integration is not necessarily agentic, right? Agentic AI means you're building a robot that's going to be able to make decisions based on data that it gets from its environment, inputs and outputs that are not provided provided directly from a human. They might be provided by a machine, they might be provided by an environment, and it can automatically make decisions and do things based on that environmental data. Now, there's reasons— I mean, this is good news, right? There's plenty of models I wish would do a better job at this. For example, my little robot vacuum cleaner, I wish it would just sense when dirt hits the floor or Cheerios or whatever and go and clean it clean up, right? And me not have to click the button, me not have to look around and go, oh, this place is a mess. Click. Like, I would like that type. Or maybe I have, you know, a little Blink camera, and as soon as it sees stuff fall on the floor, it sends a message, agentic AI, to my robot and says, hey robot, go clean up mess on aisle 9, right? These are agentic, means that we are taking these purpose-built machines my robot for my cleaning my floor and the sensors and computer system I have monitoring my house in case anybody falls and putting them together and orchestrating behavior and allowing us to orchestrate how do these systems talk to each other and execute tasks, get things done, deliver value. For me, it's for my household. For you, it might be for your business. Um, yes, yes, exactly. Uh, let's see. Oh yeah, of course. Yes. No, of course we're— no, there is no bottleneck. Who is this person? Um, so Sherilyn, please tell who— Minnie, uh, where we are located and where we have this conversation. Yeah, I am definitely not the bottleneck. As a matter of fact, you'll see in this list in the Zoom, which you are in, all of our incredible leaders here that help not only facilitate conversation, but just share what they're doing so you can get insight into all sorts of different venues. Um, Matt is one of them, of course. Javier, I think Mauricio is here, maybe. Um, Don, like, who else is here? Padmini. Uh, so yes, please know that this community is super vibrant. We have over 1,500 people in our community and it is growing like crazy and everyone is doing different things. And so the power is in sharing what we're doing with each other and realizing we're not alone on this path. So thank you for that question, uh, because it absolutely is amazing. Um, yeah, so you'll have to join us here. I'll send you the message. How'd you get in here? Oh yeah, you probably just saw me maybe. Um, but if you are, uh, go to— I'm sure Sheryl Lynn's gonna throw this in the chat too. Um, but it is in the school.com, or you can get there by noelle.ai, uh, but school.com/iheartai. AI. And when you go here in the community, there's a community chat and we chat all day long. People are throwing in ideas. As a matter of fact, I have a new challenge. Um, we are gonna be, uh, doing a challenge. Uh, how many of you've seen the action figure challenge? That is, well, it's not a challenge, but people are playing around this new action figure idea. So I want y'all to play around with it too. I'm also going to have you, I tested these out with our AI Leadership Community earlier today. Um, I also want you to play around with cartoonifying yourself. So let me go ahead and, uh, I'm gonna demonstrate a couple tools that I'm using today. I also wanna give you a tool that you might not be aware of that you can use depending on what you do every day, of course. Um, but a couple of things. One, don't forget to sign up. I put the link in here. Sign up for Skills Fast. Join a couple of these sessions you'll likely see. I see, innovation nation in the house, some of us from this community. I, I've seen people in our community in every one of my calls, in every one of these events so far. So that's been really fun. So that's action item number 1. Go do that. Sign up for the Fest. There's so many cool pieces of content, and even if you can't attend live, they are available on replay. And I would encourage you, I mean, these are like pushing the edge of what's possible. And you also, though, will have an opportunity by going to these sessions to complete one of the challenges, which is a Microsoft Learn challenge. And in completing one of those challenges, you will get submitted into a sweepstakes for an Azure AI certification. But what I will be doing is also creating a— we are going to do 50 days. So as a matter of fact, starting on Monday, our 50-day journey began. But we have leaderboards, and in our community, we have leaderboards for 30 days. Place. And so starting now, well, actually starting last week, we are gonna have a 30-day leaderboard and whoever the top 3 people on the leaderboard are. And like the way you get to be in the leaderboard, um, is that you provide content that people like. You can't just spam the community with nonsense cuz that won't get you anywhere. Um, it'll get you banned, I guess. But, um, but what you can do is just say, here's what I'm working on, or what are you working on? Or ask questions. We have 3 different ways to interact. One is to provide a win, something that you're doing that's working for you. Another is to ask a question and get the community to come to your, you know, side and move along with you and help you solve that problem. So there's wins, there's questions, and then there's aha moments, things where you've learned something and you're just like, oh my gosh. And you can submit Loom videos, you can submit content, but be aware we are a self-policing organization. Not everyone who applies to get into our community gets in. I'm saying this to those of you who are, you know, wanting to come in and just do nonsense. We watch each other, everyone. It's usually pretty quick. Mauricio was the last one to tell me like, hey, I think this one is no good. And I went and looked. I was like, oh, no good. Delete, delete, delete. So we are very careful about who gets into these communities and who can participate. But once you're in, it's such a vibrant opportunity to communicate with people who are on a similar path than you are, as you are., and even if they're not on a similar path, like have, you know, this inclusive engineering vibe where everyone's coming and we're all using AI, but in very, very different ways. Uh, we have people that are, you know, getting, that are getting jobs. We have people that are starting companies. So it's really cool to see, uh, the success that people have over time. And remember joining the I Heart AI community or I Love AI community, that community is free. And then only if you want to actually get support building stuff, like fingers-on-keyboard support, take classes, get certified. That's the only time that you really need to upgrade. But lots of our members, we have over 100 members now that have chosen to upgrade to that capability because you do get a chance to put your fingers on a keyboard and build not just Microsoft services, but we have lots of different, like, marketing and sales tools that we use, and you get access to all of those. What else did I want to show you? So this was the Hacks Toolkit. Oh yeah, the toolkit. So let me go and share with you what I did earlier today. I was in Sora and this was my prompt: cartoonify this image. And I gave it this image. It was pretty cool, right? That's not terrible. I love what it did with my logo. That was pretty neat. I showed you this is how I found it. Like I'm scrolling through Sora and I'm like, well, that's pretty cool. I like cartoons. So I did that. Let's see if we can find another one.. That would be interesting. Oh, Sailor Moon! Um, oh no, let's see here. Travel Tropical. Oh, that looks fun. Hmm, let's see. I have to be careful with what I scroll because some— Cookie Monster! Oh my gosh, I love that. That is so cool. What? Hermione! Um, or Emma Watson, as she's probably more likely called these days. Uh, let's see here. Oh my gosh, okay, Barbie Mobile Meth Lab. What? Oh my gosh, what is the world coming to? Um, oh, that's cool. Maybe older person. Oh, all right, I don't know about this. Let me see, let me filter, um, by remixes. So these are remixes that people have done. Um, let's hope I don't— oh, maybe I shouldn't do that. Maybe I'll go by prompts. I'm like, whoa, whoa, whoa, let's not get crazy. Yes, Sora is crazy. It's crazy. Okay, I like this one. I like this one. I also like This one. Okay, there's some good stuff in here now. I'm feeling it. I'm feeling it. Um, all right, let's see Cleopatra. I was trying to find one that maybe I could do. Let's see here. Oh, there we go. All right, that's pretty cool. So look at this prompt. That's a crazy prompt. I'm not gonna be able to edit that in front of y'all today. Um, but I will for next time. But what I found out was, is that in addition to that, there are these like action figures that people are of themselves. I'm trying to find one. There's one. I just think, and here you go, like huge, crazy, crazy, crazy prompt. What I found out was, is that because people are putting these extremely intense prompts into Sora, the, the model is learning what this means so that I can actually instead just say, create an action figure from this picture. And let's see if there's anything cool. Uh, and let's see, from this picture and include accessories, label it an AI apply— oh no, applied AI engineer. And make it look like something a kid would want to buy and play with. I don't know, I'm making it up, like for reals. So notice though, it's a sentence. Let's just see what happens. I do need to give it an image, so I'm going to give it an image and I'm going to choose from my library because I have pictures of my myself in here. There we go. So I'm gonna use that and I'm just gonna say go. Let's see what happens. Now, if I remix the one that you see on the screen, I'd have to go in and like edit that whole thing, which I have done before. But what I wanna do is just test out what happens if I just remix and I don't provide that huge long thing and I just provide a little yet. Um, thanks Kirk. What's up Kirk? I haven't seen you in a while. Um, yeah, I mean, I don't know. Uh, I think it's just Minnie. I don't know if that's really, if that's your name. So apologies cuz I can only call you what I see. Um, but consistency is 100%. The burden is on the prompter. So you do have to provide very specific guardrails in the way that it designs it and replication language that tells it to use the same content. But remember, the models we're using our generators. That's the whole point, is that they generate. So consistency is not the game. The, the game is generation. So that— I get what you're saying though, 100%. Um, no, duh, you're so crazy, Don. I see you, I see you now. Okay, so we're gonna let that do its work. It's there working. Um, I also tried, uh, to cartoonify myself myself in ChatGPT. It did not do a great job. I was a little bit disappointed where I used the exact same— you know what I should do is I should try it without doing— I should just try it like— oh, Ctrl+Z, sorry. Um, I should try it without saying create an image and see what happens. Um, because— oh, maybe I should do a brand new chat. Let me do brand new chat. Let me see here. Create an image, cartoonify this image. Cartoonify is not even a word, but the model figured it out. Oopsie. Eh, Mitch. Then I'm going to give it the thing. I'm going to upload it from my computer, which I think I have a few of. There we go. Open. Let's just see what happens. So now it's gonna think they're both gonna go after it and think about it. But what I found out was interesting, which is what I wanted to share with you all, was that we are now using, like, the remix concept that Sora has introduced has really created an opportunity for us. And I'm just seeing if the using, asking it to create an image, which I think is gonna go through DALL·E, I don't know. We'll see in a second. But the process, like, it's interesting 'cause we're training this model. It's called Meta Prompts, right? These big massive prompts that these artists are using, we now can get access to, to by remixing and leveraging terms that kind of collect all of that detail and nuance into a phrase. So like Funko character, I think I showed you all my Funko character. Oh, I don't even, I, yeah. Oh no. It's a 2D. Maybe I needed to say 3D, but I didn't have to say that in the other one. It's just interesting. Oh, I don't know. What's with the crow's feet, yo? Like, I don't get it. Like, I feel I'm doing pretty good. Like, I take my collagen anyway. But you got some freckles. Now that's, that's an improvement. We did get some freckles. I actually like this one. It's not terrible. That's not right. That's right, Michelle. Like, come on. You're right, Matt. I get you. I guess smile lines, I guess. That's a little bit better. I don't know what they did with this. All on inspires. All right, I see you. Um, that's a little bit better. It's not as good though as Sora. Like, Sora just— I was so— I like, you know, got excited when I saw this pic. I was like, that is so cute, it's so cute. All right, let's go see the magic that it did. Oh, that's interesting. I, I see, I see. Okay, that, that— it's almost there, it's almost there. I'd have to give it a little bit more guidance. Number one, my picture is from 2 years ago and I was a little chubbier back then, so I guess I see what it did. But the camera— the AI adds 40 pounds, as my friend Cynthia told me, um, which is terrible. Uh, anyway, but I like it, Chip. I mean, I mean, I see, I see what it did there. I like the color. I mean, I think it's actually pretty cool. The only thing is that it didn't really know how to put me in there. And I get, and I, I get why, because, um, what I built it from, I, and I almost went and looked at it, but I was afraid I'd lose it. Here's another version of that same thing, but oh, I mean, maybe I used the wrong term, but here's my prompt. I'll throw it in there too. I'll throw it in the chat so y'all can see. Right. So this is my prompt for applied. It's not as pretty because I didn't tell it to do the whole sell it like a toy thing. Exactly, Kirk. I was like hallucinations. This is like hallucinations or whatever, like the anti-confabulation. Like instead of generating, it contracted. So I do have another one. There, look, I have feet. But, and I'll show you, this was the picture I used to ask it to generate for me. Again, I was a little chubbier back then, but it's fine. It's fine. I don't mind. Let me go back to do, to do, to do. Oh, how do I? I can't see the X on my, on my screen because it's all dark. So I just have to go back. Sorry. But pretty cool, right? Like, yeah, it's pretty good. So now, like, you could see, but here's what I think is so fascinating is that to create this, this. Like, this is my prompt. Like, the most detail is not in the action figure itself, which is why I just, I think it's so crazy. Oh, I'm sure they would, Don, because they are, they're, it's similar to like Codex. So Codex is like the model that generates code that was integrated into ChatGPT. So I don't see why not. Um, the only reason why they wouldn't is if it gets so good that they can charge $200 a month for it and everyone who's using ChatGPT won't pay that. So they'll keep it separate for people like me who are like, cha-ching, cha-ching, I want more credits, cha-ching, right? So from a monetization perspective, they might keep it separate. But I see that— oh, Rick's in the house! Um, oh nice. So Rick is gonna share a couple of, uh, things that he did. Uh, Rick is known though for your very elaborate prompts. So I think it's really cool that we may or may not need all of the elaborations to move forward. And that is why one of the things I do want to show you here that I like about Sora is that it has created this concept intentionally. Sam Altman was just talking about it, created this concept of a library so that you could do remixes so that you wouldn't have to craft a prompt by yourself. Yourself. Literally the goal of like OpenAI, right, is that it's open, that now these prompts— I mean, there was a big thing like Midjourney, right, around prompts and prompt integrity and who owns them and all of that. But it's neat 'cause this is now shared with people, right? So I mean, look how, like, how simple that prompt is. I think this is the world we're going to. And so in, in a world where we can have very simplistic prompts, now we really need to think about about what we're doing and who we're copying. Um, I wonder if I have— oh, I think it's actually in ChatGPT. Did I show you all my Legos from last year? Uh, I built like Lego versions. Um, let's see if I can find it. Oh yeah, I think I did this last time. Um, but like being able to— or maybe it was Funko. Oh, I think it's K. Let's see. Yeah, so this was me kind of playing around, and I like looking— the reason I love history, I love this history because this is like, I don't know if it can tell you where it was, but this is how it started. Look at this hair, people. So I'm like, "Creme one, Afro Latina, long curls." No, not a Latina with an Afro. And so I started to get a little closer, start playing around with the hair, started getting more elaborate. It. Um, it finally— I think this one— no, no, I didn't like that one. Started like, no curly hair, and then it got too curly. Um, then I finally landed with this one, and this is the one I ended up using, which I really like. Um, and it like got some of the— I mean, minus that, but it got some of the content right. Like, it was cute. Like, I loved it. So here's my challenge to you. I'm gonna encourage you this week. I will post it in the community for you to create your own action figure Funko or Or, oh, let's just get crazy. You can even do a— let's see what I did. You can even do a Lego figure, right? Which I'm sure I did a couple of these. Lego Noel! You could even do a Lego figure. I'll take you through the history of this as well. It's not pretty. So this, you know, didn't look very Latina to me. Maybe it was the shoes, I don't know. But as I got better, I was like, like I was trying to do like liberty, justice. I wonder what I said. Uh, it might have been like Jedi. Um, oh yeah, yeah. So anyway, so I was like trying to represent all the things I care about, and finally I was like, you know what, never mind. But the hair was still very Locklear, you know, Heather Locklear-ish. So I was like, no, no, that's not it. Oh, we're getting a little closer, but what are you doing with the pants? So then I said 3D illustration So all of these prompts, um, I just got— I got really excited about it. I ended up, I think, choosing this one, which was pretty cool. So have some fun with it. I think my goal is to like use— now that's a prompt I used a year ago. I wonder if I could just say like I did 3D illustration, LEGO minifigure, blah blah blah. I think I could just say LEGO minifigure now. Um, but let's test it out. So I will put this prompt in the prompt challenge. We will pick one winner of the prompt challenge and you will get either— well, I'll come up with some cool thing that I can give you all. But let's do a prompt challenge this week. I will publish that in the I Love AI community if you want to join us. You'll get like a swag bag. Maybe it'll have an I Love AI t-shirt in it and some stickers and some fun things. I would love to do that. Um, let me see here. Did I? Oh yeah, right. Me too. Me too. I love the Lego one. Uh, so super fun. Use, just have some fun with it. Um, oh, I think, okay, last one. Let's see how much. Oh, 502. Okay, last, last, last one. I get all excited. Let me close this out. Um, I also did LinkedIn. I don't know if it's gonna, oh, it's not gonna pick this out. LinkedIn. Image, maybe. Oh no. Oh no. Oh no, I don't think so. I don't know what it was called, but remember everyone did like their LinkedIn thing. Anyway, but okay, I couldn't do it. I ended— I couldn't do that one. So in addition to that though, please also, I showed you all Gamma last year. What was one of the other things I wanted to share with you? One other tool. I think it was not Wonder Dynamics, so— oh yeah, Heydra. So video, when you're publishing content on social, video is everything. It engages people more if it's moving on the screen. It stops— sometimes stops the scroll, stops people from like— like they're just like, what's happening? And so there are a lot of really cool video tools out there right now, one of which I am playing around with is called Heydra. Not like the bad guys. I hope it's not like the bad guys. I'm going to pretend it's not like the villain, but this is an example of an AI-generated video, but it's a bunch of them. So any one of these you could create, like something like this, um, where you hyper, you know, create, uh, an individual or a person talking. Um, or you'll see there's a cartoon that shows up. You can actually have it. Isn't that cute? Like, I love that little Pixar. I'm gonna make one of me with curly hair, maybe pigtails. I don't know. Um, but all of this becomes available. So I'll just throw this link just for those of you who wanna play around with a new tool and have it. Oh, thank you, Sherlyn. Um, if you're interested and wanna play around with it, but our challenge this week will be create your own action figure, Lego, Funko, and post it in the community. Also share your prompt cuz if you do something cool, and my hope is you'll do something cool, we wanna be able to replicate it and, and play around with it. Right? Ex— oh, see Matt, this is why we're friends. I don't watch that much TV as many of you know. Um, so I don't know these little nuances. So Hydra is the one I'm worried about, and haydra, oh, is Arabic for presents. I love them even more now, except they didn't spell it the same, but still. Okay, I'm gonna go with the present side. I love that. Hydra. Oh my gosh, what is wrong with me? Um, I will watch more Captain America, I promise. Uh, but thank you, I appreciate you. Uh, yeah, So this is a cool new tool. It is yours for the using. It does— they give you 400 credits out of the box to play around with, and it's plenty to create an image. What was the other one? Okay, maybe I'll just show one more, one more, which is where I'm having some fun, and you are going to see some content that I've been working on. It's called Wonder Dynamics. Wonder Dynamics allows you to build content um, video content with robots. And so you could see here like some of these, you know, really cool robotic things, but I get to act out something myself, record myself doing something, and then supplant a robot in my place or vice versa. So I'm creating actually a conversation where we're walking down the street. My husband and I recorded this, but I'm replacing my husband with a machine and we're gonna have a machine to Noelle interaction. Session, and I'll be posting that later. It's pretty fun and pretty cool. So very, very exciting. Awesome. Cool. Absolutely. So really neat. Hope you enjoyed the session today. Any questions or comments, you're welcome to throw them in the chat. But that is always my, my hope is just to keep you just a little bit excited about the world that we're in. That yeah, there's a lot of darkness and weirdness and bad usage, but I feel like it's our responsibility to be the beacon in the world, be the lamplighter, right? Light the way and say there is a way to safely, responsibly, and profitably use this technology in the world. People need our help. So the more you learn, the more you can teach, the more you can share, um, and the more successful we can actually make everyone. The rising tide of this technology, it raises all the boats, not just the yacht. And definitely not the Death Stars. So let's go. Um, yeah, the last two are awesome, Javier, if you're interested in video. So I'm doing a lot of experimentation in video. I'll share with you, uh, and then we are doing a little revamp to the AI Leadership Institute over the next two weeks. So I'm excited to share some of that with you. I'm just trying to make sure I didn't miss any announcements that I'll hang up and be like, darn it. Um, no, I think I did it. I think I did it. So I'll see you next week, next Friday. I think it's next Friday. Yeah, next Friday. We'll also have in the community opportunities to engage. And like I said, there are prizes to be won in our action figure challenge. So jump on in and have some fun and show people what you can do. I'll talk to y'all soon. Thank you. All right, bye. Yes, Don, of course. I'll probably generate a new meeting and we can go chat over there. Bye, everybody. Thank you so much. Thanks, Char. Oh, I love it. I love it. Thank you, everybody. Oh, I think I stopped sharing. Yes, thank you, Angel. Bye. Have a great weekend. Thank you, Cynthia. Hi.