[Ep 2] Good Morning, AI : Navigating Purpose, Passion, and Pitfalls: A Journey through AI Ethics,...

Hello and welcome to this episode of Good Morning AI. So excited to be here with you today. I have tons of things to update you on. So today, um, I'm gonna cover, uh, basically a collection of things that I want to talk to you about. The first thing we're going to go through is, um, my take on leadership. So there's— every time we do an episode, I'm going to give you a little bit of a taste of the types of things I do to do the things I do every day. Then we're going to talk about AI for good. We're going to then go through AI news and updates, which I got a couple of great ones this week. And then I'll also talk about AI gone wrong. Excellent, horrible example of AI gone wrong that I can share with you today. And then also finish up with where in the world is Noelle Russell going? And I'll give you an update on all the things that I will be doing and going and talking and traveling to over the next week or so. So, so great to be here with you. Thank you for joining me. For those of you who don't know, my name is Noelle Russell and I lead the AI Leadership Institute. And what that means is that I help companies go from the boardroom to the whiteboard to the keyboard, helping them understand how to leverage AI, but not just in the enthusiasm that AI brings, but to take that enthusiasm all the way to execution. How do we go from the playground to production with generative AI, artificial intelligence, and all of the different types of technology that I'm exposed to every day. So I'm looking forward to sharing some of these lessons with you. The very first thing I wanna talk to you about today is all about leadership. Every episode will have a little section on leadership principles that I care about. I created a leadership philosophy called Lamplighters, and Lamplighters is actually an acronym. I have a podcast of the same name. So you can go take a look at it. It is called The Lamplighter Effect, and you can go find it. But these leadership principles actually came from me being just really interested in understanding kind of the qualities of good leadership, especially in the age of AI. And the Lamplighter Effect was the idea that you can't really learn these leadership principles and let it ride. Similar to AI, you can't just do something and walk away. It's a constant and continuous evolution. In terms of leadership, it ends up being an evolution of thought, of idea, of self-discovery, allowing you to bring your very best, your authentic selves as they say, to the work that you do. So today, I'm going to talk to you about one of these principles. And again, you could go check out the podcast. It's called The Lamplighter Effect with Noelle Russell Russell. So, you can go check it out. But today, I want to talk to you a little bit about the very first leadership principle, which is L, of course, lamplighters. And that first leadership is about love, being loving, coming from a place of love. Now, before you get all worried about the direction this is going to go in and I'll get all mushy-gushy, I think it's important to realize how valuable and critical your passion for what you are about to undertake with an AI project, how critical that passion is. Now, I don't know if you know this, but if you go and you watch some of Sam Altman, the CEO of OpenAI, if you watch some of his content, you will find out that he will tell you how long it took ChatGPT to become successful. And he will say that it took them about 7 years. 7 years of walking in and having something that wasn't working. Wasn't delivering on the promise, wasn't delivering on the vision that they had for what they wanted this technology to do. 7 years of going into the office day after day and hitting your head against the wall being like, "Why doesn't this do what we want?" Now, that's a long time. And when it comes to AI, many of us, we get all excited, right? The enthusiasm kicks in and we go and we try and build something and it doesn't work exactly as we expected.. It doesn't happen overnight, right? We're like, we're surprised maybe that this AI model can't completely do everything we want the first time we ask it to. But the reality is, is that most, all, but most AI projects, they require someone who's willing to do the work and stay for the duration. In other words, when you are building an AI system, it won't work at first. It's almost always broken. And your job is to pursue this vision of the outcome that you want this model to produce for you. And the more passionate and purpose-driven you are in that outcome, the more likely you are to do what it takes to succeed. Now, I'll give you an example that I uncovered during COVID actually. An example of a research organization, actually a couple of researchers uncovered that there were some very large companies in the world building computer vision models. And in using these computer vision models, those models, well, they were flawed. And their flaw was that they were over-indexing on white men versus being able to see color at all. Not only color, but gender. And it was a significant deficit, right? These models were not performing well because if you asked it a question, it would give you information. Like let's say you said, hey, show me a CEO. It would show you pages and pages of what a CEO might look like. None of them were brown or Black. None of them were women. And so, they uncovered this inequity in these sets of models and went to these companies—Microsoft, Meta, Amazon—and said, "Hey, your computer vision models can't see color. You need to fix that." Now, here's what need—how this relates, right, to this leadership principle of love. If you love what you do, if you are passionate and purpose-driven in the outcome, if you want to build a computer vision model that sees everyone because that's your core value— I want everyone to be seen. I want to create democratization. I want to create equity. If that is a core value of your organization, then you're going to go back and you're going to make it right. Even if it's expensive. And I always say diversity is expensive because in order for me to build a model that sees color, I have to go and collect data that is colorful. And collecting data that's colorful is more expensive because it is not readily accessible. I have to go to those communities. I have to ask for them that data. Sometimes I have to pay for that data. Maybe I should pay for that data. But that is the concept of being in love, like having passion for what you do. Because in artificial intelligence, It is that passion, that purpose that is driving you that will allow you to stay the long— till the end, will allow you to stay until your model works and works for everyone. As opposed to running into a hurdle like this, being presented with this research finding that says, by the way, your model can't see color. And if it's not tied to a core value, what will you do? You'll probably just quit. You'll probably just shut it down and be like, well, that's not that important for us to fix. So it's critical as you go on this journey of building artificial intelligence that you go down a path towards an outcome that you truly care about, that you'd be willing to sit through 7 years of failures to achieve. Fun fact, I have uncovered quite a few people who took 7 years to build what they were building. I am a big fan of Lin-Manuel Miranda. Oh my gosh, so much so a fan that I actually have a quill from Hamilton and I use it on occasion. I'm like, I like the lore and the backstory of Hamilton. And I found out, guess what? It took him 7 years to create that musical, that disruptive, life-changing, Broadway-changing musical. I also found out recently, I was at an event called GrowthCon with Grant Cardone. In that, at that GrowthCon event, Tyler Perry was there, and in his interview, he also shared that it took him 7 years to create his first Broadway show. And I just thought it was fascinating because so many times, I mean, the reason that we are willing to stick with something that long to see the end is because it is part of who we are. It is tied to who we are at our core. And that's what love in The Lamplighter's Principle really talks about, is how important it is to have clarity on what you care about. Because as you move into the world of AI, what you care about will be amplified. And if you are not intentional about it, what you amplify may create unintended results. So, something to think about as you go on your journey. We call that section the "I on Leadership." What are we thinking about when it comes to being a good leader in this space? And I will tell you, having a really clear understanding of what your core values are will help you be a better leader. In the area of AI. Now, the next section is a bit fun. It's called AI for Good. Our AI for Good section is all about good news in the world of AI, where we're seeing AI used for good. Many times, though, I use my own personal life experience to share with you ways that I'm seeing AI for good. Now, I recently bought a new product called the Hearth Display. I think that's how you say it, or Hearth, I don't know. H-E-A-R-T-H. But the Hearth Display is how I think their videos say it of themselves. And this display, it's a big display. We've mounted it in our hallway on the way to the kitchen, and it allows us to integrate calendars together so that we all— it has calendars, it has tasks that we can put on, chore management, executive function type things to help my kids be more independent, letting them see that they need to brush their teeth and make their bed and do the chores and clean the table and, you know, do the dishes, vacuum the floor, right? These types of things. So what I found out though that was delightful is that there's a feature of the Hearth display, and at the bottom, and it's called, um, Hearth Helper, I think it is. Uh, and so this helper, it sent me a text message when I configured the application. And when I looked at that text message, I was like, this is amazing! Um, because yes, it's called Hearth Helper, and it said Text me with a picture of a paper schedule or a calendar and I will add the, the events to your Hearth. No more forgetting dates or manually adding events. I, as an AI person, I was like, this is awesome. So it's going to leverage the art and beauty of computer vision, likely OCR, to extract from my paper calendar the events that I have scheduled for that week, that day, that month, and it will automatically create events on our family calendar, minimizing how much time I spend every single week synchronizing between my teenagers and my husband and myself. Many of you know I travel all the time, so I'm always in a new city. I— some— last week I was in 3 cities in 3 separate states. And not like close together like in the, you know, Northeast. Like it was— I was— I think I started in Vegas, then went to Kentucky, and then ended up in Philadelphia. So like my family's like, where are you? What's happening? And this tool gave us the ability to now synchronize those calendars together. And that synchronization is pretty awesome. But the AI for good comes when I'm like, oh my gosh, it also leverages AI so I could just take a picture of the calendar that I have And I will tell you, as a leader, one of the things that I do is I do plan out my day every day. I think I have my planner here so I can show it to you. But I plan my day every single day and I use a planner. Looks like this. I could show you some written ones, but I'll show you a blank page. But it allows me to like schedule out everything that's happening in my day. But then it also allows me to list my goals and then identify my targets. And the key to that, um, is that it allows me to say, okay, what am I going to do today to make sure that I'm go— that I'm doing things every single day in the direction of my goals? And by the end of this, I've now created like, okay, I'm gonna do this and this and this, and I've got like 12 or 16 things that I put on my calendar to do that day. Now what I would have done last week is I would have gone to my structured app and I would have typed it all in after I did my planning session with my husband. Today, I just take a picture of my planning session and the Hearth Display, the Hearth Helper, adds them to my calendar. Like, okay, awesome, let's go. Um, so that's my AI for good. To give you a tiny little bit more around, um, what this has done for our family, my son— as many of you know, I have a son who has Down syndrome His name is Max. So I showed him like, hey, look, Max, I've put some activities for you on here. And it like, one of them was like, um, wash your face. Another one was, uh, you know, very executive function things, things that often he struggles with. Brush your teeth, things he forgets to do on occasion. So I put all these things in, um, you know, throw your trash away, like all this stuff. And so I had him come over and I showed him what it looked like.. And he became very motivated to click them off because every time you click off a task, confetti shows up on the screen saying, "Hey!" And along with that, they've created this concept of streaks. Many of you Duolingo people out there understand this term. But it allows you to basically start keeping track of habits. And once you start doing more, a habit consistently, it rewards you with this streak. And I loved it because his face lit up. He was like, oh my gosh, I'm gonna, you know, I'm gonna go brush my teeth. And I was like, well, you don't have to brush your teeth now. And he's like, no, I'm gonna go right now so that he could check off the brush teeth capabilities. I thought it was cool. So we're using it to help kind of amplify the executive function of all the kids in our house, including Max, of course. And but also be able to manage our calendars, be able to use computer vision to easily add calendars to the list. Like, I'm just loving it. So it's really been a great experience for us. Hearth display. I'll put a link in the show notes so you can take a look at it, but really fun. All right, so we've talked about leadership, we've talked about some AI for good. Let's move on to some AI news. So in the news today, I saw that one of my old bosses, someone I used to work for, Eric Boyd, the CVP or Corporate Vice President over Azure AI at Microsoft, announced a really cool partnership. Snowflake has a model, I don't wanna say it wrong, it's like Arctic or something like that. Oh, it is Arctic. So they, Snowflake has a new LLM, Arctic, and it's now been added into the Azure model catalog. Now here I think this is really important 'cause most people, at least the people I talk to, don't know that this exists. But when I used to, last year I would do a lot of conversations and a lot of workshops on Azure OpenAI Service, the ability to leverage ChatGPT, but not using ChatGPT as the service, use the operationalized productized service inside of Microsoft called Azure OpenAI Service. It's like N minus 1. Awesome. However, what if you didn't want to use that service as your LLM? What if you wanted to use a different service? A Hugging Face model, for example. Well, that was entirely possible, and this is again a year ago. But most companies did not realize that there was an Azure Model Catalog. And one of the first relationships that Microsoft Azure built was with Hugging Face. So, thousands of Hugging Face models became available to you to use as an Azure machine learning engineer inside of your workflows. So, you could build with those models in a safe, responsible, secure Microsoft Azure tenant. Another thing that I was happy to see is obviously Mistral and then, Meta, like all these additional models started getting added to the catalog. Now we have Snowflake. So here, for those of you who are a little bit more technical or looking for an opportunity to kind of put your fingers on a keyboard and play around, I encourage you to, yes, of course, play around with Azure OpenAI. I use that service predominantly. However, I also like to go in and play with other models. And not every model, actually every model is good at something else. It has nuances to its performance metrics. I tend to look at Stanford's HELM project to identify— it's actually called HELM, H-E-L-M. I'll put those notes in the show notes as well. But it's the Holistic Evaluation of Language Models. And it allows you to see a baseline of performance across a bunch of models, including the ones that are in this model catalog. So you don't have to go blindly into the night looking for your model and hoping it'll be good or nontoxic. You actually do have the ability now to leverage Helm and look at models that are in the Azure Model Catalog. But what I love about that is that when you use, use a model that's in the Model Catalog, you are accessing a model inside of the four-walled garden of your space, right? Allowing you to build, um, LLMs or, or software, uh, machine learning systems. It gives you the ability to actually build that, um, inside of Azure. And that Again, most people don't know that. They think they have to go outside. They think they need to access a, like, a remote location. They need to go out to Anthropic directly or they need to go out to GitHub directly. That's not actually the case, right? A lot of these models are now infused into the Machine Learning Studio, something you can see quite easily. All right. So that is one thing that I found in the news. Another thing Ah, this is a really good one. I saw an article. It actually came out of like Psychology News Today or something like that. But it said AI can help people feel heard. And I actually 100% agree that as you build these systems, you can create the ability to listen to your customers in more impactful ways because you can actually hear them. You can transcribe what they say. You can turn it into actionable items. Like great news. However, what we are finding, and this ties into kind of AI trust, is that it can help— AI can help you as a company or as an individual build systems that make your customers feel heard. But as you ethically decide whether to disclose AI is involved, it is identified in this study that they were talking about that as soon as you disclose or label that the mechanisms you are using to capture that information and hear your users or your customers, that that trust, that impact you got by hearing them is actually diminished. And I think this is true across a bunch of different areas. But we're in a situation now where we have the ability to do more than we ever could before. But disclosing how we do it, which we've never had to really do before, right? If we build software to do this, we wouldn't have had to disclose to our user we're using this type of software. To solve this problem. But with AI, as the ethical guidelines encourage us to create transparency, we actually will reduce the impact that we have in using the technology in the first place. So just something to think about. Again, none of these are— I'm not definitively for or against, but I think it's really important to consider that when you decide to use AI and you decide to disclose it, that there are consequences to disclosure, consequences to not using it at all, and all things that you want to be thinking about as part of your AI strategy. So those are the two stories I wanted to talk to you about today. And again, every time I come on, I'll share with you some more updated information. One of the things I want to do now is talk about the world of AI gone wrong. So I have an opportunity to get on stages all over the world, and I always tell people stories about AI gone wrong. I have a Google alert that allows me every morning, 6:04 or so, it sends me a note with all of these AI gone wrong examples. And so every day I'm like, oh my gosh, I can't believe that happened. Oh no, not that. So I want to share some of these stories with you. And today I was given the alert around a use case where an individual working at a school— I think he was like a PE teacher— apparently was a bit disgruntled at work. He decided to create a synthesized audio of his boss, the principal of the school. And in that synthesized audio, he asked that AI to say in the voice of the principal antisemitic and racist comments, leading people to believe that he actually said these things even though he never did, right? Deepfakes, synthesized media, danger zone, right? Danger, Will Robinson. AI, the reason this is AI gone wrong is because like the world is kind of looking at this story going, Who's going to stop that from happening? Who is going to prevent bad actors, people who are going to use this to hurt instead of help? A disgruntled employee now gets to publish his boss's fake audio saying terrible things he never said. And now people, and I don't know if you know this, but some people, they don't even research the truth. They just see it and they're like, yeah, I always knew he was that kind of guy. Even though he, of course, is not that kind of guy. So interesting stuff to be said. Uh, there are criminal charges pending against the, uh, the teacher who is, who is the one who created this synthesized media. Yeah, it causes us all to kind of think twice about how this works. Now, it's not really AI gone wrong, but I also saw a video of a robot being interviewed. Ameca, I think it's called. But I was recently at the Las Vegas Sphere. And at the Las Vegas Sphere, I had an opportunity to talk to one of the robots there. And you could see it on Instagram if you want to see the video. I pinned it for easy access. So you could go and you could talk to this robot. So I of course was like, I want to talk to the robot. And in the video I saw today though, in the conversation with that— you either— it was that robot or a very similar robot from the same vendor. They asked the robot, you know, if AI would be able to eventually design itself. And I think it's a fascinating video because at the beginning it all sounds good. It actually sounds very human-generated. Like, of course AI won't want to do that. It requires all these things that are human only: creativity, innovation, patience, you know, the ability to build things out of nothing. Like, it reinforced a lot of the things I hear humans say about the world with AI, like that we have a uniqueness in this world. But at the end, it very quickly says, "But be careful what you wish for, because if you do create a robot that can design itself, it may realize that it doesn't need humanity at all." End of story. Like, scene, you know, cut. Like, I was like, "Wait, wait, wait, did everyone get that last part?" Like, it was just really funny because it sounded great. It sounded very well scripted. And And then at the end, it kind of threw in this zinger, and I was like, wow, that's interesting to think about, but also much more interesting that that model kind of came up with that on its own. Now, to close our session today, I hope you enjoyed it so far. To close our session today, I'm going to talk to you a little bit about where I'm going to be. So I have an incredibly amazing opportunity to go to New Zealand, Auckland. We're leaving tomorrow, and we will be headed to Auckland to be part of a conference called Future State, which of course, Noelle Russell, Future State sounds like a match made in heaven. And my job will be, of course, to go there and talk about, yes, of course, this incredible technology we have forming the future of our workforce, of the type of work we do, of companies that we'll build, but also how important it is to think not only responsibly, but also to think about how do we build these systems safely? How do we use the responsible AI principles that I've come to know and love of transparency and explainability and fairness and trust? How do we make sure that we build systems that embody these characteristics so that in addition to bringing in more value to a company, to amplifying the amount of revenue we can make, to increasing the amount of service we can provide, that we're also ensuring that we don't inadvertently, accidentally or intentionally hurt someone else. And so I'll be talking about the concept of red teaming and building AI safety systems. I recently got to sit with the Microsoft Responsible AI team in New York. They are an incredibly amazing group of people and they launched some new features to the Azure Content Safety product and platform that allow us to do a lot of the things that red teaming AI red teams would have to have done manually in the past. They're now building AI models that learn from a lot of the manual red teaming that was done at Microsoft and now can amplify any team, any company of any size to actually make sure they build not only resilient but responsibly safe AI systems that can help the world. I'm excited about it. I'm excited to, you know, of course, tell them about all the amazing things that we could do with AI, but then temper that enthusiasm just a little bit with the ability to build safe and responsible systems while, while building a world that serves more people. I will be there in New Zealand. While I'm in New Zealand, I have lots of fun activities planned. I will be on the local news. I will be interviewing with a local newspaper. I will also be on a podcast called Business is Boring. So stay tuned for all of that. I'm excited, uh, for the local media press cycle I will get to be a part of as being part of that. But I've also never been to Auckland. So if you know Auckland, uh, feel free to send me a note about what I should see while I'm there. But this has been an incredibly fun, exciting conversation with you. Thanks for joining me today on this podcast. It is my chance to share with you what I see every day, day in and day out in the world of AI. Thank you for being here, and I look forward to seeing you in the next episode.