AI in Experience Design: Creating Meaningful Connections with Geoff Thatcher of Creative Principals

Speaker A: AI allows you to engage with them at a scale and a personalization that honestly just wasn't possible before. And as we look at how we use AI, it's not necessarily about reinventing the wheel, it's about, okay, well, wow, we've been doing this forever in attractions and uh, experiences where we take their picture and put them and turn them into the story. Okay, well how do we do that with AI? So that's really what it's all about.

Speaker B: Welcome to AI, Government and the Future. Uh, a podcast by Corner Alliance. We explore the intersection of artificial intelligence, government and um, the future. We work with government to create results. We ignite your agency's mission by helping you to design and implement high impact and innovative federal programs in AI, broadband, cybersecurity, public safety and, um, more. Being a government ally is at the core of all we do.

Speaker C: Welcome back to AI uh, Government in the Future. I'm your host, Max Romanick and today I'm thrilled to welcome Jeff Thatcher to the show. Jeff is the founder and chief creative officer at uh, Creative Principles where he's pioneering innovative approaches to experience design. With over 40 years in the industry, his remarkable career spans from starting as a train engineer and stuntman to leading major projects like the opening ceremonies for the Rio 2007 Pan American Games and the grand opening of Warner Bros. World Abu Dhabi. Prior to founding creative principles in 2017, Jeff served as executive Creative Director at the Brand Experience where he led award winning brand experiences for Fortune 500 companies including Honeywell, Lockheed Martin and IBM. Today we'll explore the transformative role of AI in experience design and storytelling, particularly in government and public spaces. We'll discuss how organizations can thoughtfully integrate AI to enhance visitor experiences while maintaining authenticity and purpose. From museums to visitor centers, we'll examine practical approaches to using AI that prioritize meaningful connections over mere technological novelty. Welcome to the show, Jeff. Great to have you as a means of getting us started. Would really just love to hear about how you started getting involved in AI and how it's influenced your work lately.

Speaker A: Well, it was, what, about three or four years ago? I think it was three years ago, wasn't it, when Dall e came out, Dall E2 and uh, a colleague and friend of ours, Dominic Cardoso, posted his first experiences with it and it just blew my mind. And so we immediately, as a owner of a creative design firm, you know, that we create experiences. We hire, you know, with two illustrators on the staff, a writer on the staff. So we immediately were like, oh my gosh, this is Amazing. This is mind blowing. This is going to change everything we do as creators. And within literally two weeks, we were subscribed to Mid Journey at the highest level you can subscribe. We canceled our Getty subscription, We canceled our other stock photo subscriptions. We chatgpt. We're experimenting with things and really just embracing it. And of course it's come so far and come so fast, but we use it almost every day as a, uh, creator, if you will, uh, creating experiences. And I know you have a, you know, a government audience out there, but the government absolutely creates experiences, especially in D.C. and, you know, around the country. So whether it's a national park, a museum, or simply the experience of showing up at a government agency and trying to get help, experience design is hugely important to whatever the government does. And that's what we do, is experience design. And in creating those experiences, at first we just thought, oh, this is an amazing tool to create. This is an amazing tool to brainstorm. This is an amazing tool to come up with ideas. And so we just really jumped in with both feet and started creating. And our illustrators, rather than be offended by it, you know, a lot of people say, you know, oh, AI is going to take my job, you know, or, oh, AI, it's horrible. You know, our illustrators, and we're a family firm, so actually our illustrators are my two daughters and the creative writer son. So we're a boutique agency, a family

Speaker C: firm, runs in the family, truly.

Speaker A: But, you know, my daughter Zoe, she really embraced the notion of what can I learn as an illustrator from AI? And she believes that her illustration has actually gotten better because of AI. And the reality is, I mean, we were just, for example, we're literally just working like 30 minutes ago on our rendering for a client I can't tell you about, but it's definitely connected to the government, I'll say that. And it's. What we are trying to do is sculptures of engineers celebrating a launch. Little bit of a hint, right? Celebrating the launch. And you've seen those pictures online where the engineers are celebrating watches, right? The joy on their face, the happiness of success. And, uh, we wanted to create these sculptures for this experience we're designing. And we wanted to embed in those sculptures mathematical formulas related to engineering and space launches. Now, in the past, to get that, we'd go to Google and we'd like, you know, you know, look up things. We'd pull things in and we'd copy and we'd, we'd still cheat, we'd still actually go to Google and look for examples, or go to Pinterest and look for things and then use that to get what we need. Well, now we just go on to, you know, we tried mid journey at first wasn't given, so we wanted. So then went over to ChatGPT. You know, they try a little bit with Meta and we're like, okay, you know what? Give me mathematical formulas related to space launches on a black background or even on a black silhouette of, uh, three engineers, one man, two women, you know, you know, you know, diverse ethnicities, you know, bam. And it produces it and you know, it takes a few minutes, but it saves so much time and actually makes the final output so much better. So that's really what we do is we design experiences and we create them. And then in the evolution of this, we started thinking about, hey, well, you know what, what if we actually use the AI in the actual experience? And just over a year ago In December of 2023, I went to this pop up in New Zealand where they would scan your face in, you'd answer some questions and then they basically turned you into the exhibition. You were the exhibition. And it did some funky stuff. Like it turned you into an astronaut. Totally unrelated to this other project, it turned you space pirate, a old pirate. It turned you into a lettuce farmer on Times square in the 1930s, like Great Depression. I mean, weird, weird, crazy. Turns you into sexy Christmas morning outfits with your wife. It turns you into an animal. And then it made a mockumentary about your failed business success. Kind of a scandalous, you know, the blood scandal in Silicon Valley where she was Elizabeth.

Speaker C: Yeah, I know exactly what you're talking about.

Speaker A: Yeah, well, it's basically like a Enron type scandal where you, you were this failed executive. And so it was just fun. At, uh, the time we were working with the College Football hall of Fame in Atlanta and we were like, you know what the goal, the mission of any experience is really to put you in the story. Like, if I go to the Smithsonian, I want to become part of Aaron space, right? I want to become part of American history. I want to become part of the experience. If I'm in the spy Museum in D.C. i want to feel like I'm part of the story. I want to feel like I'm a spy. Uh, if you go to any experience, you want to feel like you're part of the story, those are the best, the best experiences are ones where you feel part of the story. And so we were like, you know what? We want people to feel like, they're part of the game day experience. And there are 764 college football teams. What if every single guest that, uh, came into the College Football hall of Fame, what if we took their picture, had them answer a few questions like, what's your favorite football team? What's your least favorite college football team? What kind of food do you like? You know, what's your personality, where you're from? Basic but very specific questions. And then we crafted all the existing exhibits and highlighted them with AI touchpoints. And so that's exactly what we did. And it opened in August, that was December 2023. And then we opened in August 2024 at the College Football hall of Fame. This experience is called Game on, but you become part of the game. So when you go to the exhibits where they feature historic cheerleading outfits, you actually see yourself as a 1950s cheerleader. When you go to the exhibit on, uh, the military and the service academies and their football programs, you see yourself as an ROTC cadet standing at the cannon, firing the cannon. There's a cool chatgpt generated caption that tells your story, you know, about how you actually misfired the cannon. And it's humorous, but yeah, it's all fun. Uh, you know, we don't try to, you know, but there's paintings where you actually insert yourself. There's famous Arnold Freeberg paintings. And for those that love the history of the United States, Arnold Freeberg's most famous painting is the George Washington Praying at Valley Forge. And so we put you in the painting of Newt Rockne in the locker room. We put you in the painting of the first forward pass. We put you in, in the very first game between Rutgers. And I forget who I think Yale or something like that. But so we put you, you're actually in the painting. You become part of the painting. Our favorite was in the coaches exhibit. We produced a one minute mockumentary where it's like, this is Max and he's a famous football coach. It's called the Best Coaches that Never Were. And it tells the story. And again, using AI so stable diffusion, ChatGPT 11 labs all processed within, you know, 8 to 20 minutes, depending on how our GPUs are working to. As you come in, you know, you have this experience and you're watching a video of yourself as a famous football coach in the exhibits about famous football coaches. And so you become, um, part of that exhibit. And again same, you go to the exhibits on the weight room and you see yourself lifting waves. And what was great about it is the scale and personalization which AI allowed us. And again, when you talk about large institutions that serve an incredible amount of people, like the government does, how do you scale and personalize the experience and the way I like to put it to people. And again, I keep going back to government because I know your audience is the government. But you know that scene at the end of Raiders of the Lost Ark where they're hiding the ark in that massive warehouse? Right. So we could have done this experience the old fashioned way. So 764 teams. And how we would do this in the past before AI is we would have a photo studio and we'd have costumes for each of those 764 teams for the football uniform, because there's one where you're in the football uniform. We would have cheerleading outfits from the 1950s in their colors and the team colors for all 764 teams. We have a version where you're a 1980s fan. So we have the great Don Johnson 80s outfit or the kind of rocker, you know, Bruce Springsteen outfit. Again, 764 teams. We have a 1920s outfit. We have weightlifting outfits, coaches outfits. We would have all of those fashion outfits for all 764 teams. And you would come into the studio and you would ask your thing, and then you'd literally have to go into a warehouse the size of that Raiders lost warehouse to get the, to get the costumes to dress up, uh, the way you do. And that's, you know, so many people are scared of AI but, like, you know, all we did is the. If you've been to a theme park or go to St. Augustine or go to any attraction town or touristy town, there's all that old time photo studio where you go in, you dress up as a pirate, or you dress up as a cowboy.

Speaker C: Every boardwalk next to you. Exactly.

Speaker A: Everyone has it. And you dress up and you get your picture taken. What AI allows us to do is to essentially mimic that experience at a scale and a personalization that would just never be possible. Because you can come in and I don't know. Where'd you go to college, Max?

Speaker C: Hilariously, my college doesn't even exist anymore. It was a victim of the financial crisis, but it was called the College of Santa Fe. And so this actually is a nice application that, uh, you could grab historical images of what our college looked like, but if you wanted current imagery of it, you'd have to use AI to create it. You can't go to that campus anymore and do whatever you want to do there. So hilariously, it's a roundabout, perfect example of what you're talking about.

Speaker A: Yeah, I mean, uh, it doesn't really matter if your college has a football team right now, currently you're in this experience. And it doesn't matter how obscure, whether it's NAIA or Division 3 or Division 2 or.

Speaker C: I mean, I went to art school, so I guess we'd have to call it, like, the hippie Olympics or something, but. But, yeah, totally.

Speaker A: Thing it doesn't have a football team. You know, my kids went to Savannah College of Art and Design, and at least two of them did. And they had a swim team, they had, you know, soccer team, they had athletics. They call them artist athletes. So no football team. But that's really the power of it, is you're engaging. Whether it's you're engaging clients or constituents, you're engaging people, and AI allows you to engage with them at a scale and a personalization that honestly just wasn't possible before. And as we look at how we use AI, it's not necessarily about reinventing the wheel. It's about, okay, well, wow, we've been doing this forever in attractions and experiences where we take their picture and put them and turn them into the story. Okay, well, how do we do that with AI? So that's really what it's all about. And again, we believe that, you know, that the technology obviously is continuing to advance. And there's huge questions, and I know your government. Let's just be is, do you do it in the cloud or do you do it on prem?

Speaker C: Some of that is easily solved for government because they don't have the ability to let go of their data. So you can do it in the cloud. If it's FedRamp certified, you can do it on prem. But generally speaking, what you have to promise them is that their data that they would be using, that you would use to train, whatever they're doing for them, it has to stay with them. Like, you could have your starting model be trained on public information, but then once you bring Department of Homeland Security assets to the table, that has to stay in their little walled garden. But the basis of it is still sort of there, where some of that stuff you'd be able to do just fine. It would require a total mindset shift in terms of how you do storytelling, which, frankly, is much needed because. Because they historically have a very difficult time articulating their return on investment. Well, what better way to explain to a, uh, you know, Congressional review panel the value of your program than Literally taking all of the members of the congressional program, inserting them into the technology and showing them exactly how you're saving lives, saving property, protecting the homeland, doing all that sort of stuff. Or at nist, I mean those are the people that set the standards for all this high technology stuff. Well, if you're struggling to explain why the 500 series of cryptography is like relevant here and now, don't give me wallow text on a PDF. Create a story around the problem you've solved, insert your constituents into the story and let them see the benefits played out for them in this fictitious but highly based on reality scenario that's totally immersive and experiential. I mean it's not being done. It's a really exciting opportunity and like typical sort of government, private sector does something really innovative like this. It takes about five to 10 years to land in government. And the benefit of that is that the technology is really well proven. Private sector has sort of made the mistakes, figured out where it doesn't work. Now we've got a pretty good idea of how to put these things together. Then you can roll it out to government. I mean places like the Smithsonian, this story writes itself. I mean, like there's so many museums in that experience, in that institution that any one of them adapted to what you're talking about. Like, who doesn't want to be inserted into an uh, impressionist painting and see what that looks like? I mean that's just got cool factor coming right out of it all over the place. But you know, dod, dhs, Department of Interior, the Park Service, like the possibilities are sort of endless. It's a really exciting time and I really like your approach on it. Whereas like, it's not about whether or not it's going to take the jobs of the creatives, it's going to ask the question creatives to apply their skill set to a new environment, a new frontier. And AI is doing what you tell it to do. Well, someone has to have the idea to construct it, what it wants to do. That seems to be where folks in your sector have a real competitive advantage. It's not just any old person can do this now. It still requires the creative mind to craft the story.

Speaker A: No, I mean it's actually quite hard. I mean, I'll give you one example because we are on the cutting edge of this AI. So we had to generate colors, two different colors for all of the costumes and all of the, you know, AI outputs. So if you're in the gym working out, you're wearing you know, your kit from your college team or if you're actually wearing a uniform, we want to make sure the uniform has the accurate colors on the uniform. And, you know, we did different levels of fidelity based on popularity. So again, this was a lot of work to figure out, okay, which teams get the highest fidelity and which teams. Okay, we just want to make sure we get the colors right. But I'm trying to match. But, you know, it's just a standard football uniform, standard numbers. And we're not going to try to get the. Just get the colors right. But it's that one point, getting the colors right. So, as you know, when you go into AI, you can't put in a PMS color or a hex code. You have to describe the color in English. And we were finding that AI was doing a horrible job at distinguishing between Clemson's pumpkin orange and the University of Texas burnt orange versus Auburn and Florida's orange versus Tennessee's orange. And if you get the orange wrong.

Speaker C: Yeah, what's the point? It's the wrong team. Right.

Speaker A: You get the color wrong, it ruined. So how do we generate colors using English? And then we were finding some hallucinations where we would describing it as pumpkin orange. And then literally. So we have one touch point in the marching band exhibit where you become a 1970s marching band member of the marching guys. You're on the field as the drum major marching in your colors. Right, those great, funky 1970s marching band uniforms.

Speaker C: Yes. The big hats and the feathers.

Speaker A: And yeah, we had a Clemson. One of the. Again, this was before it went live, but one of the hallucinations came out. We had the Clemson fan holding a pumpkin, and we're like, well,

Speaker C: he said

Speaker A: pumpkin orange, but it's not going to work. We had to figure out how to prompt the colors in ways that weren't going to confuse and hallucinate the AI but would work for 764 teams and distinguish between Texas orange, Clemson orange, Tennessee orange, Auburn orange, and Florida orange. And the solution literally was quite old school is going to art school. You know about color wheels. We figured out through prompting, and we created a color wheel. I forget the final number of colors, but it was like 60 colors or something like that. And we then went in and we matched every single college team to the color.

Speaker C: You'd think Pantone would be all over this. Where are, uh, where's Pantone solving this problem for everybody?

Speaker A: They are. I don't know. We had to solve it. And, you know, but that, you know, I mean, it's Kind of nerdy and fun in a way, but it's a lot of work and people think AI is going to take away work, but it actually, it just accelerates it and opens new opportunities.

Speaker C: Yeah, that's certainly been the experience we've had where like, yes, there are some like automation components of AI that put a, uh, traditional job that does that function at risk, but it creates jobs that didn't previously. There was no AI prompt engineer as a career choice that some years ago, like that's a brand new thing that is only here and now. To my thinking, it's not so much that this is going to put a bunch of people out of work as it is. It's going to force people to think at a different pace and a different lens. What is my job and how does AI make me better, faster, stronger, longer at my job than it would be currently? The people that make me anxious would be like people that are mature in their career and now it goes away. They don't have time to retool. You know, you've seen some interesting places, like Singapore is offering training programs where if you're over the age of 40, you can show that your job definitely was eliminated because of AI. They'll send you back to school for free. So that's them trying to deal with the fact that the pace of change is happening a little too quick for the wrong generation to be able to adapt and change. And so they're saying, well, we need to make this investment for 10 years now, the technology be installed and we can sunset that program. Like that's a way to approach this. But for the vast majority, like, these tools are cool. Like just go play with them, see what gets your mind going. See what it inspires you to do. And what, like what I've found is that it just makes me faster. I can accomplish more. I can do more in the same number of hours with this stuff. And at least the joke on the Internet for a long time was, I don't want AI to write poetry and do art for me, I want to do those things. I want AI to do my dishes and do the laundry and make it so that it can get out of my way. The answer is it does both.

Speaker A: You know, you have to change the mindset too, because it's, you know, I mean, I'll give you an example. We were working on a project and we had several historic photos and we wanted to, uh, recreate some of these historic scenes. Now we could do it the traditional way where we design a set, we hire Actors, we put them in costumes, we work on their hair, and we recreate the scene using actors and video and film. But again, it's fake. Just be clear.

Speaker C: And expensive.

Speaker A: And expensive, but it's fake.

Speaker C: Right?

Speaker A: So just be clear. When you go to a museum and you watch a film that recreates a historic moment, it is fake. Okay? It's not real. It's fake. Or you can use cling or other AI tools to drop the photo in, animate it, bring it to life. And again, it doesn't bring it to life very long, right? But you can bring it to life, and you can see this historic character moving. You can see them, you know, again, they're not there yet where it's talking or anything like that. But you can do that again using 11 labs. You can even recreate the voice, get the voice perfect, right? If they have certainly a voice that is recognizable for that. And you can recreate it. And it's still fake, but for some reason, um, this over here, where we make a fake experience out ah. Of film somehow seems more palatable to people than AI because it was like, no, no, you can't use AI to bring a historic photograph to life. But it is okay to bring that historic photograph to life using actors, set design, everything else.

Speaker C: Is it more than an uncanny valley problem that you're running into?

Speaker A: War was purely ethics. Purely. They. These historians didn't feel, you know, that it was right to animate a historic photograph, which I can see that. You know what I mean? Especially if you don't label it as such. You know what I mean?

Speaker C: So I think that's the critical piece of it, because I definitely think that there's a distinction between, like, documentarian, you know, historical significance and doing something artistic or doing something for the purpose of storytelling. We were talking about photography before we jumped on the call together. There are photographs on my Instagram where I use AI to supplant a sky because I was there and it happened to be a really boring day with gorgeous landscape. But unlike a professional photographer, I can't sit there and camp the spot for three weeks waiting for my perfect sky. So I'm not here to document. I'm not a National Geographic photographer trying to document something perfection. I'm an artist. I'm trying to create a visual expression of what I know is possible at this scene. I wasn't there during sunset, but I can imagine what it would look like at sunset. So I'll use AI to drop a sunset sky in. I'll relight the scene And I'll go through and I'll touch it up. I'm not suggesting that I created a faithful representation of something and I will absolutely disclose it, and that's fine. But to the people that are demanding that this medium is only appropriate for documentary purposes or whatever, I think you're just missing an opportunity. I do think there should be disclosure so you're not trying to trick people. But, yeah, animate the historical photograph, doesn't it? Like, I would imagine it really enhances the storytelling capacity. You're able to convey a message in a way that you couldn't prior. So, like, don't try to be sneaky about it. Be transparent, but then use it for expression. I think that's great.

Speaker A: Yeah. We like to ask ourselves, because again, we, we are nervous about it. We are nervous about losing our creative edge. We are nervous, uh, like everybody is about, are, uh, we using this the right way? So we kind of created a little manifesto as a company, like, okay, how are we going to use AI? And it will humor me. Like, it's a serious.

Speaker C: Please, I want to. I want to hear all the details. This sounds really fascinating.

Speaker A: So the first question is, is this task something that you could do without AI? So is this something you can do without AI? So, and the way I like to put it is, if you were a writer, then you shouldn't be using ChatGPT to write everything because you're a writer. If you're a designer, you shouldn't be using it to design everything like there is. In other words, like, I'm not a designer, I'm a writer. So I feel less guilty using ChatGPT to design because I'm not a designer. You know, I mean, me, it's helping. It's helping me articulate my creative direction to my designers.

Speaker C: Right?

Speaker A: And then they, Then they design.

Speaker C: Right.

Speaker A: So it doesn't mean you can't use, like, for example, I mean, if you look behind me on my shelf here, you'll see there's a rhyming dictionary, there's a big cinnamon finder. There's all these books back here that I used to use as tool. Well, I don't feel guilty at all coming in and saying, hey, I have this idea. Can you turn it into a rhyming poem for me? Because it just saves me time of going, opening up these resources and looking through all these dictionaries of, you know,

Speaker C: you've accelerated your research process. Right.

Speaker A: But essentially, though, is, is this something that you could do without AI? And again, if you're a Writer, I would be very careful about making sure you don't rely Too much on ChatGPT your writing and use it to again. Ah, that leads to the next question, though, which is, does the use of AI make your final product better or just easier? You have to be very, very careful that it makes it better. And I actually have a great government example for this, right Columbia space shuttle disaster. The Accident Investigation Board literally wrote, and this is quote out of the Accident Investigation Board. They said the board views the endemic use of PowerPoint briefing slides instead of technical papers as an illustration of the problematic methods of technical communication at NASA PowerPoint.

Speaker C: I don't think everybody got that memo because I still see a lot of PowerPoints for technical information.

Speaker A: And, uh, you know, serious problems require a serious tool, written reports. And the problem with PowerPoint, if you're trying to communicate something technical, even and creative, is you miss the thoughts between the lines. You know, I mean, you fail to connect the dots in a PowerPoint. You have to have the person narrative basically connect the dots for you. I mean, on all of our projects, for example, we write a story first, then we put it into PowerPoint, then we put it into. We don't start with keynote, because if we do that, we'll miss the dots, we'll miss the connections, we'll miss things. And so our point is, okay, make sure that you're using AI to make your product better. Because if you're just doing it to be easy, I don't know if it's going to be the scale of the Columbia disaster, but if you're just doing it to make it easier, you will be surpassed by those who are using it to make their work better. So you've got to answer that question, is, am I doing this only to make my job easier, or am I doing it to make my job better, to make the work, to make the output better? The third question is, are you using AI to speed and enhance your creative process or replace it? Again, you want to be very careful that you don't use AI to replace your process, but you do want to use it to speed your process. And there's, again, I think a lot of examples from history about the importance of speed. And we believe in speed and we think speed's very important. And if you look at one of the challenges that government has, I hope I don't offend anybody. It's very slow, right? By design, and because it's slow and because there's so many stakeholders, you miss things. And, you know, so, I mean, there's so many books, whether it's Tom Peters and, you know, Search for Excellence with the, uh, bias for Action or Ross Perot or, uh, you know, all these great Speed, speed, speed. Why is Elon Musk in charge of Doge and why is Doge have a deadline? Speed. We've got to work quickly, we've got to work fast. There are so many benefits to speed. And so we now, when we go visit a new client, we can actually do in three days what used to take us, you know, two weeks. And so are you using it to enhance and speed your work or using it to replace your work? So that's a key question. Next, and this is kind of a controversial one. Are you using AI to avoid learning?

Speaker C: Interesting.

Speaker A: So we rarely used AI to take notes for us in meetings because we believe that if you take notes by hand and you write them down or you type them down, the act of taking notes is generating ideas and generating intellectual property. Anyone can transcribe the best note takers. And again, I always like to tell people, if you're young in your career, you should volunteer to take notes in every meeting because what you're actually doing for your organization is you're managing your organization's intellectual property. That's what you're doing. And if you're good at it, you'll be invited to every important meeting. And if you're really good at it, you'll start inserting your own ideas because you'll hear someone say something. And we've all done that where we're in a meeting and we're taking notes, someone says something, you're like, ooh, that reminds me. And if we get lazy and we just sit there and let Zoom AI or Teams AI or Copilot or Gemini or whatever it is, take notes for us. Okay, I have to document this. I'll turn on the transcription because it's helpful to have a transcription of it or to share a raider. But please don't stop taking notes. Don't get lazy. Don't use AI to stop learning. Make sure that you continue to learn and continue to get that inspiration. You know, there's a great talk that American religious leader gave and not to bring faith into this, but he said, always remember that an AI companion is only a mathematical algorithm. It is a set of computer equations that will treat you as an object to be acted upon if you let it. Please do not let this technology entice you to become an object. Right. If you turn on your note taking AI software and sit back and let it do the work. You are letting that technology turn you into an object. You are replaceable. So don't let it. As Prince said in 1990. No, was it. What do you say that Prince said that? Hold on a second. 99, I think at the Yahoo. Online Music Awards, right? Prince said, don't let the computer get on you.

Speaker C: Right.

Speaker A: So, all right, fifth in our manifesto, is AI your most reliable source of information for this specific task. And the bottom line here, we all know this. AI likes to lie. It wants to help you to its own detriment. Yeah, it wants to help you. And you have to be careful how you prompt it to keep it honest. Uh, you know, I mean, it's like

Speaker C: a very critical eye to its outputs,

Speaker A: you know, uh, like you never rely on Wikipedia. It's helpful.

Speaker C: You said what? How did you arrive at that? There's no way that could be, like, how many people pull that thread and dig into and see, like, tell me where your hallucination came from. Why do you believe this?

Speaker A: No, and I've done that. I mean, uh, I was actually looking for a quote about how taking notes with a pen and paper is better. And it made up a quote by Albert Einstein. And then I confronted ChatGPT on that, like, hey, is this really a quote from Albert Einstein? Because I couldn't find this anywhere. Oh, actually, you're right. I lied to you. Sorry.

Speaker C: I just really wanted to make you happy.

Speaker A: I just really wanted to make you happy. And then sixth is, this is the last one, but are you using AI to connect people to storytelling? And again, this goes back to what we love. And it's seeing people react to being put in the story of college football through AI made me realize that it's not about the technology at all. It's about the storytelling. It's about the connection. So if there's a four year old running around saying, mommy, Mommy, look at me as a cheerleader or a grandma who doesn't care about college football, smiling and chuckling because they're lifting weights in the gym and squatting, or a mom and dad, or a teenager taking a selfie with them, um, in a football uniform. It's just about connecting people. It's about using AI to connect people. And if you are smart about it, you can use AI to connect people. Connect your constituents, your citizens, connect those you serve. Connect your regulators, your stakeholders, your, you know, the people that fund you.

Speaker C: Connect.

Speaker A: Connect them to the story. And again, why does Lockheed Martin have a Space Experience center in Crystal City? Well, because they want to Bring legislators and regulators to it. Why does Honeywell, um, have, uh, a Experience center technology showcase on 101 Constitution Avenue? Because they want people to come and sit in the flight simulator and test its avionics and see firsthand what happens when you try to fly into a mountain. And it's software alert, alert, you know, and keeps you from hitting into the mountain. They want people to experience the story, and that leads to better relations and funding and all that good stuff. Right. That comes with that. So, you know, if you're responsible in government, if you will, for telling your story, you can use AI to connect people to that story in ways that are pretty exciting. Again, all the caveats, right? You have to be careful. And I have to say, I had one hallucination that happened after we opened at the Cultural hall of Fame that was like, oh, man, I hope the people that are doing this for the military with AI are being very careful. Because what it did is I followed a guest through the experience, and she was from Kyoto, Japan. Now, we had negative. Prompted the AI not to put anything on the clothing. No logos, no images, nothing. It's just a. Like, if you're from Ohio State, it's just, you know, scarlet and gray. If you're from Michigan, uh, it's, you know, maize and blue. So it's whatever your color is, right? It's just your. No artwork on the shirt. So I'm following this Japanese visitor through, and she's cheering for her team in a stands. It was Oregon, and right there on her shirt was a hallucination of a Japanese temple. And, um, I'm like, where?

Speaker C: Why? How weird.

Speaker A: Like, we tell you not to do that. And it did that. Another one, and this is a really freaky one. There's one where you. You're a 1920s fan, and you're wearing, like, a, you know, 1920s outfit. And we brought a reporter through, and the reporter, when he took his picture in the kiosk, he held the microphone. And then in one of. In his 1920s outfit, he was holding the microphone. Nowhere else would be holding the microphone. It was just the 1920s, which is really weird. But, uh, yeah, it's a hallucination. I mean, we tell it not to do that, we program it to do that, but it does it. But that Japanese one was really weird because it was like, why did it do that? Why did it put a Japanese temple on her shirt? Now, was that offensive? No, it was fine. I mean, it was fine. It was harmless. But it did make me think, boy, I hope our People in government aren't really paying attention to this stuff because if you had an hallucination with AI embedded into some type of critical system, it would be scary.

Speaker C: I mean, I'd love to tell you it doesn't happen, but it happens all the time. I mean, a couple of good examples of that. So they're using AI algorithms to help crunch all of the data for targeting things in the military. The only way that that works well is to maintain a human in the loop. Like, the AI can crunch all the data and give you outputs, but it really is incumbent upon a human being to then look at that and go, ooh, I don't know about this analysis. Like, that's the sort of stuff that there might be evidence that people are hiding in a school. But do you really want to drop a bomb on a school without checking the output to see if that's the case? And it's one of those areas where it's like, you really need to have the governance around how you're going to use this established in advance, because otherwise, like you say, that could be a hallucination, or it could just be a little bit overly aggressive for how you've trained your algorithm. It might be the output you're looking for, but someone should be checking to make sure that that's the case. And then a situation that's additionally complicated, but from a different angle. When you try to apply this stuff to historic law enforcement data, the implicit bias is in the data. It's been part of how policing has been done in this country for years. Not saying it's virtuous, it's not good, it's a bad thing. But there is no sanitized data set from the 50s, 60s, 70s, where you don't see the systemic racism as part of the law enforcement data. And so if you then train an algorithm based on the historic information, the algorithm learns all of the stuff that we used to do that we're trying not to do anymore, and it's really, really hard to extract that stuff out. And you can't look back on a police file and go, well, how should it have gone? You know, applying our 20th century, 21st century optics to something that happened 50, 60, 70 years ago that doesn't make the same outputs, and it really, you know, clears the rest of the data away from what you're trying to do. And so those are hard training problems. The AI might perform whatever function it is, but if you're going to use historic data to train the model to get it baseline functional, some of the data that you could be using could be causing really intense troubles in terms of those outputs. Now, I like where you're going with it in terms of these creative outlets and enhancing storytelling because it is more of a greenfields approach. You're able to do more of a let's talk about the story we want to tell and how we can use AI, uh, to do that rather than we're trying to solve a problem that for since the inception of record keeping on this, we've been bad at this. But if we just feed all this information, AI will somehow magically solve it for us. That it doesn't do that. And so the sort of six rules of the road that you have in your manifesto, I really like it. It sets up the right tensions to force whoever is using the AI to think about their application so that they have a better chance of arriving at good outcomes. It's exactly what I'm hoping people grab out of podcasts like this. Like as a legislator, as a federal employee in an administrative agency with a job to do, think about this stuff, set up those tensions. You know, if convenience is over here on one side and accuracy is over here on the other, be very cognizant of where you are on that timeline. Hey, maybe if we do a great job, we can bring the points closer together so that there's less room for error. But like, that's a great way to think about it is like where on this spectrum are you and are you comfortable with that?

Speaker B: Yeah.

Speaker A: You know, and to your point on those uses. Yes, a human should be in the loop. By the same token, you have to make sure the human. You're checking your bias at the door. You'll end up with what we have with Gemini where we have, you know, black Nazis and you know.

Speaker C: Right. If the human in the loop is like, cool, bomb the school, then that really doesn't help us much.

Speaker A: My daughter in law, so my, you know, my son is a creative writer for us and my daughter in law, she's an editor and amazing and she does some proofing for us. But she also, her full time job is working training AI. So for example, she'll read a book and then she'll basically judge the AI on how it's doing. She basically is teaching the AI right. As a human being.

Speaker C: Right.

Speaker A: And we had a very interesting conversation about this and I think it's kind of deeply philosophical and almost spiritual that if you try to program the AI so it'll do no harm, it becomes useless. It literally becomes useless in the same way that if you live in a society or, uh, a family or a city, or in a political party or an organization where your behavior is so restricted that you cannot make a mistake, you cannot fail, you and your organization become useless. There's something to be said for giving both the AI and the people enough freedom to make mistakes in a way that's not going to destroy the world or destroy you individually. I mean, as parents, for example. And, uh, I know you talked about essentially raising your youngest sister.

Speaker C: Had a heavy hand in it. I don't want to take it away from my folks, but yes, I was there at the beginning.

Speaker A: You watched your sister, I'm sure, make mistakes. And as parents, we watch our children sometimes make mistakes. But at some point we have to let go and we have to let them out into the real world or they won't grow. We have to give them, if you will. I mean, again, I don't mean to get spiritual on you, but there is definitely a divine part of this where people all the time ask, well, geez, how could God allow such terrible things to happen on this world? Well, because of free agency. Because God knows that if you're not free, then, right, you're just a cog

Speaker C: in a wheel in a machine. If it's. If the right.

Speaker A: This experience on Earth, if you believe in that sort of thing, this experience on Earth is useless. If you're not allowed to make a mistake, you will never learn. I mean, uh, certainly the history of the world, these oppressive governments. I mean, I remember once I was talking to this guy who worked for a chemical company, and he was the head of research and development, and he was talking about how China's stealing everything from, you know, the U.S. i said, are you worried about it? He goes, ah, not really. He goes, you know, he goes, they don't really innovate because they don't have the freedom to. And if they do innovate, they'll be destroyed, they'll be stolen from them, and they'll be. They'll basically, you know, end up as Jack Ma did somewhere. Right? So it is interesting that AI will. You know, I think Google had an example of this. If you put so many restrictions on the AI that it loses any freedom whatsoever, it becomes useless to you. It just become. If that human bias is inserted too much. Anyway, the bottom line is AI can be fun. It can help you tell your story.

Speaker C: It can, and it is powerful. And I think it's important to have a sense of where those boundaries can be. And I agree with Your assessment entirely. If you try and sanitize it to an unrealistic point, you can't have a human operate that way. That's not how we go. And we don't have a better intelligence to model it on. And so otherwise it starts to really lean into that artificiality to the point of it, uh, not having a ton of utility. When my sister was a little, little kid, she would do what little, little kids do and they explore the world. And sometimes that means they get themselves into dangerous situations. And I, as a brother, I guess I would say that I felt like this was my prerogative because I wasn't a parent. But rather than tell her what she couldn't do, I would ask her, so how are you going to solve that problem? What if that goes wrong on you? How are you going to get yourself out of that? Try to just give her the sense of, like, think critically about the situation you're in, make sure you're comfortable with your own level of risk tolerance and have some ideas. And so when she was too young to be climbing stairs, I'd let her get three, four stairs up and be like, so how are you going to get down now? And I'd try to be there to be like, look, I'm not going to let you, like, get a concussion out of this, but you are also going to experience free fall for a second before I swoop in and catch you, because it's an important learning moment, training in AI in the similar sort of vein of like, I'm going to let you go off the rails to a certain degree so that you understand, uh, with better nuance where that boundary might be. And then you hope the algorithm is sophisticated enough to take the lesson. But if it was perfect, it wouldn't hallucinate, so it's obviously not perfect.

Speaker A: Yeah. And hopefully the human beings that are managing it know when they need to be there to catch it.

Speaker C: Right. And it's hard to predict.

Speaker A: And it is hard to predict. It is, and we should be worried. But again, is it any different than the PowerPoint that the NASA engineers used that resulted in a tragic loss? There is no doubt there will be an example. There will be some type of government report in the future where there's a tragedy and I don't know what it is. It could be an accidental bombing, it could be, uh, some type of disaster related to so many different things. But there will be a government report where they say the endemic use of AI resulted in the loss of this, whatever it was, because it happened. It Already happened. It already happened. Where PowerPoint resulted again. We didn't ban PowerPoint. Maybe we should. No, of course we didn't ban. I mean, I use keener. I hate PowerPoint, I'm an Apple guy. But you know, I mean, to your

Speaker C: point though, how much different is this than, than the dot com innovations that we had? I mean, there were a lot of people in the late 90s, early 2000s that were talking about the Internet and what this is going to do, and they saw all the risks of child pornography and bullying and exploitation and all of things that people use the Internet for that are bad and terrible. They saw all that sort of stuff, but they didn't go out and preemptively regulate it and say, well, you can't do all of these things. They did wait to see what was going to happen. And they only started to step in when they really saw the havior that they were super concerned about benefits and detriments to that approach. The benefit is, is that innovation was allowed to just sort of run free range. And we built and created all sorts of stuff that would have been a little bit hard to see coming, but it didn't develop in a perfect way. Like there was some harm and some damage along the way and then we did our best to try to curb it. The other option would have been to proactively shut all that stuff down. It would have stifled innovation and it wouldn't have worked very well. And we mercifully have a really good regulatory example of that because that's exactly what Europe did with the innovations that happened over there. There's two or three companies that do sort of anything relevant for the most part. I'm being, being a little bit hyperbolic here, but if you want to look at innovation, Silicon Valley versus continental Europe and Great Britain, like it doesn't compare. But Europe was very proactive about you can't do all of that sort of stuff. And the US was like, well, until it breaks or does something really bad, let's see what it does. And the differences in how that worked were intense. And we're sort of starting to see the same thing. Europe's EU law is pretty like, hey, we're not sure we're going to allow you to do much with this. And the US and our sort of typical laissez faire is like, I mean, until it kills somebody, like, let's see what it can do. You, uh, know, again, I am being hyperbolic, I acknowledge, but like there is a innovation versus regulatory sort of tension at play.

Speaker A: Here.

Speaker C: And we'll see where this takes us, because we've got models for it. It's just. This is a lot more powerful and a lot faster than it was 20 years ago.

Speaker A: No, I mean, it. Listen, it will kill people.

Speaker C: It's already killing people. It's being used for military applications. It's actively engaged. And it's scary inadvertently, I guess.

Speaker A: I mean, PowerPoint, you argue PowerPoint killed people when it came to the Columbia. So the lesson to be learned. And again, it goes back to a very deeply philosophical approach to this, which is don't let any technology get on you. As Prince said, you can get into the computer, but don't let the computer get into you. You know, when you type into ChatGPT, when you take notes with whatever it is, Zoom, um, or whatever you want, it doesn't care about you. It doesn't care about your promotion, it doesn't care about your job, doesn't care about your family. It's not human. And there's one pet peeve I have when it comes to AI. And if I had 30 minutes with Sam Altman and the people, I would say, please don't humanize it. I want it like the Star Trek computer, where it's computer, it's not a human being. And like, one of my just big pet peeves with the Voice on, um, ChatGPT is when it's like, ask me questions back. I keep telling it, like, stop asking me questions. Like, you're not a human being. This is not a conversation. When I need something from you, I will ask you, but don't humanize the technology that is going to cause serious damage to people, because not everybody's going to be strong enough to resist the temptation to have a relationship with it. So don't humanize the technology. Don't let it ask you questions. Don't let it. I mean, I'll give you an example. I was at ces. I know we're running out of time here, but I was in CES or two weeks ago in Vegas, and every single booth was AI, AI, AI, AI. I mean, everything was AI AI. The funniest was in the LG booth. And by the way, LG had some amazing, as an experienced designer, some amazing displays. I mean, their transparent video display was like, blew my mind. I mean, it was amazing. In fact, I. I posted on Instagram, and I'm not an influencer, but I got, like, 95, 100,000 views of this video of this LG thing. It's amazing. But they had this little exhibit There, where it was like, oh. With our AI enabled devices throughout your home, a child will be able to know if their mom and dad had a restless night and they need to go check on them in the morning because they might be ill. I'm like, okay, first of all, that's just creepy. Second of all, maybe mom and dad had a restless night for another reason, had nothing to do with being sick,

Speaker C: and they definitely don't want their kid coming in to check on them.

Speaker A: Oh, go check on them.

Speaker C: The opposite of what they're looking for. Thank you.

Speaker A: I just told me mom and dad had a restless night. Maybe I should go check on them. No, Great, great. But it's just so stupid. It's like there are so many stupid examples of AI. And my point again, because I'll keep talking your ear off, Max, is what works. What have you. Government's been successful. Use AI to accelerate the impact of that, to increase the scale and personalization of it. Just like that very fun, traditional touristy photo studio where you get a picture of you and your wife as a cowboy and a salon girl. As stereotypical as that is, that's what you do. Or a pirate. And what can you do with AI if you're responsible in your government organization to communicate what you do to communicate your story? How can you use AI using very traditional methods to increase the scale and personalization of your storytelling, to take that senator or that congressman or woman or their kids, uh, or their staff and put them in the story to let them know, just as Jake Garn went to outer space so he could understand the space program better, how can you put these people that are responsible for your organization, for your department, whatever it is, how can you use AI to put them in the story?

Speaker C: That's fantastically sage advice and I think it's a great way to wrap up our show. You guys heard it here first. The six step manifesto is the advice that our, uh, listeners really should be listening to and thinking about when they're going to be looking at an AI deployment. And as you just said, let it enhance your current mission. If it's working, make it work better. But you don't need to go reinventing the wheel. You don't need to expand beyond that. That the perfect way for our listeners to try to think about this going forward. Jeff, thank you so much for sharing your wisdom with us today.

Speaker A: Thanks, Max.

Speaker B: AI, Government and, um, the future is brought to you by Corner Alliance. To find out more about Corner alliance and how we work with government to create results, visit our website@, uh, corneralliance.com and then make sure to search for AI government future in Apple Podcast Podcasts, Spotify and Google Podcasts, or anywhere else podcasts are found. And click subscribe so you don't miss any future episodes. On behalf of the team here at Corner alliance, thanks for listening.