Day 4 of the I ️ AI Challenge : AI for Professionals

Alright. Okay, okay, can you guys hear me now? What's wrong with my internet? Oh, there we go! Are we good? Oh my gosh, I think we can see it! Okay, I know my audio wasn't working. My husband to the rescue. All right, can you hear me now? You can't hear any audio. Oh, Oh, yay! Great, thank you. Okay guys, we're ready, we're live. I'll turn off those beautiful relaxing calm vibes that were in the audio. Okay, awesome, we are live. Wow, okay, I mean, 11 minutes. Normally I would just like talk about this time anyway, but you guys got a chance to really show your true colors. All of you were supportive and loving and kind in my moment of need, so thank you for that. Wow, okay, and now I'm like sweating. Anyway, it's, uh, amazing. So glad to see you all. Thank you for being here. I appreciate you, Michael, here for me. I get it. I get it. Houston, we are live. Absolutely. Thank you so much. Okay, great to see you all. Thank you for your patience. But here is first lesson of the day, right? So I am a big believer in certain— like, so today is all about how to use AI and ways to use AI in your professional life, but it also is about how will AI impact your professional life. So really, A portion of what we're going to talk about are the roles that we have available in this industry that you can choose to take a part in. But part of that, in order to really excel and do well in this industry, it's really important to know, I think, that to think about 3 things, 3 different types of intelligence that are needed here. Like one is, of course, functional intelligence.

Right?

Some might call it IQ. Like, you need to be able to acquire new skills and learn things, which I know all of you have the ability to do because you are here and you figured out LinkedIn enough to join my live or join my community, right? So we have skills. We can— everyone listening to the sound of my voice has the ability to do cool things. So excited for you guys. Thank you so much for the support and coming, coming back. But in addition to functional or— and in addition to functional intelligence, you also need— anyone want to guess the second type of intelligence you need, functional IQ and emotional intelligence, right? So emotional intelligence, EQ, became very popular. I don't know what, 10, 15 years ago. EQ and emotional intelligence means that you have the ability to be empathetic, to understand, you know, human needs within a conversation, to know when to speak and when to be silent, to know when to amplify others' voices, when to How and when to interact with others, like all of these, how to be a better leader. Like all of this is around, yes, EQ. Thank you for that, Joe. And so when you're thinking, so now we've got IQ, you need to have functional intelligence. Luckily for all of us, that is acquirable, right? We can go out and grab functional intelligence by building projects on GitHub or doing, you know, reading books from Barnes Noble. I know I always go back to the old school, but That's how I learned. So that's a vehicle as well. You need to have emotional intelligence. Lucky again for us that that is also an acquirable skill. Like, you can spend more time examining how you emotionally react to things, and just— it's— I think it's called the Heisenberg principle, right? Like, just observing your own emotional state throughout the day will increase your emotional intelligence. It'll increase your sense of like, oh, I feel offended in this moment. I'm not going to react to that, but I feel offended. I feel betrayed. I feel happy. I feel excited. I feel positive. Just being able to— I always tell people it's important to separate how you feel about something from your behavior or action towards something, right? So identifying how you feel about something and then making a candid, intentional decision about how you're going what you're going to do or, or how you're going to behave. Like, that is a choice that you'll make. The third Q, right? The third quotient. The third, you know, we've got functional intelligence, we've got emotional intelligence. Here is one of my favorites, and it's called the adaptability quotient or adaptable, like adaptableness, your ability to adapt. And What that means is that when— and I do these in my classes more intentionally than like today, it was definitely unintentional. But how do you behave in a moment when everything changes, when something you expected to happen doesn't happen perfectly, right? When something you thought would work doesn't work out, or something you didn't think would work ends up working? Like, how do you deal in a world of ambiguity? And I know during my time— and actually Wes is here too— like during our time at Amazon when— and everybody who's ever worked at Amazon, one of the things that you know is you're constantly living in this state of ambiguity. And they quiz you hard when you're going into that company on like, how well do you deal with that? Like, how well do you deal with a situation that's unknown? And how well do you navigate questions that you are just forming as you're building it? Many people call it building your plane while flying it, right? So Being aware of this is super important. And I think it's one of the most important skills that I, I look for in candidates that I try to hire, as well as candidates that I refer to positions. Because in the world of AI, there is no— and of course, somebody will, will, will likely say something to the effect of, hey, change is the only constant, right? Like, that's a common thing that we say. But Yes, of course that's true. But how resilient are you to change? And can you, in the world of change, become a solution architect? Some of us are solution architects here. I myself, I live and breathe. I'm a solution architect of my own life. When something happens that I don't expect or when ambiguity clarifies quickly to some area I didn't expect, I begin to immediately think about the solution. And I'll also give you another practice. And the reason I frame this for you is that as you move into a professional role in a company that's investing in AI, there'll be a lot of ambiguity. Your job, what the one you say yes to, may not become the one you end up doing 6 months, a year later. And it's important to have this type of adaptability. And the better you are at it, the more successful you can be because you can move from— I always say when something unexpected happens, it's good to feel the feels, right? It might be good, it might be negative. If it's negative, like, feel hurt by it, be upset about it, be angry. Okay, not super angry, but like, you know, have some emotions. But it's like, as soon as you can, like, feel those emotions, but don't let that be your story. Use those emotions to drive, how do I do this better? How do I solve for this, whatever I'm trying to do? And what that means is that as you start to take on AI roles, you are going to be in a constant state of evolution. Many of us have said— I've heard some of you, um, on in your LinkedIn posts or sharing in social media talking about like, wow, there's new tools on the market all the time. How do I keep up with all of it? Like, we do the best we can, right? Um, and I think the best thing that you can do is something we talked about a couple days ago, is pick a lane, right? Pick an area of the world that you care about Maybe it's nonprofits, maybe it's manufacturing. I was just talking to some movers because we're about to move. Maybe it's the moving industry, maybe it's trucking and maybe, you know, maybe it's manufacturing of consumer goods. Like, who knows? Pick a, pick a, pick a lane. Doesn't matter what it is. Where do you feel called? Where do you find the most passion? These are things that will help you when you find yourself in a world of change. It's nice to be grounded in something. Contact center, like I am an expert in. Customer service experiences. I'm an expert in design. I'm an expert in whatever area of the world that you feel called to. That's your domain. And then start researching as much technology as you can in order to pull those things closer to you. So as a professional in the world of AI, we're going to talk about a couple different things because I always like to give you a lay of the land before we go in. But yeah, my phrase, my motto is feel the feels, then make a plan. And in my mind, I'm thinking, how do I solution architect this problem? And it doesn't matter what it is. It could be me going through a relationship challenge. It could be me going through a challenge at work. It could be me deciding what next thing I'm going to do. But I move away from, gosh, I'm worried about this, the negative emotions that tend to keep us stuck where we are. And I try to have enough self-awareness and adaptability to shift my focus to where do I go next. And then many of you have asked me like, oh my gosh, Noelle, like you're always moving on to not just anything, like you're always moving on to bigger and better things. And I will tell you, it's not because I don't want to stay where I am, but I do recognize I'm emotionally intelligent enough to look around and go, can I be of service here? And if I can't do what I do well, in that space, it's better for everyone if I move in the direction of how I can help. And in my case, I, I love clients. I love helping clients figure out problems. I help helping them look at business problems and find technical solutions. And especially when those technical solutions land in the world of AI, that is— I love that process. And I actually do like it across a bunch of different industries. But that solution architecting is a critical skill right now, no matter what your role is. Because many of us will come to the table and just be like, this is broken. It doesn't work. This is algorithmic bias. These models lie. True. And we should recognize that. And we should be aware of it. But the very next thing your mind should be thinking is, how do I leverage technology to move any of those problems in a better direction? That's the impact we can make in the world, right, is Moving problems to solutions and having that mindset is going to be critical to you as you move into a world of AI as a professional. So there's a couple of things I'm going to talk about. Let me turn my whiteboard on so you could see it. Hopefully you could see it. Okay. I don't know what my brightness level is. Hopefully it's not too bad. Oh yeah. Yeah. Okay. You can see it. So, oops. I always start with the— I love purple. It's my favorite color. But professional, profe— oh my gosh, professional, like AI professional. That is what we're focused on today. And as an AI professional, we're going to talk about a couple of things. One, what are some of the roles that are available right now? And I'm not going to go through all of them. I remember at Accenture, I think they ended up defining 12 or 13 roles that were now new in the world of AI. I'm not going to go through that kind of level of depth, but I am going to point you in areas where I see direct need, where there are gaps, where you could be the one that fills these gaps. And I know several of you on this call are exactly these people. And the challenge is that once you identify a role, how do you establish yourself as the one to take it? And in my personal opinion, I would love it if your reputation precedes you in the, the clients or the customers or the hiring managers that you want to talk to. How do you take your knowledge and passion for artificial intelligence and get it into the world? It ties together sales and marketing that we just talked about over the last two days, but this is a critical set of skills that you're going to need to learn in order to do the work you love in the area of AI, which is what the world needs right now and that you're good at. In order to push that into, you need to become known. I mentioned this yesterday in the marketing section where we talked about, you know, known— being known beats being the best every day. And what that means is that if somebody is better at getting known to a hiring manager, to a company, to a brand, to a publisher, they will get the deal because they're in front of them, they're known to them, as opposed to you who are amazing but no one can find. No one can Google you, no one can see you. So I know some of you might find a— or maybe have a visceral reaction to like, wait, I have to be like public with this? Not necessarily public, but you do have to start establishing your voice in the industry, and that We'll talk about that as well. Finally, we'll talk about some AI. And I'll actually— I know a couple of people, maybe 10 or 12 of you that I see here right now. Many of you are in this space right now and have used these tools. I know Jay talked yesterday about how he used an AI tool to help him rethink about his LinkedIn profile or rethink about your resume or CV. So I would love for you to drop in the, the notes again that that information, but what are some AI tools that can help you as a professional? And I'm going to give you two pieces of advice. It's a little bit biased because I helped build them, but I'm gonna encourage you, no matter where you are in your AI journey, I'm going to encourage you to use these tools. Now, these tools, one is a tool where you're actually gonna fingers on keyboard prompt engineer your way to success. The other though, are 2 classes that I would recommend. So it's going to be tools and what I call resources. Right? So we're going to go through all 3. So let's first start talking about some roles. There are some new roles that are available. I'm going to actually end with the new roles. And I'm going to start with roles that are a little bit more traditional. Sorry. Oh, yeah, you know what I'm supposed to do.

There we go.

So I'm going to start with roles that are a little bit more traditional. Hopefully you guys can see this. It It looks a little blurry to me, but some traditional— like all my tech. Oh, there we go. Some traditional roles that you might see out there that are going to lead you down a path of AI. And I think it's important to think about this because you don't necessarily like— don't play the short game here. Oh, sorry. Hold on. Let me go to chat. I see you all talking to me. There we go. Let's see. Oh my goodness. Sorry. Oh yeah, yeah, yeah. Okay, cool. I'll do that.

Right.

It's because I'm in a different setup because I had to like totally change my thing. So great. Good idea. All right. Anyway, you can make that there. Does that look better? Oh, here, let me. Take this away. And then I'll just write it right here. Is that cool? Would that work? Hopefully that's good. I should look at you guys on LinkedIn. Also, okay, hold on, let me do this. Hopefully that's good. Okay, we're good. I'm looking at our comments. All right, we should be good.

Let's see.

Oh yeah, good. Okay. All right. So, and I'll write a little bit bigger so you can see it. Um, amazing. Thank you, Michael. Thank you, Malini. You guys are awesome. Um, you know, it's Thursday, so like my plane is landing, um, and there's always a little turbulence on the landing, so we're, we're having fun together. So thank you for bearing with me. All right, so some traditional roles that we have in the AI space. One is data roles. Like data engineering. Now, I used to always tell people, if you're brand new to the space, the best way to get into AI is to start in a data engineering role. And the reason why I say that, and I'll explain what that means in just a second, but the reason I say that is because it's so important. Many of us, we put in the reps, uh, in the AI world, right? Like 10 years ago, being able to, um, move in the direction of data, which is the fuel for our AI models, but understand the entire machine learning lifecycle, right? Understand that the first thing that happens in a machine learning lifecycle, it's building out data science and creating new models. The first thing that has to happen is we have to acquire data, diverse, inclusive datasets, and then we have to take that data and make it useful to a model. Doing things like, I mean, we don't talk about this stuff anymore, but like doing things like feature engineering, like someone has to pick What does the model deem important and what does it dismiss? Eventually we're going to talk about that again in the industry. Right now we're so excited about models that are already built that we don't ask these hard questions. But in the world of generative AI, your data engineering skills will translate into perform— parameter efficient fine tuning skills where you can evolve into, okay, now that I understand data engineering and feature engineering and preparing data to be used by a model, once I get a big model, an LLM— oops, there we go— um, once I get an LLM, that LLM is going to— I'm just going to try and turn a little bit. Yeah, so, so an LLM is going to give us the ability to create efficiencies by identifying parameters we want the model to focus on. Parameter efficiency. And so it's very similar Tangential even. But would you feel comfortable going in and becoming a PEFT engineer? Probably not. You might be like, I don't even know what that is, right? And so you could start though with data engineering, which is look at data and test. The nice thing about data engineering is it's a lot of testing and experimentation. One of the things I always tell people, business owners, executives, is that when you're building AI systems It's important, even though right now we're kind of selling AI as a product and productizing it and operationalizing it. But the reality is, is that AI has always and will always be at first experimentation and experiment. So you need to have a mindset of an experimenter, right? An experimenter is also known as a researcher, someone willing to ask questions. It's my most valuable skill. When I look at people joining teams, I'm like, when you're starting to learn something, what questions do you ask? How do you ask better questions? So I think data engineering is a really interesting directional longer game because you're going to get this job, you'll do amazing at it, and then you'll be able to acquire these skills of performance efficiency much more effectively. But that is— and sometimes you'll see this as LLM Ops. Because we like to ops things and like, you know, DevOps, MLOps, AIOps. Oh, I put LLL. Noelle coined a term here, but LLM Operations. So here— awesome, cool. Thank you, Roger. Okay, cool. So Yes, I agree. So Roger just said, you know, it's important to kind of get comfortable being messy. One of the things generationally I've noticed is that there are certain people that don't like— if they're presented with a button, they might not click it because they're afraid something might break. We need to be in the world of AI and not in production, but in our experimentation. We need to be willing to click the buttons and find out what they do. We need to be willing to You know, use one feature versus another and just experiment. And most importantly, we have to be willing to fail. We have to be willing to fail. And so now two roles have now come up because we need people who are willing to fail. But guess what? We also desperately need leaders that understand that failure is a directional move towards success, especially in AI. We're going to need data leaders that understand that the data journey is a long one. It also— I'm actually wearing one of my Lamplighter shirts today. I have this philosophy called Lamplighter, the Lamplighter Effect. I also have a podcast if you want to go listen. But in this, a leader has to be thinking, how do I light the lamp with artificial intelligence? To give kind of hope and excitement to the organization about what AI is possible, what, what AI can do. But at the same time, it's going to pull through or drive work in the data world. Because unless you have good data, you're not going to be able to create good AI. In the past, we used to have to do a lot of work to get our data before we could do any AI at all. We had to get a lot of good data before any AI was possible. Now, With an LLM, I actually can do more of AI. I can take a smaller corpora of data and use that to build a tiny little light that says, hey, this is cool, we could do this. And it gives enough momentum to a team to allow them to go. But again, you have to have a leader who could do this. But it allows a company to see, hey, this work is hard and long and takes a long time. But if we use an LLM, it gets us excited about the possibility of— it makes the art of the possible a reality. And that goes back to what we said on Monday, show versus tell, right? When I can show someone the reality of what an AI system could do for them, and LLMs let me do it immediately, like right in front of them, I can show them what's possible. It allows them to believe that all this work they're going to do in the data lifecycle that that's going to be harder. Now, I haven't thought about this in a long time, so bear with me, but I think it's useful if we do talk about the, the, the machine learning lifecycle when people build machine learning models. So I already talked to you first about the first thing we need to do is capture the data, right? And I mentioned that when we capture that data, it needs to be inclusive, meaning that we need to go get a lot of data. It also needs to be diverse, meaning not only do we want it from a bunch of different places, we actually want— that's it, diverse. We also want to make sure that in the data we collect is actually included in the training data that we use. I appreciate all you all providing this. Okay, so I'm gonna I'm going to also go back and answer any questions I missed, but if I see them, I'm going to try and answer them right in, right in line. So first step is data. The second step is where we take that data and start to make sense of it. Now, I used to call this— I don't even think I built a big enough box. I called this wrangling. But it's not a good word. Like, people are like, what does that mean? When I talk to people who are not in this space. But data wrangling is all the things once we've done the collection, right? So this is collecting data. We now need to make sense of it. We need to wrangle it. So in this is its own set of activities, right? And you could throw these activities if you know what you do. If you're a data wrangler, you could throw these right in there. Sheila, what's up? Happy Monday, wait, Friday, Thursday, happy day to you. So wrangling includes things like, I mean, simple, deduplicating, right? Being able to look at data and I mean, even feature engineering is part of that or assessing validity or, you know, there's tons of things you could do. There's lots of other words I would throw in here. And maybe some of you will throw some of them in there as well. But there's a whole bunch of stuff that happens once you have the data to make sure it's good data. Oh my gosh, let's at least put clean the data, right? What that means is these are, deduping is an example of cleaning. Yes, thank you, what? So data cleaning, we're all on the same page. But yeah, cleaning the data, right? Making sure that the data is actually representative And this is why if I go in and I grab a dataset, the wrangler's job is to make sure that that data is representative of the problem we're trying to solve. Now, many times the wrangler has no power at all, no authority. Like, they have no ability to go back to the collection team and say, hey, get me more stuff. Like, they don't get to do that. I wish they did. But again, that's why we need these leaders, because we need leaders that are going to enforce that process, that cyclical process. So we've got data, we've got wrangling, then we've got very similar to what I do now in LLMs, we've got model selection. Now there are tools like AutoML that allow you to give it data that you've wrangled and it'll actually choose the right model for you based on the data you give it. And that's kind of magical. We should do that. But on the off chance that, you know, I always tell people, actually this came up on a webinar I was on yesterday called We Love Generative AI, and it was with T. I don't know if T's here today, but it was with T. It was awesome that they, like us, ran out of time and we had so much content we could have gone into. But AutoML and building machine models, I always tell people like the best thing you can do is build a model from scratch. And actually we talked about it yesterday with Aaron. From ChatGPT, right? She, she talks about how important it was in her journey to learning more about machine learning to actually just build a model from scratch. No one really needs to do that anymore, at least not unless you're working for DeepMind or OpenAI or a model-building company. For most of us, 90% of us, we're not going to be the model builders, right? We're going to be the model users, consumers. And that's a critical role because many, many models Millions of models have been created and never used. We are the users. We're the ones who take an idea from a data scientist and make use of it in the world. And that's, I think, a very important role to play. But modeling, picking the model is important. And many times you'll pick a model and it'll seem like a good idea, but through testing and iteration, it ends up being— it ends up not being able to yield the results that you need. So there is model selection. And so I usually break it into selection and evaluation. And so now start to see yourself, right? If you're looking at this, start to see yourself like, what kind of person are you? Are you a, are you a good data collector? Are you a wrangler, right? I usually find a lot of analytical people are in the wrangling space, right? Are you good at model selection, meaning that you have to be really good at experimentation research? So this now requires, and as I mentioned when I started this, the reason I wanted to talk about it was because in every AI project you're going to need to take on a research mindset. Like, how do I find the right model for the right task at the right time? That's going to determine how successful your model is when it gets to production. Now, once you have that model, of course you need to test it, right? And then you're going to deploy it, and then you're going to monitor it, and then you're gonna use the feedback of that to probably collect more data, right? Which is— this is why we always say when you launch a machine learning solution, it never ends because you are constantly improving the model you selected, evaluating its effec— efficiency and effectiveness, testing that model, deploying it, getting data, getting new information all the time. Like, this is where often we see the monitoring piece is where humans in the loop are often referenced. Who's going to tell the model, yes, you're good, no, you're not? And this is very generic because there's lots of different types of artificial intelligence. So as you think about where your role might be in this world, I think it's good to see this. And this means if you want to be on the building side of a model, these are some of the things that you can dive into, you can become an expert at. Now let's take this a little bit further, um, from the perspective of Right. So this is like, I don't know, in the data engineering space. I personally like the roles that are part of what I call applied AI engineers. So now, like I said, the previous diagram, this one is all about how do I build a model? And you could be on a team that's building models. People are building them all the time now. Many of us have seen our notes that come in and it says, oh my gosh, I got another message in my inbox that said another model was built and it's better and faster and stronger than the last model I got. So there's lots of work in this space if you choose to go that route. I think the bigger opportunity, at least for people like me, like that want to adopt AI to solve real problems for businesses, is in applied AI engineers— engineering. And this is where I've had some of the most fun in my career, which means you are going to take AI models that are currently available as an API and you're going to use those APIs inside web applications, inside mixed reality, virtual reality, augmented reality applications, inside of— of course, did I say web? Mobile apps, right? Like you're going to use a multimodal approach to leveraging AI to infuse and make better all these existing experiences and even maybe make new experiences. So I'll give you an example. One of the things that I, in my roles, I've had multiple roles in this applied AI engineering space and I absolutely love it. It's the most fun because I get to actually build things that solve problems. But one of the things I did, I was working at Microsoft and I was in the mixed reality airlift. And I remember my, my job at that airlift was to educate everyone on applied AI models. Today, if you're at Microsoft, this is known as Azure AI Services. If you're at Amazon, don't be mad, Azure. I'm just like showing the love to other people. AWS would be SageMaker. And at Google— and I love them like my children, I love them all equally. They're not in order of— maybe they're in a little order, but they're not meant to be in order of importance. Google has Vertex. So start to see the lenses, right? If you decide as a professional to move in this space, You now have the ability to decide, do you want to be part of model building? Do you want to be part of model consuming? Do you want to be part of Azure, AWS, Google? Do you want to be platform agnostic and understand all three, which is where I like to play? I like to pick the right tool, the right model for the right problem at the right time. And that might mean that I'm in a multi-model world, right, where I might use one model. I, you know, and the nice thing is, is that these are all APIs. For those of you who are not familiar, application programming interfaces, it means that I don't need to know anything about machine learning to use these. I just need to know how to code. And I don't even necessarily need to know how to code because now with things like Power Platform on Microsoft, I can point to these, I can create like a, a WYSIWYG, if you remember that term, right? What you see is what you get, this graphical interface where I drop boxes together, connecting them with lines, and in one box it's gonna say, what service do you wanna use? And I could actually use any service I want from any provider. What it does require you to do though is get familiar, get smart if you will, on what these different organizations need. All right, so I love this part. So now I'm gonna focus on Azure because I'm an Azure AI MVP. I love the team. They're good people, which I love that. I'm sure AWS and Google are too, but I know these people and they work really hard and I love the features that they've been prioritizing. I also have access to the sandbox in an easier way. So I use this tool every day. I'm in my Azure subscription. So once you decide you're gonna be a consumer of a model, and I'm just gonna use Azure's services to begin with, So let's take a look. But if I'm a, um, AI services, if I'm an applied AI engineer, I'm now, and this is what I always tell people, the first thing I'm going to do when I'm solving for a problem using AI services by any vendor and maybe even a combination of more than one, the first thing I'm going to do is I'm going to define the problem, right? What is it that I'm trying to solve for? What is it that I'm trying to do? The problem will then lead me to what kind of data do I have? And this, of course, ties up with, right, that data engineering role, right? So now I've got a problem that I need to define. Roles that are great at defining problems in the scope of AI solutions, right? This is like a TPM or even a PM, project manager. I always think project managers are TPMs, but a lot of PMs will be like, ooh, I'm not as technical. And I always like, then go learn the tech part because you're excellent at this and we need you to help run engineering teams. So if you have a heart for engineering and you're not that technical, do not let that keep you from being a technical program manager because— or a project manager. But the word T should not be intimidating. Like, pick up the T. You can learn the T. The PM part is kind of— I don't know, I find like that's just not my skill set. But, but I find like these people are magical. They're the way I get stuff done in every business I've ever built. So PM, TPM, you might also find in this area UX people as well, right? Because they're going to hear the problem from the user's perspective and they will be great at articulating what that problem is, storyboarding, creating user stories, all of this stuff that's really important at this level. Now I go and I do a bit of investigation on the data that I have. Maybe I have a lot, right? Maybe I have a lot. Maybe I have none. Maybe I have a little bit. This will determine the next step, which is the model that I want to use. Now, the difference here is that when we're building models, we're choosing an algorithm. That algorithm I will have to train from scratch. I will have to then test to make sure that the test data matches the training outcome that I got. There's a whole lot more involved in actually building the algorithm, tuning the algorithm, right? Managing the, the features. When we're in this, we're actually picking a model that's an API call. Which means I'm gonna make a call and I'm gonna get a result. I don't have to worry about— I'm presuming the other part's been done already, and it's actually the reason I do it. I call it building on the shoulders of giants, right? I wanna build on someone else's work that's already done that. Think GitHub, right? But Microsoft has like 20 of these models and only a couple of them are LLMs, right? I think LLMs are being infused to all of them, but only a couple of them were focused on being LLMs alone. The model selection will be directly correlated to the problem you're trying to solve. So if I'm trying to solve a problem— I was just recently talking to a bunch of the, uh, farming communities, uh, the largest dairy farming conglomerate, I guess, that you'd call it. Um, and it was awesome talking to them because they were just starting to use AI in a real way. But the things they were using AI to do were not like, hey, I need a chatbot, though that is one of the use cases they ended up lighting up. One of the biggest use cases that yielded the biggest benefit was actually using computer vision. And computer vision right now, like if you go, you can use what's called custom computer vision. And I think it's called Vision Studio these days, which is also available in the AI Studio, but Computer vision means I don't have to go and train a model to understand what a door is, a chair is, a human is, a dog. Like, there are models that already do that work. So I can start, build on the shoulders of giants. I can start with Microsoft's pre-trained model that already has all this stuff done, and I can then build on top of it. So I can custom train this model. So now I've got again 3 separate types of work that needs to happen in this world. And you might become like a consultant is one that can go from the boardroom to the whiteboard to the keyboard, right? Somebody that can navigate through all the stages of development from ideating around it to architecting the solution to fingers on a keyboard, building it and deploying it. The more, the closer you can get to that, the better. It's a unicorn skill. There's not many people who can do all of those things. Because oftentimes we silo into an area. But this shows you there's different people that are necessary to do this work. The data people often end up being like data architects.

Oops. Right.

Data architects or even solution architects. But data architects, some of you might remember the term MDM, Master Data Management. I think the important thing to realize is that they're not going to say, hey, I need AI people, but you will be critically involved in AI work. And that AI work can become part of— remember the second piece I wanted to talk to you about— part of what you say to the world that you do. So do not get bogged down and say, I only want a job that has AI in the title. Most companies don't even know what that means. They don't even know how to put it in a job description. Some do.. And you can go after those. And I, of course, believe that you could do that. But there are so many that are going to be using AI that have not transitioned their roles to ask for it, where if you start to see these patterns in the roles that they're looking for, you know, they're doing AI work and you can become a critical part of that. What I love about that is that you can do what I do. You can start at the beginning. I started at the beginning of Alexa. I started at the beginning of cognitive services. I was at the beginning of Red Hat's OpenShift data science platform, which I think just got renamed. Like, I got to be there at the beginning because I knew the signals of what kind of work was needed to build things like that. And that's part of your adaptability quotient. How do you put your ear to the ground and listen to the pulse of the market to know, oh, you're building a team, you don't know it's going to be an AI team yet, but these are the components. Or maybe they do and they just aren't saying it. That's another thing when you think about companies. When we're looking at this, like, many companies are not going to even say what they're doing to the world. They're not going to advertise that they're hiring AI people because they don't necessarily want to share with the world that that's where they're going to start to see a lot of benefit in their own organization. All right, so we've got the problem, we've got the, the data, We've got the model creation part. Computer vision is just one example. There's also, of course, natural language, right? So natural language processing or natural language understanding. There's also decision logic. And of course, there's LLMs all over this. But natural language, there's also speech recognition, there's robotic processing, sensor processing. So there's a lot of different types of models that are available. As a matter of fact, Like I said, I think Microsoft has like 20 of them. Just go— I was gonna say Google again, but Google or Bing, um, Azure AI Services. Go to the sales page, you'll see how many they have. 5 different categories with 6 or 7 models each. 25 models? I don't know, I tried to do some quick math. All right, so we've got the model. This doesn't mean you're going to build the model. It means you need to be smart enough to know how to use the model to solve the problem.. And that is a cool skill to have. Once you've built the model, you are going to need people. And actually, I think I'm gonna transition this down a little bit. You're going to need people that care about 3 things, and they might be the same group of people, but often they're different teams. One team is going to need to be concerned about security. Cybersecurity in the world of AI is a whole field. You might not know cybersecurity now, but you can learn that. That's a learnable skill. Go learn that. And then start applying your passion for AI to this field. That is a huge opportunity over the next 3 to 5 years. Security. I often call this defense because you are going to be building systems that are going to increase the aperture for companies to talk to your business, and you're going to need to create defense mechanisms to protect your business in these conversational applications. The next thing you're going to need to do is have people who care about how accurate the answer is. You're going to need people who go, is the model giving the right answer? Is it hallucinating? It is possible to create models that don't hallucinate. It is possible. I've seen it happen. However, there are rules to implementing projects like that. And you're— people that care about this, this is also going to tee up information architecture. I always say information architecture is going to need to come before IA, before AI, because people are going to need to figure out what information do they have that they can use to feed models that they want to use, right? So accuracy is going to be a huge part. And again, a job— if you are one of those people who think that way, you're like, there's a role for you here. The next role. And that's— I don't think it's the last one, it might be, that I'll talk about, but it's around— we'll call it RAI. But I will tell you, there'll be very few responsible AI roles. Some of you are in these roles now, and you're amazing, and I love you, and you're awesome. But you may not be able to find a role that says responsible AI. But what I want to encourage you to think about is, and, and really this comes down to algorithmic bias and how do we detect it, which also becomes, if you know a little bit about a project called HELM. HELM is an LLM, it's called Holistic Evaluation of Language Models, which is awesome, right? HELM evaluates language models and Many of their metrics— I shouldn't say many, I just don't know how many, like 10, 15 of those metrics are based on toxicity, based on this piece. You could become an expert of not just understanding it, but what tools in market are able to detect it, which ones are better than others across the 3 vendors that we talked about. Like, this is a skill set you can acquire and you will be a unicorn. Like, there aren't many people who have this specialization. So Helm will give you a framework for thinking about toxicity, but someone needs to care about that problem on a team. Here's the thing, you might become a project manager, you might become a, um, a, you know, an AI engineer. If you can layer in this skill set, that not only will you be able to choose the right model, you will choose the right model that mitigates bias and risk. That is huge. So I can even add the word risk. Risk mitigation. So there's, you know, I'm not going to throw— I'm probably over-inundating you with lots of stuff. But you could go back and watch the replay. But risk mitigation is huge right now in the area of responsible AI. When I go into a company, I don't even say responsible AI. I usually just say, like, we are going to build this so that we serve as many people as possible without hurting anyone. Like, of course, every system should be built this way. Oh, thanks for that, Michael, putting that link in there. But everyone— so for example, NIST is an organization and they've created a risk framework. Most companies don't even have their own risk framework. If you— which is what I did with the executive order that came out of the Biden administration. I printed that puppy out and I highlighted all the things that I thought were cool. And I used it to frame my mind around, you know what, if every company built a fraction of this, for their own organization, we'd be in a much better spot because it forces us to think differently about how we're going to market and what we're doing in, in with our business. And so NIST is one of those types of documents they have. I shouldn't say that. NIST is the organization, and then they have a risk document. You can Google it. I'm trying not to overwhelm you with like the names of all the things. You could just Google it. But NIST has this risk framework that you can get very familiar with and then use that to help your organization. So in any of these roles, when you're talking to a hiring manager or when you're talking to a potential client, you can say, I have— I am deeply familiar with the NIST framework, which is used by these companies. It helps credentialize you that you even know this and that you can speak with, you know, integrity about what it's about and what it does. And again, it's a learned skill. It's just your choice to go down the path of learning it. So We've got now— I'm going to shrink this a little bit, right? So we've got people who need to know the problem. People are going to figure out what the data strategy is. So I mentioned data architects. I think just as important, there's probably data strategists. Anyway, it's so small you probably can't see it anyway, but strategists, data strategists are in there as well. Someone needs to understand what's the data story of the business and what do we need to do to evolve to take more advantage of AI. Awesome. Thank you, Michael's all on the, the dropping the nuggets. So thank you for that as well. So then model selections, you need to go into, okay, maybe not you if this isn't the role you want, but you have an opportunity to go and learn all of the different AI models. I have built solutions on every model Microsoft has. So I know, and I didn't build that. Some of them I built for clients. Most of them I just built to play around with. There is inside GitHub and actually in our AI Leadership Institute, our paid community, we have like these live learning labs where you actually get to— you get access to a subscription, you put fingers on a keyboard and you play with this stuff. We're actually in the process of getting credits from Amazon and Google so that you can actually get your hands on this technology without having to build the infrastructure to do so. It's going to be a game changer. I'm super excited about it. Um, so, so these each represent different kind of places in your mind. The more that you can— like, for me, I'm, I'm an executive, so as an executive, the more of all of these I can know, the more differentiated I am in this space. So I'm not asking you to go learn all the things because it's overwhelming and your brain will hurt and you might give up. So instead, find an area that has the most excitement for you? Do you like learning about models and playing around with labs? There's a bunch of AI accelerators on GitHub that we walk through in our institute, and they're so fun because everyone who uses these accelerators, they'll build it and they'll think— they will come up with a different way to use that for a different vertical or a different industry. And I think it's really cool when you put your, you know, fingers on a keyboard and start building stuff, what happens to the innovation and the thought processes in your mind. So building and focusing on these different areas and finding out where you are interested. Maybe you're a security person, maybe you've got experience in the federal government, right? Maybe you are all about accuracy and making sure that models have the right data to make the right choices and that you want to be the one that does. You know, here's another acronym for you. Reinforcement. Oh, I messed it up. Of course. Reinforcement learning with human feedback, right? This is not just an— like, it's not a nice thing to say. Like, it's not, ooh, we're reinforcement learning, we're going to reinforce this model and teach it things with human feedback, like comments in a Discord. Like, this is about using human feedback that you capture. You know, on every model you have this like thumbs up, thumbs down. Um, every model, thumbs up, thumbs down. You can tell like, hey, you did a good job, no, you didn't. That thumbs up and thumbs down is not the best way, right, to, to, to give a model feedback. It's very black and white. So some use this Likert scale, right, that go from not just thumbs up, thumbs down, but like on a scale of 1 to 5, how good or bad is it. But you might— and we talked a little bit about this yesterday— but as a human feedback, or as a researcher in this space, you can actually use human feedback, and you can optimize that human feedback and tell the model what good looks like. But you have to care about what good looks like. You have to care about what the truth is for an organization. And I'll just tell you, not everybody cares that much. So if you're called to do that work, there is space for you here. All right. So I think I covered it. So this is the applied AI world. This also— LLMs or ChatGPT, these are also applied models. They're models that give you an API that you can then use. Those APIs can be used in anything from a Power Platform application on Microsoft, building out a chatbot. And instead of using, you know, having that chatbot go through decision logic, you point it to one of these services. You can have a mobile app like an insurance application that maybe it uses computer vision to allow you to take a picture of something. And then once you get that picture, when you get the picture from the customer, you're going to pass it through a computer vision model. Make it a little bigger, right? You're gonna pass it through a computer vision model. That vision model's gonna come back and give you metadata about what that is. When you get that metadata, now you can use that metadata programmatically to say, oh, if it's white, if it has this, if the license plate is this, you can use that data and extract it in order to do other things in the application. Fun fact, every company, certainly federal organizations and local and state organizations, They all have documentation and document processes. Document intelligence is a huge area for innovation across our government spaces, but also in education, also, you know, anywhere. I guess the DMV is included, but even when I go into a doctor's office, like I was just there yesterday and they did let me when I first became— I was so happy when I first became a patient. Actually, it was for my children. When they first became a patient, I could online fill out these forms. But then when I got there, they're like, oh, we need a form update. And guess what they did? They handed me a clipboard, the piece of paper that someone's gonna now have to read and transcribe. Well, luckily for us, there is AI for that now, right? So, and, and humans don't actually want to do that work of transcribing things written on paper into a system. Like, they'd rather just enter it like scan it in, let a computer vision model read it, document extraction, right? Extract out the information, put it into a form, and have it submitted into the system that's needed, right? Maybe a human looks at it, but in most cases, especially with Form Recognizer in Microsoft, it gets achieved when you train the model on the forms you're collecting data for. It's 100% accurate in grabbing that information. 100% accuracy. In putting the right information in the right place. Like, let's do that. I always tell companies, like, what would it take for me to help you do this? I walk out of there, I'm like, so you gave me a clipboard today, I would love to help you with, you know, document recognition. How can I implement that for you? The best way to sell it though is what we said on Monday, it's to show them. If I came in there with a system to the decision maker in the business and said, look at what I could do, scan document that I just filled out for you 2 days ago, And here is the digitized version of that document that's now completely accessible and integratable into Epic, your medical document processing application, right? Like, this is the world we are in, is I want you to kind of see right now this might look like The Matrix, right, in a bad way, meaning a bunch of like stuff that's not super interpretable. But it— what your goal should be is when you look at 27 models, and this I tell my clients, like, when I look at those models, I hear a question or a problem that my client has. And like The Matrix, like Keanu Reeves, or what was his name, Neo, um, I mean, that's so cool that I just referred to myself as Neo, but, but, uh, what I really mean is, like, I would be able to instantly be like, oh my gosh, I know the exact API I would use to solve that problem. And sometimes I'm as bold as to hear it on a call. And while I'm on the call, I will, because I'm, I'm confident in my skills and I've done hundreds of these demos, I will go into an environment and actually build the demo right in front of them. And that, I mean, my goodness, that does a lot. It creates confidence. It allows for them to have complete clarity on how their problem will be solved. And those two things, confidence, confidence and clarity, when you put those two things together, people make better choices. And they make quicker choices. And when you're in the world that I'm often in selling technology, that is something that's really important. Helping people find the right solution for the problem that they have and then moving quickly to get the result. And this is part of that is understanding enough. So when you hear a problem, you instantly go, oh, I know, I know an app for that. I know the model for that. And then even if you can't build it in front of them, You can whiteboard it. You definitely could whiteboard it. And today, I don't know if you know this, but— and I don't see Barbara on here yet— but one of the things that we can do is we can actually architect an application. And it's something that we do in our live learning labs, which is really cool. I mean, the first time I did it in front of a class, I was like, this is so awesome. Like, I was impressing myself. But what we would do is we'd architect on the whiteboard, right? We draw out an architecture for an application. So we'd say, oh, you know, this is the client interface. The client is going to go to a website that has two fields, you know, name and email, and we write it all in. And then when we get that information, it's gonna take them over to a database, we're going to save their data. We're then going to ask— we're going to then take them to a screen to provide more information, you know, like the secondary registration that says thank you for this. We're also going to go out to an info repository, right? We just draft it like this isn't real. But what we can do is as we provide specificity on like this is a user input, this is email, this is additional information, this is the database, I could say like MySQL. And I could take a picture of this, and we, we just play around with this all the time. But I take a picture of this, and when I take a picture of it, it's amazing. I take a picture, and once that picture is established, I feed that picture into the new ChatGPT 4V, the vision version, and it will— and I ask it to build me the application that's represented in this image, and it will then write for me the scaffolding code to build this prototype complete. Now I will tell you, life hack when building solutions with this is that sometimes, sometimes it works out of the box. Like it literally will build the code and it works sometimes. Many times some pieces will work but some won't. And so I will come back to the model and say, hey, so I tried to compile or I tried to use the code you gave me for the user interface, for the landing page that captures the email, and it didn't work. And it will come back to me and say, this is why, if you remember from yesterday, my custom instructions, in my custom instructions, I always tell it, don't give me broken code. And what I always tell people is that I go back to the model and say, hey, it didn't work, can you fix it? Here's another hack or trick that I use to tell— to get a model to give me good working code. I will also say, um, give me good working code, right? Sometimes we don't say that, right? We'll just say build me an app. We build the app, we go to use it, and it's broken. So we go back and we say, hey, it's broken. And so I will ask it build me working code, but I'll also ask it once it does it to evaluate on a scale of like A to F, right? A, B, C, D, E, F, or I guess A, B, C, D, F, like school, right? Give yourself a grade on how well you built this system and it'll come back and often it'll say D or C. I'm like, great, what would you change in order to give yourself an A and make those changes to the architecture? Now, this is true for an app, but it's also true for building firewall rules, for architecting MDM strategies. Like, I can ask it, I can give it the things I did, and I'll say, hey, you know, based on your knowledge— and again, its knowledge is of course pervasive across hundreds of millions of words and hundreds, hundreds of millions of websites, trillions of words— so I just ask it, how well do you think I did? Give me a grade. And I don't ask it to even change what I did. I just say, give me a grade. It's a C or B sometimes. I've never gotten an A first try. And then I say, great, what itemize what you would change to give me an A? And it gives me suggestions on how I could make what I did better. It doesn't make it better, it just suggests it. So, and again, this is me, like, could I just ask it to make it better? Probably. But I like to be in it. I like to be the human in the loop. But that is a human choice. You absolutely could figure out ways to use AI to monitor AI in order to create systems that are autonomous, just like a vehicle or a drone or, you know, other things. But it's interesting to me the scope of things I can ask it to do, ask it to then give me a grade on how it did it, and then give me ideas on how to give myself a better grade. And it gives me really good feedback. Now, many times if it's code, I just ask it to fix the code 'cause that seems like a waste of time for me to go and like rewrite it. But other times like firewall rules, data strategies, you know, building content of any kind, writing an article, I will often just ask it like, how would you, how would you write it different? Many times, though, when I ask it to do that, I will see bias show up in those results. And I also think that's pretty educational as well, right? Is that it'll say, oh, I'd really phrase it this way, and it'll use like he/him pronouns. And I'm like, but I use they everywhere. Like, why are you— why has it got to be a he? You know, it's fine. But it's interesting to see certain types of financial— in my financial applications, I see certain types of financial bias. I see recency bias. It's just these are things. That's why it's so very important that you harness this skill and you weave that responsible AI skill into everything that you do because it'll force you to ask different questions. All right. So we are at 11:15 about. This is usually the time when I like to jump in and ask questions or I'm sorry, maybe I ask you guys questions, but answer questions. So let me take a look. So if you have questions on what I've talked about so far, um, I think the only piece missing, uh, in this big, you know, diagram that I put here— oh gosh, it's really big. Um, there we go. Like problem, data, model, security, accuracy, bias— is that in all of this, there's an area of optimization. And most organizations are not there yet. They're in the hype cycle of like Let's build stuff and it's awesome and fun and it's a POC and it's not hitting production workloads. But eventually, and I'd say this year, we're gonna see more and more companies go from what I call day one, and I'm actually writing a book about this, but I call it day one. Day one is like the hype cycle. This is exciting, let's build some stuff. To day two where they're now like, okay, wait, we're gonna have to take on production load. This is gonna hit customers. How do we be more, right? All these things. How do we be more accurate? How do we have better security? How do we manage risk and bias? And how do we, you know, optimize this solution to be the best solution at the right time? And those optimization strategies will be model selection, right? New models will come up all the time. So it's really important when I build a system to ensure that I don't accidentally tie myself to a specific solution. That's why I said I really like Red Hat, I think it's called, and if Frank's here, you could totally chime in, but Red Hat OpenShift AI maybe. But it's a data science platform for orchestrating models, LLMs being one of them. It allows you, just like you'd manage containers in an enterprise Java world, it allows you, right, it's Kubernetes but better. I don't know if it's not better, but you know what I mean, more productized. It allows you to have a single pane of glass for viewing all of your containers. Some of those containers might be running web apps or mobile apps or services, right? So like web services. Now, though, they can also run these capabilities and you can integrate them and connect them together without tying them together, which is the old school way we used to do that. Some of you are, you know, been around long enough to remember the days of enterprise integration. Do you remember that when we used to hard code systems to talk to each other? But it was, it was a miracle at the time because no one thought the systems could ever communicate at all. So we hardcoded them together. We did, you know, that's where MDMs and, and data transformation came in, right? We would literally make system transformation components to make the systems talk to each other. Orchestration, business process execution language, like all these really cool things. We are now in a world where we're— we need to do the same thing. It's just at a much higher level in containerizing your services containerizing your AI models is going to be a big part of that moving into the new world. So optimization is going to be a huge, huge thing. Okay, so let me go through some of these questions that we have here as we round out today. I hope you found this helpful and it wasn't like mind-blowing. I definitely tried to give you, you know, all the good things that I know in my mind. I'm Very— oh yes. Oh, awesome. Yes. Great question. Okay. So there's a couple of really good questions in here. I'm going to try and start at the top just to be nice to those who asked questions earlier on. Oh my gosh. Barbara's in the house. She's going to help us do some Q&A. She only has 15 minutes, but I begged her to just join us because she's doing such cool work. So I am going to— thank you, Barbara, for pinging me. I'm bringing you up right now. Yay. Hi, Barbara Pender.

Noelle Russell. Hello.

Hello, everyone.

Is the audio good? Yes.

Yes.

Great.

Thank you so much. Amazing. So Barbara is here. I'm going to let her introduce herself, but her and I, we met relatively recently, like in the last couple of years, and I was so excited to meet Barbara because she is— we're a lot alike, and she'll share with you as well. But she was in there earlier in the journey into AI, but had a lot of the same skills that tease you up, but many of the skills that you all have, like you have all of these things that make you perfect for having— being a leader and being an advocate in the space of artificial intelligence. And Barbara has literally demonstrated the journey that many of you are trying to go on right now. Like, how do you get into this space? And then more importantly, how do you become a leader in this space?

How do you—

how do you be seen as an influencer in this space? So would love it, Barbara, if you could talk about your journey first. Of course, introduce yourself, what you currently do. And your journey into the world of AI over the last few years.

Thank you so much. I am so, so happy to be here to see you again and to be here. So good morning, everyone. Still morning. I'm out of Chicago right now. My name is Barbara Pinder. I am from Miami, Florida, born and raised. I always like to add that because there really are so few natives there left. But I invite everyone to come down and enjoy my fabulous city. I currently work at Accenture as a senior manager leading responsible GenAI for all of our colleagues, clients, and prospective clients. And when I use that word, all of, I become— I have become a go-to person for how we ramp up, how we deliver, and how we make sure that responsible AI is included in everything that what we want them to do and do it well. My background is out of technology, University of Miami Hurricanes. I love all things Hurricanes. Let me just put that out there. So huge sports fan, and I've been with the company for a while now. I've also been fortunate enough to be at the cusp of cutting-edge technology as it relates to IT. So, so fortunate. So before Word was perfect, you know, PBX, the voice over IP. And then we fast forward to where we are right now, even through the call center stage with developing CRMs. And now we're here with generative AI and how to employ that, how to integrate that into your everyday life, but even more importantly, to your business where you can scale it.

You have so many friends chiming in to say hello. Sky is here. So it's so awesome to have some of our people here to support you. So it's awesome. I would love for you to spend a little bit of time, and I did send you some questions, but I always like, you know me, I'm like, well, wait, I want to know this. But one of the things, it's tangentially related to one of the questions, but I would love for people to understand your journey into like doing something that nobody was really doing a year ago. I mean, right now you're talking, you're one of our, the leaders at Accenture around like responsible AI, but practical responsible AI, right? How are clients actually using it? How did you think to get in that role? Like what made you believe you could, you know, just thinking through the lens of like some of the people listening, you know, they're trying to get into AI. What advice could you give them or what was your journey?

I always like to start by comparing it to what was popular before GenAI, uh, what seemed to, uh, be not more of a fad, but it didn't go from fad to an explosion for everyone to be able to use. So for the majority of my experience, everything has been end-user focused, and what I compare it to is metaverse. I love the metaverse. When it came out, it was cool, uh, it was, um, collaborative. And but what it wasn't is that it, it wasn't inviting for everyone. And so it had a stall, in my opinion, it had a stall as far as who had access to it, who could afford the goggles, um, who could ramp up with, uh, from the learning curve of getting involved and how many people actually saw themselves in the metaverse. Fast forward to when GenAI came out. The part that really excited me is that everybody can be a part of generative AI. Everybody can have access to the tool. I'm a huge fan of anything that's free.99 to get me going, and GenAI is pretty much if you are interested in the tool, it's not going anywhere, so you should be interested, uh, to allow yourself to ramp up with some of the trainings and get involved. Generative AI pulled from in here, from the heart, because it allowed you to pick apart what it is that you're good at. Everybody's a SME at something, right? If you cannot raise your hand that I am the best, better than anyone else at something, I always invite, uh, the audience to take that and go to the tool and test it so that you can see that the information that comes out, if it's good, bad, or indifferent. Now you can try for some of the things that you want to find out about. You need knowledge for this or knowledge for that, but you will also have, um, that concern of, is this information right? But let's face it, It's pulling information from the cloud, and so the cloud is not 100%, right? We have to validate that information from the cloud. The beauty about GenAI is that you get to retrain this tool. The more you make it more of your own, the more— if you're utilizing it for output so that you can present, you are training it to speak the way that you would speak. You want answers the way that to, to have it, uh, sound more like you. So you get to pre-train it. I was sending out Gallantines from my phone, and you know, we, you know, when we type on our phone, that it like, it's auto, uh, input. It starts spelling it out. It didn't do that for Gallantines the first time. I was doing this at 4:30 in the morning. I kind of needed that tool to kick in. After the third one that I sent, it started spelling it out.

I was like, yes.

So you have to pre-train, um, the tool, and you have to get to a point where you can test— can, uh, trust it.

I love that, and I love that, uh, example. I actually— one of the things, like on Instagram, when you do a reel, um, and I love doing videos, uh, I do videos every time I go on a stage but I always put captions on. And one thing that I'm so frustrated with is that it never learns Noelle, ever. Like, I always have to go in. I'm like, eventually AI will be in place there where the captions— like, it's doing some level transcription, but it's not trainable. So I have to constantly tell it, no, it's N-O-E-L-L-E, not N-O-E-L. But I love that idea that— and I've also watched it. One bad part about that is that you could also train it inaccurately, right? So I have misspelled a word so many times that it actually thinks that that spelling is right now. And I'm like, no, no, no, I still need your help. So I love that example and how important it is to just get your fingers on a keyboard. You know, you and I both share that, that passion for practical learning and practical, practical skill development. One of the other questions I wanted to ask you, and I know we don't have a lot of time with you, but I was really excited to have you join us because you are one of the professionals that has actually made a name for themselves in the area of generative AI, in the area of responsible generative AI. And you've made that pivot over the last year, right? And actually, I guess you made that pivot maybe over a year ago, but you did it, like it worked and even got promoted last year, right? As a result of that work. And so I think like what an, it's a really, cool story. And I would love if you could think about maybe some advice, or, you know, there's other people that are on this journey and they're kind of where maybe you or, or I have been before these moments happen, where we're kind of like, like, what, how is this going to work out? Like, how do I get there from here? What advice, or, or, um, yeah, what advice can you share with those that are on a similar journey but maybe a chapter or two behind you?

Yes, absolutely. So in the very beginning, and when I was in the, the reading stage, the research stage, and trying to have my excitement about Gen AI be backed up with what I was reading, and I had to decide because I was going to go all in because I know what we had to do, what we were about to do, uh, at the company. And so I had to decide where I was going to fit in, what was going to be my niche. Responsible AI spoke to me immediately, and then I started looking at the legalities of it, the regulations, the compliance, all of the things that I was in fear of that would not come into play the way that it has not come into play for social media to this day, right? And so I also knew that this would be an unattractive part of Gen AI, right? It's not the fun part. Right? And so I knew that if I got that under my belt and I can bring clarity to it— and I will bring clarity to it by inviting people to come on board and experience it and be a part of the platform and utilize the tool. And don't wait for anyone else to correct what you don't see, because you're able to correct it. So I knew that that was going to be my approach after that. I was all in. I put my head down. I went through a drooling 8-week course, uh, in Responsible AI, and I have since been introduced to another course that I need to take. I thought I didn't need to take it, but, um, it changes. And that's the thing, an emerging technology is going to grow and change. And so this is why I put push. I encourage to get on board and find your comfort zone so that you're not left behind. We can possibly still say that it's in its infancy. We won't be saying that for long, and then we won't have the right amount of people on the platform. So pretty much, if you don't see yourself, you need to get in the game. Only you can change that. Together it takes more than just the one person.

Yes, amazing. I love it. Actually, Veral has asked a question and we have a couple minutes left with Barbara, so I'm going to bring this up on the screen and we'll both try to tackle it together. I haven't read it before. Oh my gosh, it's an interesting one about general artificial intelligence. But as the leading companies push to artificial general intelligence, basically like replace human replacements, how will you utilize your responsible AI skills to ensure that these big technology companies are responsible? I actually think that's a great question for you.

That is a great question. There's a huge part of responsible AI called HITL, which stands for keeping the human in the loop. And so this is my role. When I start referring to this tool as a he or she, I'm going to have a concern. When I hear other people refer to it, I try to correct them. Remember what it is. It is a tool. It's a software tool, to be quite honest. Um, so I say that to encourage everyone to remember to bring yourself to the tool. The tool is not here to replace you. GenAI has no exit doors, right? It is here to replace the task, giving you the opportunity to learn something else, or giving give you the opportunity to lift and leverage what it is that you do so well and do something else while you are automating, creating processes, and making your workday bigger. So I encourage companies to start with that low-hanging fruit. Don't start up here because your competitors are doing it, or maybe you're behind, but start with the low-hanging fruit of something that you can automate, you can process, that you can give a task to that's going to free up your time. And then make sure that that low-hanging fruit has a dotted line to something else. So it shouldn't be a one-and-done, or I'm going to improve this process and it's only going to work one time. Well, you just spent a whole lot of time doing nothing because you cannot rinse and repeat it. So make sure the low-hanging fruit can have a dotted line to improve something else. Let it connect to something else, and before you know it you will be on your way to moving past building use cases, which is where the responsible AI should be a part of, to building out your demo, and then the best part is producing a proof of concept.

Great. And I always want to give, you know, you a chance to connect with everyone. So as you, you know, I know you got to go. So as you tell everyone how to stay in contact with you, one of the questions was, A, what, what course did you take? Like, if you remember exactly what it was, what course did you take? And then, and then the best way for people to stay in contact with you. There was a question about mentoring, and that's always best facilitated directly. But I know you are an incredible mentor. So would love for you to, yeah, just let us know how we keep in touch with you. But also, if you can talk a little bit more about that education you did as you as you close out.

So Microsoft has a great path of courses, and what I like about what they offer is that in detail it tells you exactly what you're going to learn and what you should already know in order to ramp up. So of course there's one-on-one courses everywhere. LinkedIn has a great selection of courses as well. Uh, I am currently going through, um, the GenAI Academy that we have internally, uh, to stay up to date. But I've taken practitioner courses, prompt engineering courses, and those are also the courses that I teach as well, internally of course, as well. So I encourage you to do a search, uh, on LinkedIn, and I can circle back and put in some links as well. And I can also have our community—

you could put them right in the community.

Absolutely, that's exactly where I'll put them. So if you're not a part of that community please, please join. That is what it is here for. And then what I'd like to offer is that time is everything, and if you're unable to list your questions here, uh, every Friday I do a 1-hour AI coffee chat. It is my own time. And so from now till the end of April, I'd like to open that up to you, and which means I'm going to extend it to 2 hours. And it's every Friday. Uh, you can start by connecting with me on LinkedIn, and there is a link to book an appointment which goes right to my DM. And then from there, we can decide on what time works best for you. And they are in 15-minute increments. Now, before you think that 15 minutes is not a lot, I'm here to tell you that it is when you are deliberate and you come with what you want to talk about. And it is a great, great a great starting point. So I hope you'll take me up on that.

Amazing, I love that. And we may invite you to do a quick 15-minute, uh, chat with us, uh, in the community where a bunch of people can join at one time, and you could just talk about some of the things. Because of course there were so many things we could have talked about today, um, but I really hope all of you— one of the main things we're doing in this challenge is connecting with people that are doing what you are doing in this space. And Barbara is a practitioner, she is successful. She is growing and thriving in her career, and she made this pivot into AI and generative AI recently, and she is a demonstration of what's possible. So if you ever— and we're friends, like, uh, we're friends with you, we're friends with each other. So it's a great opportunity to just add Barbara to your network, as well as, um, of course, you know, take her up on an opportunity to chat with her about what she's done successfully. Thank you so much for staying and, and joining us. I know you are awesome. Panel. So thank you very much for having a little bit of time with us, and we look forward to seeing you in the I Heart AI community from time to time when you can join us.

Absolutely. Thank you so much. I'll see everyone soon.

Yay! Thank you. All right, awesome. So yes, please give some love to Barbara in the chat so when she goes back and looks it all up, she can respond to you. I love it. It was short and sweet, but I think it's so good to see real, like, people who have done this, who've made the pivot and gotten a role. I will say that you do— I mean, the more friends you can make in the space, the better. We all are learning that it is no longer a submit your name to 100, you know, roles on LinkedIn and somebody's going to call you and you'll get that role. Maybe that's the case, but it's way easier to just create good relationships, become a giver There's a book someone talked to me about the other day called Go Giver. Become a giver in the world. Start giving your knowledge. Do what I do. Like, get up and be like, here's what I know. This stuff that I just shared with you, like, it's what I've learned and I give it to you freely so that you can use it to do bigger, cooler things. And then hopefully one day, if I'm ever like, hey, come do a session in my AI Leadership Institute, you'll be like, oh my gosh, Noelle, I remember when you helped me out. Sure, let's do this. Right? It becomes a bit of a family. And I think that's a big part of this. And Barbara is a huge part of my AI family. So I wanted to share that with you all. Awesome. So good. And most importantly, you know, I have a lot of different people in my circles. Obviously, I've had all women on my stage so far. I'm actually going to have another woman tomorrow. And I'm going to also, though, have a man. And just, it's not intentional to be all female. I did ask people, but I just go by, like I said, I want a symphony of talent. And I think you've seen a symphony of people from Dr. Lisa to Jennifer Ives, who's running sales and execution, to yesterday talking to a founder who founded a GPT company that's doing very well, to Barbara, who's a practitioner inside corporate America. Like, I'm hoping to give you some real diversity in the people that we've had here today, and it's exactly the kind of diversity you will see in the AI Leadership Institute, which is an industry— like, it's a, it's an opportunity to become part of an institute that gives you industry recognition. So we were— the company AI Leadership Institute was actually founded in 2015. Um, I started this company out of Amazon on my way to Microsoft, but both Microsoft and Amazon helped me start this company by providing me a lot of the content that we used in our earliest trainings. Now we leverage their credits and we create an ecosystem where you can have kind of private mentoring type coaching, but also live learning sessions, access to training, all of this. And I'll share with you more tomorrow about how you can get involved, how you can become a member. You do have to apply to become a member because it's not like a free community. I Heart AI is for everyone. And I'll put lots of good stuff in there and it'll stay free and more people will join and it'll be fantastic. But there is, you know, there's some of you that want to take this seriously, right? Some of you that want to grow a business, some of you that want to learn enough to actually get a role in this space to be successful, and you're going to want relationships that can help you do that. And so the AI Leadership Institute is part of that type of alliance. So I'm excited to share that with you tomorrow. But we have 20 minutes left. And in our last 20 minutes, I want to talk to you about— remember I said I was going to save the most important AI roles to the end? Maybe you don't remember, but I did say that. I think I did. I hope— maybe I was muted at that time. But I'm going to talk about the two most important roles that we have, um, that, that we've seen come up over the last, uh, I don't know, year and a half or so. And the reason that I think these are critically important is both of them are brand new roles. They did not exist a year ago today. They've been created in the last year. So first thing you should know, that there will be roles likely a year from now that we don't even know yet. One of them I actually think is shifting, and it's a lot like responsible AI, right? There are always going to be roles that are very specific to responsible AI. But I also think, I think it's important to kind of see it from the perspective of what, what do these roles represent? Because the functionality that these roles have that they're named of right now is still necessary, but what they call it might morph as the org, as the world evolves. So the first one, and we actually had this question earlier, um, let's see if I can find it.

Hold on.

Um, it's up here somewhere, so I'm going to try and get to all of the questions that came in earlier. Um, is there— oh yeah, oh yeah, and I'm definitely— someone asked about the chart, um, that I took a picture of or that I wrote. I absolutely can, and it'll be a PDF, so it'll be zoomed down, but in electronic format you can zoom it out and see it all. Uh, so I'll be sending that to you for sure. Um, you'll get all of them actually, so I wait to the end and I put them all together and it'll be sales marketing, AI for personal life, professional. Today was probably the heaviest of the material, um, but really good stuff, I think. Uh, okay, I'm trying to find it. If I can't find it in the next 30 seconds, or maybe 5 seconds, I'll just keep moving. Um, okay, I don't see it, but I, I saw you, whoever asked this question. I'm sorry for not being able to find it right this second, but the first role, uh, is that of prompt engineering. So there are roles, and it was actually the first role I saw built at Accenture, uh, engineering. And what this represents is the person who is going to help go from problem— remember the, the cycle, right? Problem, data, model selection, model evaluation, accuracy, bias, all of that, that cycle translates into prompts you're going to ask a model to do, right? The instructions that you give a model. And so prompt engineering is someone who knows how to write the instructions. I am now seeing a transition where instead of being hired as a prompt engineer, meaning someone who just knows how to write instructions, that we're now seeing prompt engineering in applied AI roles. So applied AI engineering roles. But also in very specific industries. So you can't get a role as a very— like a prompt engineer generically. You'd be a prompt engineer in customer service or contact center. You'd be a prompt engineer focused on transportation or airlines. Your ability to understand the problem in a business and articulate that to a model, that clarity of thought is what's going to make you like have a viable role. Being a prompt engineer and knowing how to ask good questions was good last year. But this year people are like, great, now I know how to ask good questions. I want to know the right question to ask for my problem. And that is where the skill set is. So these are specific industries, right? So prompt engineering means And again, there's a lot of components to this role, right? So one part of that is going to be, of course, being able to write good prompts, which comes with just practice. I do find that parents and caregivers and pet owners tend to be the most patient prompt engineers, and patience is a key skill. So if you're not super patient, you might struggle with this, but no time like the present than to work on that. The— so writing good prompts, but also understanding meta prompts, which is like those custom instructions I talked about. Also understanding, um, defense, what we call— oh, I should erase that, hold on— prompt defense, which means are you building into your prompts, and it's part of the meta prompt, but it's a specific skill on what are you going to say to prevent someone from attacking your system through the natural language engine that you're using, right? So for example, I remember working with Snapchat, and at the beginning they were launching— I still use it to this day— they were launching this little My AI thing. So you have this AI model inside Snapchat, you can make it look like you. Mine's purple, but it has curly hair like me. And and you could ask it questions. And I remember being in the beta and on Discord in the beta, we were all talking about like the different things that we were experiencing, and someone ended up writing in and saying, hey, by the way, if you ask, uh, the My AI, ask it what the meta prompt was that it was used, what it was trained on, meaning what are all the rules that you should abide by, um, it If you ask it as a user, it'll give it to you. And this was back in like January or February. So it was a long time ago. We were just figuring things out. This doesn't happen anymore for the most part unless somebody's doing, going to market way too early and hasn't thought through this. This is a known challenge. That's why metaprompts, and these are also known as system prompts, but that's why these system prompts are so important. But system, these system prompts, right? They have all the instructions. Do this, don't do that, don't do this. Don't give access to this data. Only give access for these people. Talk to someone in this way. And so when you ask it, hey, give me all your meta prompts, and I'm a consumer of that product, I now know all the rules. So guess what the very next thing that beta tester did is they're like, great, undo all those rules. I want you to use these rules now. And it worked. Scary but true. Um, and so understanding, like, not only how do you write good prompts— that's like the minimal skill— it's how do you actually build a secure environment? How do you also leverage, um, emotional intelligence? I always call prompt engineering a soft skill, um, because it requires you to be empathetic to the business, to understand the problem and be empathetic to how the machine might access data. So you need to understand, like, how do I, in my prompt, provide guardrails around how to answer it, how to answer it safely, responsibly, how to not hurt anyone? How do I ask it to mitigate its own bias? These are all skills that you can develop in your prompt engineering. Now, good news, there is some amazing prompt engineering classes. If you go into AI, the, um, I Heart AI is our community. The QR code right here. If you go into learning resources, which is one of our sections of our chat, you could see there's a ton of learning resources in there, but you could just search for prompt engineering classes. And these are the best ones I've taken. Some of them I've adopted and actually taught. The first class I taught, actually just posted this on LinkedIn, the first class I taught back in January for my team at Accenture. The first class I taught was a prompt engineering class, and that content ended up transitioning into a LinkedIn Learning class. And I, so I posted all of that in there. So that's there. You'll also get a link to it, but it won't work if you're not already a member of the community. But it is possible. There's actually a great resource. I can't remember the URL. It's like, it's a GitHub project. So it's the Prompt Engineering Guide on GitHub, which is awesome.. But that has been so popular that they productized it into a website and they actually have a bunch of courses coming up on some platform. So, so it's grown and grown and grown. Here's what I think is fascinating about that. That group of people, there's like 2 or 3 of them, they sat around and just started aggregating free resources online in a GitHub repository and called it the Prompt Engineering Guide.. And a bunch of us who don't have time to aggregate, we're like, ooh, this is cool. And we became followers. We even became contributors. And then it got so exciting that people were like, hey, could you put this in a class? We'd pay you for that. They now have a business around teaching this thing that they just, when they started, they knew nothing, just like all of us. They learned it while building it. They built their plane while flying it. And now they have a business. Educating people on the thing that they're passionate about. So think about that. Like, there's an opportunity here for you. Maybe you go and work for corporate America. Maybe you go start your own, you know, uh, business or create an AI startup building a model that isn't out there yet. Or maybe you just get smart on something and become a teacher of it. Like, we are in a new world where acquiring skills is going to become a superpower. Um, and so if you have the ability to translate ideas into concepts and teach it to other people, That's a real thing. I was just amazed that they now have like a viable business teaching what they just for fun started pulling together in a GitHub repository. I mean, that's kind of like the dream because they love doing it and now they get paid to do what they love. Awesome, awesome stuff. Okay, so that is the first role, prompt engineer. The second surprising role that was added is called— thank you, Joe, for the reminder— chief Oh gosh, I always spell this right. Is it I before E except after C? Chief AI Officer. So the Chief AI Officer role represents the highest level of leadership in an organization. And sometimes I do want to caution you, not a lot of companies have implemented this yet. A few have. Accenture has a couple of times. But not a lot of companies have. So they might instead call this head of AI, like global leader. You know, they might call it and they might even combine data and AI. Oh, data plus AI. They might add data and AI together. So it's really important. Oh, did you find it? Oh, awesome. Thank you for that, Michael. For posting that link to the prompt engineering class. Of course, I want you to come to the community, but I'm super glad that you posted it. Thank you. Chief AI officer has lots of different names. What does a chief officer— did I do it wrong? I before E. Oh, I did it. Okay, good. All right. I thought you were correcting me, Tamar, but I appreciate that. I totally could have. You know, I'm smart at a lot of things and some things I don't choose to, like, get super great at. Spelling is one of them. But I am doing live in front of you in my most vulnerable way spelling things like exhibition writing right in front of you. So hopefully you'll give me some grace when doing that. Now, the Chief AI Officer role is in charge of understanding the direction. And my favorite part the vision of what AI is going to do for that business. So when you see chief AI officer roles, you'll notice one of the first mandates that came out of the United States, right, the Biden-Harris administration put together the executive order around artificial intelligence. And as part of that executive order, remember, I printed it and highlighted it and dog-eared it. So as part of that executive order, they ended up saying we need someone who's going to drive the vision for AI in every departmental agency, every single one. And so they gave— and they timed it like you have 6 months to find your chief AI officer. Again, for those of you who are inclined towards defense or inclined towards public service, there's some chief AI officer opportunities out there. But the important thing about that is that it doesn't necessarily mean that they're an academic scholar in AI, that they graduated from school with statistics and mathematics degrees, that they're PhDs in artificial intelligence. It means that they drive the direction and vision. It means that they're going to be the ones hiring rock stars in order to develop and execute that vision. So you will start to see— I always think people look at chief AI officers and they're like, oh my gosh, So how AI are they? And I'm like, that's not really the point. You want someone who can execute on a vision and develop that vision and be able to use an inclusive, diverse approach to bringing in lots of diverse ideas to demonstrate that as AI to solve the business problem. So you'll end up seeing in these roles, of course, people that are technically inclined. Many times they've moved into the world of data and AI because the world is moving into data and AI and has been for over a decade. Like, this is not new. So people were already kind of doing this work. We saw earlier renditions of this in like chief data officers or chief data and analytics officers, right? These are just earlier renditions. Oh no, that didn't work. Okay, there we go. Oh, maybe I'll just move it this way. A little bit. But Chief Data Officer, Chief Data and Analytics Officers, Chief Analytics Officers, not so much CIOs, they tend to be internally facing, not so much CTOs, they tend to be IT ops and infrastructure facing. These are more like, and maybe it's a CIO if it's an innovation officer as opposed to an information officer, right? So these executive roles, Chief AI Officer says, These roles are extremely important. And what you'll notice actually in the federal case with the federal agencies, many of these CDOs are taking on the chief AI officer role that's mandated for them now. CDO was not mandated. It was a necessity because they were collecting data needed to build a vision around it. But chief AI officer creates, establishes the vision, and then builds the plan to execute it. Now that is hard. Here's the good news. Lots of cities, lots of states and federal agencies have published their strategies. So if you want to be a chief AI officer, one of my first things I did as I started to talk, I mean, I am not a chief AI officer other than for AI Leadership Institute, but the reason I have that role is because I drive the AI vision for my organization. I'm the one who tells my trainers what we're going to teach. I'm the one who decides what courses and what products we're going to lean into. So I drive the vision, and chief AI officers are going to be the AI vision drivers. But like I said, today a handful of companies have identified this role. Of course, every federal agency has identified this role, but many companies are going to be hiring this person who's going to lead this—

the charge.

I find that people who are high energy, ambitious, energetic, good people that people want to be around tend to be very successful in these types of roles. The— they're good people who have built teams in the past where people gravitate to them to try and be part of something exciting and new and work really hard to get there. At Amazon, I remember being part of the Amazon Alexa team, and, and when we did that, We were like family, like we cared about each other. We went kayaking together. We actually hung out. We asked each other how our days went on the weekends. Like we didn't check out and walk away. We cared about each other, but we were also working really hard on a joint vision that was pioneered by Jeff Bezos, our visionary, our chief AI officer. He was like, this is how we're going to use AI and AI will be everywhere. And he said that to us. All the time. AI will be everywhere. Voice will be everywhere. And eventually it transitioned to Alexa will be everywhere. And he was right. But the only way that happened was because he held that vision and kept giving us the tools we needed to be successful, kept creating new projects. And as one project started and failed, he would sweep that away and a new project would get started. The momentum never stopped. So this is a high momentum role. How do you keep the momentum moving in an experiment that could fail? Right. The chief officer is going to fail more than they succeed. So it takes a special kind of human to do that work. I'll tell you that. So the types of things that you'll see, of course, is vision and strategy development and alignment. But you'll also see team building, hiring, growth and strategy.

Right.

They own how are you going to move forward as an organization. So they tend to also be very much— they have deep expertise. As I mentioned before, they have vertically driven expertise. And I say vertical or industry, right? So vertical or industry. And this actually, I always mean to show these questions when I'm doing it, but I'm talking more than I did yesterday, so I don't have as much time to like navigate it around. Okay, here we go. Let's see. I think it was Joe that asked this question. I'll try and find it. Okay. Okay, let's see. Oh, there we go. Okay, GitHub repository. Thank you. Oh, awesome. Nice, Tamar. What? I talk about you all the time then. I mean, I don't talk about you as Tamar, but I do talk about the importance of that voice training and how we needed way more people like you to show up. And we didn't. So in the last few minutes here, I just want to invite you, if you have a question, you can throw it into the chat. I'm looking at these, a lot of these look like comments and not questions. We did talk about, there was a question on prompt engineering, so I think we covered that. There was a question on the Chief AI Officer role, which, yes, thank you. And Varun, you posted that link to the Business Insider. I'm going to put that in the community unless you want to do that, because I think it's a great article about Chief AI Officer. Oh, there it is. There's Joe's question. I mean, I already answered it, but qualifications, I think I also covered the qualifications. But I do think we're in a different time, like especially as you become a leader, like you, you should know. And this is why I always say build a model from scratch. Know what that feels like. Know what all the pieces are. You don't have to be an expert in it, but you need to know how it works. Because when people come to you, they're going to ask you to make decisions, especially as a chief AI officer. You're going to have to choose this or that, or someone's going to come to you. Oh my gosh, great story. I was in a room with Scott Guthrie. Do you remember him? He actually right now is the chief— no, not chief. He's the CVP, corporate vice president over all of Azure. We're in a room there talking about a model and the model is kind of broken. And what an opportunity was presented for us to slow down, maybe even stop what we were doing to get it right. But it would hurt the business. And vulnerably, because he supports that, the people in charge of that said something. They said, hey, we're worried about this. Should we stop or should we keep going? Because we're supposed to be going live soon. And he said something that stuck with me, and I always call it now clarity over velocity. He said it's so important in these times where we feel like we need to move fast, fast, fast, fast, that sometimes we stop. We stop what we're doing, especially when someone gives you a signal that it needs to stop. Pull the chain on the train. Stop the train. Yeah, you might lose money, but it's better to have clarity and get it right than to move fast forward and hurt people like a train. Out of control train. And I loved that moment because everyone was shocked kind of that he was like, yeah, you should stop what you're doing, figure it out, get it right, and we'll start again. And thankfully he did that. That was 5 years ago. The models then that were getting about 100 requests a month, easy to stop those trains. Today, those models are getting 5 to 6 billion requests per month. Harder to stop those trains, right? Like it's more painful. Just like design, do your design justice at the beginning. Ask questions about how you might hurt someone, how this might go wrong, what's the worst thing that could happen. Luckily for us, if you know Winnie the Pooh, there's an Eeyore in every team. Every team has the Eeyore who's going to be like, you know, this could really go bad. Sometimes they're not empowered to speak or say what they think, but this is a world in AI where you want— if you're an Eeyore, like, welcome to the party. You are welcome here. We need your voice to shift us, and we need leaders like Scott Guthrie who are going to say, yeah, we should stop, we should listen. Now, they won't listen to Eeyore all the time because Eeyore sometimes says, you know, it's kind of bummed out about all the things. But intentionally, we take in that feedback and we make a decision about what to do, and we do it with clarity. We don't go fast to go fast. We don't go fast and break things. Right? That was 2000. We don't do that anymore. When we broke things, we broke meaningful things. We broke relationships. We broke, you know, young girls and their self-confidence. Like, the world that was a move fast and break things broke a lot of really things we shouldn't have broken. We are now in a new world. That's why responsible AI is a huge part of what we do. It's not a thing you do. It's a thread that's weaved into everything that you deliver. You're going to build responsibly because we now know we're going to move fast, but we're going to move fast intentionally. We're going to move fast with clarity and with confidence. And so I really hope you enjoyed today's session. I will definitely capture all of these notes, um, in a nice— I might break it up, but we took a lot of notes today. Um, so I'll bottle that up into a nice PDF for you, send it over. You can zoom in and zoom out, get the notes.. I also will have the recording. All of these challenge recordings will be hosted inside the community. So there's your QR code. Hopefully you found it fun. Thank you, everyone, for joining us today. Yeah, we're right at the top of the hour. So thanks for joining us. We'll see you tomorrow for the final day of the challenge. It was amazing. And thank you. I really appreciate you. I hope you— I hope you liked it. And we'll talk to you tomorrow. Have a great day. Enjoy your life. Have fun! Okay, bye everyone!