---
id: "workshop:starting-with-curiosity-a-beginner-s-guide-for-public-servants"
type: "InnovateUS workshop"
title: "Starting with Curiosity: A Beginner’s AI Guide for Public Servants"
url: "https://innovate-us.org/starting-with-curiosity-a-beginner-s-guide-for-public-servants"
date: "2025-04-10"
duration: "60 minutes"
language: "English"
instructors: ["Jamie Kimes","Caleb Williams"]
recording: "https://www.youtube.com/watch?v=1EeiaSsCaos"
transcript: "automatic, from the recording"
source: "InnovateUS (innovate-us.org)"
licence: "CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org"
---

# Starting with Curiosity: A Beginner’s AI Guide for Public Servants

InnovateUS workshop · 10 April 2025 · 60 minutes

Artificial intelligence (AI) is everywhere—but what does it actually mean for your work in government? This beginner-friendly workshop is designed for public servants who are curious about AI but may also have questions, concerns, or hesitations. No technical background is required.

Together, we’ll explore what AI *is* (and isn’t), how it works behind the scenes, and how public servants can begin to navigate this rapidly evolving space with confidence, ethics, and purpose.

## Learning goals

- Learn the different terms and acronyms related to AI—ML, DL, GenAI, NLP, LLM—and what they actually mean
- Discover how AI models “think,” make predictions, and where they can go wrong
- Explore the difference between common fears about AI and the real risks public servants should watch for
- Learn how to find your starting point in the AI landscape, based on your work and values

## Speakers

- **Jamie Kimes**, Founder, The Idea Garden
- **Caleb Williams**, Founder & Principal, dataPIG

Recording: https://www.youtube.com/watch?v=1EeiaSsCaos

## Transcript

Automatic transcript of the recording: names and terms can be misheard, and the times can be a few seconds off. Each time links to that moment of the video. The transcript alone: https://innovate-us.org/starting-with-curiosity-a-beginner-s-guide-for-public-servants.txt (plain text) and https://innovate-us.org/starting-with-curiosity-a-beginner-s-guide-for-public-servants.json (every segment with its start and end second).

**Jamie Kimes**

[0:00](https://youtu.be/1EeiaSsCaos?t=0) Jess, I love to be called amazing. We love that. I want to just put out there right away, this is going to be very interactive. So I'm probably just going to be throwing stuff out there. And just if there's a question, just respond in the chat. Right, Caleb, we just want it to be super interactive. I am Jamie, founder of the Idea Garden. I love to make complicated things simple and boring things fun. That's the goal with this very complicated

**Caleb Williams**

[0:25](https://youtu.be/1EeiaSsCaos?t=25) topic of AI. Very good. Hi, everyone. I'm Caleb, and I make simple things complex. So in all seriousness, where Jamie brings just a wealth of expertise kind of on the literacy side, and really breaking down these topics to make them super approachable. I kind of come from a very hands-on technical perspective and very excited to be teaming up today and bring you guys this workshop. That's right. Okay, so we're going to

**Jamie Kimes**

[0:49](https://youtu.be/1EeiaSsCaos?t=49) get started right away with another poll. And actually, this is going to be a little bit duplicative of that poll you just took, but who cares. So Jess, if you just want to launch that poll, rate yourself on a scale of zero to 10. And Caleb, I'll let you take this part.

**Caleb Williams**

[1:04](https://youtu.be/1EeiaSsCaos?t=64) Yep. And so if you're just at the beginning, if you're on the left side here, that is a-okay. This workshop was built for you. And if you're kind of already feeling like you're on the right side of this poll, you have a solid understanding of what's going on. That's okay, too, because really what we want you guys to leave with today is feeling like not only do you have a solid understanding, but you can also explain this topic to people who are on that left side of the poll. So really starting with all ends of the spectrum here and just trying to get everybody comfortable with this topic. And we're seeing kind of results come in. The data side of me loves this. We have a very normal distribution centered around five here. So it sounds like lots of folks have at least some familiarity with the topic. It's a lot of folks that it is their first time, so really excited to see kind of a widespread of familiarity here as we dive into things.

**Jamie Kimes**

[1:53](https://youtu.be/1EeiaSsCaos?t=113) So our goal is to get you from that left side over to the right side. And we want to do that by answering five questions. But before we get into those five questions, I really want to address the elephant in the room. And that is that artificial intelligence can be a really uncomfortable topic for a lot of people. And it can make us nervous. It can even make us like downright scared. And I'd really be interested if you just want to put into the chat anything that you've heard about that's made you nervous or something that gives you hesitation when it comes to the topic of AI. You can just drop it in the chat. And I know, Caleb, that we're familiar with a lot of, yeah, deep fakes definitely is a big one. How do I know what's real? Data privacy, bias, lack of accuracy. There are a lot of fears. We hear this a lot, right, Caleb?

**Caleb Williams**

[2:43](https://youtu.be/1EeiaSsCaos?t=163) Yep. All of these ones are definitely ones. I see climate impact. I see trust taking over jobs. Gosh, they're coming in too fast for me to even keep up. Definitely one of the big ones, the accuracy with these models. And we're going to touch on some of that as well later

**Jamie Kimes**

[2:59](https://youtu.be/1EeiaSsCaos?t=179) today during this session. Yeah, exactly. And one thing, I like, Rachel, what you said. You said you're excited. I have a great mentor, David Smith. He's actually one of the AI champions in the state of Colorado. Fantastic. And he told me, he said, Jamie, if I could just say one thing to anyone who is starting out with AI, I would say, don't start with fear. Start with curiosity. And so, even though all of these things that just came into the chat are really important, and we will talk about them today, it's not where we're going to start. We're going to start with curiosity, learning about this really cool thing called AI. And we'll start with what it is. I have a rule that I live by in life. And it's any time I want to learn about a new technology, I look for what it's imitating. So, what do you think that AI is attempting to imitate? If we're honest with ourselves, like there's really no kind of technology that exists that really isn't an imitation of something that already exists in nature. So, what do you think it's imitating?

[4:07](https://youtu.be/1EeiaSsCaos?t=247) Human knowledge to learn, basic reasoning and summary information, critical thinking, language patterns, et cetera. Yeah, you guys are really, really onto it. Artificial intelligence is technology that specifically attempts to imitate how humans learn. Would you agree, Caleb?

**Caleb Williams**

[4:28](https://youtu.be/1EeiaSsCaos?t=268) I think that's exactly right. And we'll talk through some of, I see data is another one of the things down here. We're going to talk through some of how it's actually doing that as well a little bit later on and get a better sense of how exactly does it imitate how humans are learning.

**Jamie Kimes**

[4:43](https://youtu.be/1EeiaSsCaos?t=283) Imitation isn't new. Caleb and I were talking about this yesterday. It's not new. As humans, we are creatures of imitation. We looked at the birds and we wanted to be able to do some of what the birds do. Did we replace the birds? No. We just figured out a way to do some of the cool stuff they do. So, imitation isn't new, but the question really is, how do humans learn? And more specifically, how do you learn? So, put in the chat, what do you think is the most effective way that you've learned in your life? Is that data analysis?

**Caleb Williams**

[5:18](https://youtu.be/1EeiaSsCaos?t=318) Data analysis. And I think, Jamie, some of the ways that I have learned, definitely repetition, just actually grinding through, especially if you're learning languages and things like that. But I think the biggest one that I learned from, I see Jennifer here, by doing. It's exploring. That's really what hits topics home for me and how I really start to understand them. I'm also seeing watching and listening, visuals. I'm a huge visual learner. I know you are too, Jamie.

**Jamie Kimes**

[5:43](https://youtu.be/1EeiaSsCaos?t=343) Yes. Very big visual learning. And really, if you take all these things from the chat, you'd be able to bucket them all into this category that we just call experience. We learn from experience as humans. But early on in our lives, before we have a lot of experience, we have to be trained a little bit, right? And we have caretakers or teachers that literally say, this is an apple, apple, apple. And then they hold up an apple and a banana, and they're like, fruit, where it's categories, right? And then there's rules, like, don't cross the street without looking both ways. We learn from routines, things that happen again and again, like, I'm going to put on my pajamas, brush my teeth, and go to bed. And the theme of all of this is that it's all repeating over and over and over. Yeah, trial and error, too, Frank. Thank you. But all of this doesn't really matter unless we have the ability to learn. And we have to be able to learn in two different ways. We have to be able to learn directly from what we're taught, right? And we have to be able to learn directly from experience. If we touch the hot stove, we have to be able to be like, never going to do that again, even if nobody says don't touch the hot stove again, right? And all of this, our experience and our training is really giving us a perspective, a way to look at the world and to know what to look for. So pop question, what do you think we're looking for the most when we're learning? Like, the theme that goes through all types of human learning. We're looking for two things, basically. Answers, growth, understanding and meaning. Yeah, these are good. Context, explanation, recognition, not being afraid. And Edwin, if you're thinking, okay, I'm looking for not being afraid, what you're looking for is things that happen over and over that results in you not being afraid, right? So you're looking for things that happen again and again,

[7:46](https://youtu.be/1EeiaSsCaos?t=466) and then you're trying to make a prediction about what's going to happen next. So if we ask, what are we learning? Even if we're learning swimming or driving or anything, we're recognizing patterns and making predictions. And as our experience and our perspective expand and grows, we just get a lot better at it. So instead now, like at the top of this slide here, you see the shapes and the colors. That's kind of where we start in kindergarten. Now, all of a sudden, we're able to predict that the Halley's comet is going to come by once every 75 years, and humans were doing that before there was any kind of a computer, right, Caleb? So it's just like this increase of complexity of patterns. And I think I missed a slide. There we go. Yeah. So it's just this increasing complexity. And all of this doesn't really matter unless you can remember it. In order to be effective, you also have to be able to process it and connect what it means. You have to have language so that you can communicate about it. You have to be able to apply what you learn. And then after you apply it, that's learning by doing, which turns back into experience. And next thing you know, it's literally like everything's just going together.

[8:59](https://youtu.be/1EeiaSsCaos?t=539) And you can't really tell where memory starts and experience ended and training started and application ended. It's just all bucketed together into this one process that we call human learning. But even with this whole beautiful process, it's still not very effective unless there's one more key. And to demonstrate this, Caleb, I'm going to put you on the spot. Tell me, tell me five things and people can put in the chat to tell me five things that you've learned in your life.

**Caleb Williams**

[9:32](https://youtu.be/1EeiaSsCaos?t=572) Ooh, I'm looking at my garden right now. And I would say I've learned growing things from seed is very, very hard. Struggling with that one currently. I've learned that puppies chew on pretty much everything they can get their hands on. I don't have any kids yet, but I assume kids might do a little bit of that too. I have learned that water turns into ice if you freeze it. And I have learned that pens are better than pencils. Debatable, but okay, we're going to keep going on

**Jamie Kimes**

[10:09](https://youtu.be/1EeiaSsCaos?t=609) this. What I love about what Caleb said is number one, he didn't say one thing that was like anybody in the chat, right? You learn about happiness, how to ride a bike, that the one thing you can count on is change. But what's interesting about everything he said is that in this moment, in this session, none of it matters. Caleb is not here to teach us about puppies or gardening or anything. So he has to have this incredible ability to have all this stuff that he knows just fade to the background and to be able to focus on the part of his learning that matters the most right now. And that's what makes human learning so incredible. So how does this relate to AI? That's what everyone's asking. Caleb, this is where you come and make it complicated, okay?

**Caleb Williams**

[10:56](https://youtu.be/1EeiaSsCaos?t=656) Yes, this is exactly right. And so I think, Jamie, you'll kind of go through the buildup, but the big thing to remember here is when we're imitating the human learning process with these algorithms, with these models, we're doing the same thing. We are looking for those patterns and we are making predictions based on those patterns. And that's the really core thing. If you take away one thing from this session, as far as how AI actually works, finding patterns and making predictions based on those patterns. Absolutely. Like it's really that simple. Like

**Jamie Kimes**

[11:28](https://youtu.be/1EeiaSsCaos?t=688) we could pretty much stop this call right now, but we don't want to because we're having fun, right? So a couple of decades ago, there were these really smart people and they asked a very interesting question. They said, what if machines could learn? And the first challenge that they had was figuring out how to give a system or a machine experience. It can't go out and swim, right? It can't go out and ride a bike. So how do you give it experience? How do you think that you could give or simulate experience for a system or a machine?

**Caleb Williams**

[11:58](https://youtu.be/1EeiaSsCaos?t=718) And spoiler, I think the one of the correct answers was mentioned already in the chat for sure. So I'm curious what else you guys think. The internet.

**Jamie Kimes**

[12:07](https://youtu.be/1EeiaSsCaos?t=727) Great answer for Marie. I love that one. And what is on the internet? What's on the internet is a ton of information, right? The way we can simulate experience for a machine is by giving it information. So data, it can be spreadsheets. It can be books. If you want to teach it like about culture, have it read books. If you want to have spreadsheets, images, videos, and all of this for your first vocabulary word is a corpus. This is like the corpus and a corpus is to a system what experience is to the human. After that, you have to do all the same things. You have to train it. You have to label it, say, this is this. These are the categories. These are the rules. You have to repeat it, give it the same information again and again. After that, it has to have an ability to learn and it has to be able to learn from the training and it has to be able to learn directly from the data, exactly like humans. All of this is giving it a perspective. It's telling it what to look for. And in a system, we call a perspective an algorithm, right? So an algorithm is telling it what to look for and what is it looking for, what happens over and over, and what's going to happen next and what is it learning to recognize patterns and make predictions. And as it gets better and better and developers have gotten stronger and stronger, AI has simply improved in the maturity level of what kind of patterns it can predict, what kind of patterns and predictions it can make.

[13:31](https://youtu.be/1EeiaSsCaos?t=811) Um, so if you have really simple patterns, that's what we call about machine learning. It's very specific. You have to tell it exactly what it is. Once you get into like predicting Haley's comment and sentiment analysis and things that are really abstract and complicated, now we're into deep learning. And we could literally continue this metaphor all the way up. I've put all these jargon words about AI over here on the screen, and we could literally just dump them all into one of these buckets. But just like with human learning, they're going to get so enmeshed and mixed up that it's going to be really hard to know where one starts and the other ends. And we can just group it all together and say this is artificial intelligence. But even AI was not the game changing thing that it is now until AI had the functionality to be able to ignore 99% of what it had learned and focus on what mattered right now in the context. And in AI, we call that attention. And when it had that maturity level, that's when generative AI or gen AI was born, your chat GPTs, your Gemini's. And this refers to AI technology that makes, recognizes patterns and makes predictions so well that it seems like it's making something new. And Caleb, I will turn it over to you with that. Excellent.

**Caleb Williams**

[15:00](https://youtu.be/1EeiaSsCaos?t=900) And I think, yeah, there's some good foreshadowing in there with the language of, it appears to make something new. So one of the things we're going to dive into next is actually kind of opening up one of these Gen AI tools and really showing how it works. So similar to what I was saying earlier, you know, I really think there is no substitute for experience and seeing things firsthand when you're trying to learn. And so one of the things, again, we want to make sure everyone leaves this workshop today with is having seen one of these tools. So based on our entry poll, I think many of you likely have used one of these before. If you have an approved tool and you'd like to follow along yourself, absolutely, please do. We're on gemini.google.com. This is called a large language model, an LLM, a chatbot, insert whatever name you feel most comfortable with. And we're just going to go through a couple quick prompts just to really highlight what's going on here behind the scenes. And so, as Jamie said, this whole field's been around for a very long time, you know, going all the way back to the 70s and 80s, we were working on, you know, what some people would call artificial intelligence, but where it really started to kind of captivate, I think, everyone's minds in the general public here was when we saw the advent of these models like chat GPT, Gemini, Llama, all of those types of things is when it really started to click for a lot of people and started to become really accessible. But as we talked about earlier, it's still all working the same way, whether you're talking about machine learning, or deep learning, or natural language processing, or LLMs, like we're talking about right now, all of it is looking back at the past data, that corpus that Jamie was describing, and it's looking for patterns and using that to predict something. And you know, that might be predicting whether or not a stock will go up and down based on how it's performed in the past, it might be predicting whether or not you're going to click on this advertisement based on how you've interacted with a web page before,

[17:03](https://youtu.be/1EeiaSsCaos?t=1023) it might be predicting whether or not you're going to like this Netflix show, based on all of the shows that you've seen before. And when we turn to these chatbot type applications, it's still doing the same thing. So what is it doing here? It's predicting the next word in the sequence. So if you think back to, you know, the old school flip phones and texting where it would, you know, you'd start to type in the first couple words, and you do it by the number so that you could actually get the letters out, it would say, oh, you probably mean this word that was very, very early in this technology development was predicting, okay, you've typed a couple letters, you probably mean this word. Now, if you think about that, at scale, like someone mentioned, if you give one of these algorithms the entire internet as its corpus, and it's past data, well, now it can start to predict, what is the next word in that sentence? What's the next several sentences in that paragraph. And again, as Jamie mentioned, that's getting to where it starts to appear like it's predicting something new. And so to kind of illustrate that we're going to do a couple of very quick, very simple prompts. Jamie, if you want to throw the first one in there, we're going to ask this chat bot and you know, always start with your Hello, be nice. And say, you know, please finish the following sentences using only one single word. And so the first sentence we're going to start with is the sun was shining and the sky was. And so before we hit enter here, what do you guys think the next word that's going to be predicted in that sentence So the sun was shining, the sky was.

[18:30](https://youtu.be/1EeiaSsCaos?t=1110) Blue seems like a likely candidate. Like it. All right. Very good. All right. Let's see. Let's see.

**Jamie Kimes**

[18:44](https://youtu.be/1EeiaSsCaos?t=1124) Blue. Blue. Right. We learned our labels and our categories.

**Caleb Williams**

[18:48](https://youtu.be/1EeiaSsCaos?t=1128) We did. And if you look at the chat, it's I think a really great example here, which is that of the seven or so responses we had, six of them were blue, two of them were clear. So if we were to repeat this prompt, say eight times, we might get clear two of those eight times. But if we look at what the most likely response would be, it was blue. And that's what we get out of the actual model itself. And that's always going to be true when you're using one of these models. It's going to predict the most likely outcome. And that does very, very many implications to some of the concerns raised at the beginning of the session about their accuracy and different things like that. But one of the ways that you can get more accurate responses here is by giving it a better amount of context or changing what we're asking for. So now we're going to try a new prompt, which is the sun was behind a cloud and the sky was. And I see we're already starting to get some responses in chat. It was dark and stormy, gray. Very good. So let's go ahead and demonstrating

**Jamie Kimes**

[19:51](https://youtu.be/1EeiaSsCaos?t=1191) like how context changes things. And so even in your mind right now, it's happening just like that. And that's what AI is trying to get to. It wants to be as smart as you, where it's like that fast,

**Caleb Williams**

[20:02](https://youtu.be/1EeiaSsCaos?t=1202) you can change the context. Exactly. And all right, we have cloudy, misty, gray, gloomy, dark, dark. And again, we see gray. So see how we change the context just a little bit. And now we're getting a different response. And like Jamie said, just like you as a human want to do, that little bit of context, the fact that the sun was behind a cloud dramatically changes what we think the next word is going to be. So now let's get just like a little bit more complex. And we might even have to start kind of a new chat with this one. So now we're going to do something where we're not just asking for a single word, but our prompt is going to be, how would you continue this sentence? So the sentence is, a new program was launched to help students learn faster. The program uses dot, dot, dot, and put in the chat, how do you think using your human corpus, your human experience, a new program was launched that would help students learn faster. What would that be? And yes, to our question in the chat, when we start a new chat with one of these tools, you're basically resetting the context. So you are now getting essentially a clean slate.

[21:12](https://youtu.be/1EeiaSsCaos?t=1272) I like Anthony's response here, more well-paid teachers, Julia, interactive learning. All of these are great ideas here. So let's see what we come up with for Gemini. All right, and so we can see there's several different ways. We have focusing on the method, and then we see a number of different examples. We have focusing on the target, and again, more examples, as well as some kind of just more general suggestions. And I want to make two points here. One is, you notice when we did not include the context of respond with a single word, you can see we get a much more detailed response. So we have not given the model the context to say, we want this specific kind of output. So you can see a much more free text response. The same way when we ask an open-ended question, we get much more varied answers than if we say, hey, give us a single answer. And you can see some of those align a little bit with what we were talking about. I saw lots of AI. We can see there, it is not quite so biased that it immediately suggests itself, but I like where your guys' heads at. And you can see some of the different suggestions it has. Interactive software, similar to Julia's response, interactive learning.

[22:27](https://youtu.be/1EeiaSsCaos?t=1347) So again, we are getting out of this a huge amount more information and different suggestions of ways that that sentence could be completed. And so here, we are getting really more of a summarization of full ideas rather than just a single word or a single one. And so as a final example that we will not dwell on super long, but we do want to kind of make this very practical, particularly for public servants here. Let's do a prompt, Jamie. Hi, please create a draft government digital experience strategy. And so kind of before we send this, I want you to think about if you were tasked with that, what types of things would you want to see? And recognizing a government digital experience strategy is kind of an overly broad request. So there could be many different ways you can take it. And again, encourage you guys to think through and drop in the chat if you'd like different things that you think should be included in some type of successful

**Jamie Kimes**

[23:23](https://youtu.be/1EeiaSsCaos?t=1403) strategy. Yeah, Martha said risk assessment. And I'm sure that if you were doing something that was a strategy, you're probably going to have headings, you're probably going to have sections, you're probably going to have bullet areas, role-based strategy, security, roadmap, policy, capacity planning, governance. Yeah, there's a lot of different things. And I want to ask this, how long do you think it would take you to do that? Just to get to the draft. How many places, how many websites would you have to go to? It's going to take you forever, one week, a few days, or maybe a week. All right, can I hit go, Caleb? Yes, go for it.

**Caleb Williams**

[24:03](https://youtu.be/1EeiaSsCaos?t=1443) And so you can see here, we're already getting quite a detailed outline to start from. And it includes, you can see strategic pillars, getting to the strategy, key initiatives, getting to the roadmap, different goals. So all of this is the response that you're getting out of here. And I see a question from Jesse that are, what are the sources here? One thing to note, and that is one of the drawbacks that you have with these, often you aren't going to get a source. So again, remember, this is just the most likely response, thinking about all of the internet as an input. And that's where some of it's going to be really good and really, really close, because it's seen a lot of different strategies. And some of it is not going to be good at all. And so, yeah, I see Dana mentioning as well, there are, and also try to keep up with them. You are not getting, this is not preloaded with any code. This is just a raw request to the chatbot itself. You can request lists of sources, sometimes they're accurate, sometimes they're not.

[25:09](https://youtu.be/1EeiaSsCaos?t=1509) There's also tools like perplexity AI, that are really targeted towards trying to address that gap and provide better sources. But the big thing to take away from this is, again, you're getting this much more complex starting point. And again, thinking of it like a starting point, again, how long would it have taken you just to even get to here? I would not recommend citing AI as

**Jamie Kimes**

[25:31](https://youtu.be/1EeiaSsCaos?t=1531) the source, Michelle. No, that's kind of like, I mean, you say, who ate the cookies and you blame it on the toddler? It's like, no, you can't blame AI if something goes wrong. If I were to put right now, I could actually ask Gemini to quote its sources. It'll come back with links or whatever, but I would have to go back to each one of those and make sure that that was accurate before I used it in any type of an email or communication. And we'll get through, I see some additional

**Caleb Williams**

[25:59](https://youtu.be/1EeiaSsCaos?t=1559) questions on sources, and we will kind of come back to those during the Q&A. And I think that should be really fun to dive a little bit deeper into that one.

**Jamie Kimes**

[26:07](https://youtu.be/1EeiaSsCaos?t=1567) Right, I'm going to go back to the slideshow. Okay, very good.

**Caleb Williams**

[26:14](https://youtu.be/1EeiaSsCaos?t=1574) And so I think I'll turn it to you, I think, Jamie, to kick this one off. But so remember, we're trying to answer these kind of five questions today. We've gotten through what it is, you've seen a little bit of how it works. What can it actually do for you, though? Jamie, I'll turn to you for some of that. That's always the question. This is another

**Jamie Kimes**

[26:31](https://youtu.be/1EeiaSsCaos?t=1591) very human thing, right? What can it do for me? So it can do a lot of things. And we like to recommend going for the quick wins first. So I know that there's some of you on here who are definitely experts and probably more operational, and you're looking at that more large scale, full suite of AI, not just generative AI like the chatbots. But number one, right off the bat, it can save you time. What we just demonstrated here was you could get to anything. Let's say you're working in a program with public sector, and you need to do a program update. Let's say you have to summarize or look at policy, and you just need a quick idea of what it's about. Not necessarily all the details, but a quick idea. You can use it for summarization, you can use it for content drafting, drafting emails, drafting anything. And we've seen that people save usually on average about five hours on the low end per week. And a lot of people save like 10 hours a week just from using this stuff, which is time that I hope you go and put into like, learning something else like outside of work. But it can help you be more creative.

[27:38](https://youtu.be/1EeiaSsCaos?t=1658) Just the other day, I literally wrote a song, I had lyrics, and I was able to have AI generate music for it. Because you're saving so much time, you're able to reallocate that time into being more creative and more strategic. Go ahead, Caleb.

**Caleb Williams**

[27:55](https://youtu.be/1EeiaSsCaos?t=1675) And so some of the other things that can help you do, it can really help you to focus. So when you're kind of swirling around with many, many, many different ideas, and you need to try to get a better sense of, okay, you know, I have 10 different things to do, how do I prioritize those, this type of tool can help you actually look through, okay, here is what, yes, how I can address different topics right in here. If you want to go to the next one to Jamie, it can help you communicate. So I think one of my favorite ways to use this tool, and there's, I guess, there's two of them, as far as how it can help you communicate. One is, you know, I can be very passionate about things that sometimes shows through in my work, in a good way, sometimes in not so good a way. So one of my favorite things to do is help it rewrite email, say, hey, is this coming in too hot? Is this coming in the wrong way? How might this be perceived? And I can use that to then filter through my email and say, okay, here's a better way to phrase that. Here's a way that's approaching that with more empathy. You know, here's a way that's approaching that in more of a collaborative way. The other thing I think it can help you do is it can help you learn, and they're really, really, exactly, helps with tone and structure. And that's one of the things it's really good at. It can also help you learn to the communication. One of my favorite things to do is ask, you know, these tools and say, hey, explain this to me like I was 10, because I don't really get this. And it's far too complex. And it can break it down and say, okay, here's another way to think about it. And that's really, really powerful. And again, to some of the risks we've talked about, you have to be careful to make sure that you vet those sources. And you don't just take it at surface level and go, okay, great. Now I know everything about this topic, treat it as a starting point and treat it as a way to get you comfortable with breaking those things down.

[29:37](https://youtu.be/1EeiaSsCaos?t=1777) And then from there, launch out into the actual learning process that you've been doing for, you know, 30, 40, 50 years. Martha said it can help you be nice when you're not feeling it.

**Jamie Kimes**

[29:49](https://youtu.be/1EeiaSsCaos?t=1789) I love that because it's very true. It's like, yes, I have to be nice. Be nice for me. It's

**Caleb Williams**

[29:54](https://youtu.be/1EeiaSsCaos?t=1794) outsourcing nicely. And I think at a high level, one of the things We did a survey when I was working with Jamie around, hey, how is AI being used? And we were able to pull from this awesome data source with the federal government to kind of look across something like 700 different use cases about what people use AI for. And some of the big things we see identifying, detecting things in documents. So looking through and being able to say, yes, this document has that or no, it doesn't. Looking through images and finding things as well or forms and automatically extracting that data. Summarizing or synthesizing content. Doing actual forecasting. So all of this stuff you can use AI for. And it's actively being used for all across the federal, state, and local government space. Short plug here because I think you guys are really going to like it. We have kind of put all of that together into a big dashboard. So you're more than welcome to check this out. Highly encourage you guys to as a way to jumpstart your thinking on, hey, how can this be used for me? Over 500 different use cases all being used actively across federal government sector. Absolutely check it out. It's super fun. It's interactive. You can search. You can filter by agency. And you can look at all sorts of other ways that folks have been using this across the public sector. So highly encourage you guys to check it out. I see Jamie's got it pulled up here. And we're looking at using it for comment analysis.

[31:23](https://youtu.be/1EeiaSsCaos?t=1883) So I know a lot of you guys often have to collect inputs from constituents. And yes, absolutely, we'll drop that in the chat. But so I know many of you guys are using it to actually look at, hey, what are folks saying? You want to know, are these comments positive? Are they negative? You can absolutely be using AI for that. I love this so much. And we're going to put this

**Jamie Kimes**

[31:43](https://youtu.be/1EeiaSsCaos?t=1903) in. We'll have a form at the end. And we'll have all of the links that we include in this next little section in there. So we will have it in there. Definitely, definitely check that out. We're back to where we started at fear. We didn't start with fear. That was the whole point was we didn't start with fear. We started with curiosity. So now we're back to risk. This is the thing, I feel like it comes up over and over and over because public sector is very risk adverse for a lot of reasons. The work that you all do impacts so many people, millions and millions and millions of people. And nobody wants to end up on the front page of the paper. There's just a lot of risk aversion. And the common fears that we hear are, what about my data? What about security? What about job replacement? What about copyright infringement? We hear about this a lot, especially from artists, musicians, authors, artists. They're just like, does this mean somebody can just type in, please write me a novel in the voice of Stephen King?

[32:48](https://youtu.be/1EeiaSsCaos?t=1968) And next thing you know, we have all this stuff that's not really Stephen King. So copyright infringement is a big one. Environmental impacts. Caleb, I'd be interested in your take on this, but I feel like this is growing in terms of a concern. You're nodding. Go ahead.

**Caleb Williams**

[33:04](https://youtu.be/1EeiaSsCaos?t=1984) Massively. And I think we even saw that in the chat as well when we were talking about it earlier. And I think as we go through, I think there's one final one as well. Unclear reality. That was the one I was looking for. Even the Terminator or iRobot. And I think the big point that I want to make with this is, hey, these are actually all real fears. Maybe not the Terminator so much, but these are absolutely founded, whether it's environmental impacts, whether it's the copyright infringement, whether it's the job replacement, all of those are real. And we're not here to tell you, oh, it's all going to be fine. This is a big, massive technological shift, and it really requires changing the way of thinking. And I think as many of you know, because you live this day in and day out, these are all risks that have to be managed and they have to be mitigated every time you start adopting new technologies, and particularly for this technology. So what we really want to encourage you guys to do today is actually look at this. Don't hand wave it away. When you guys implement this, absolutely manage and mitigate those. And we're going to drop a link in the chat as well. But Jamie, if you want to kick off on some different ways that are really, really critical for mitigating these risks and these fears. Yeah. The first two that we say right

**Jamie Kimes**

[34:22](https://youtu.be/1EeiaSsCaos?t=2062) off the bat, and I love talking about this specifically with Innovate US, because Innovate US does an amazing job at this. And I could like share the screen of their AI literacy stuff that they've got going on. But you've just got to be educated and keep learning. And if you're a team leader, or if you have any place of leadership, make sure everyone on your team is literate about AI. You also have to have a community where people can learn together, make mistakes together, and not only do that, but fail together. Like come in and say, not only did this not work, it bombed. And we did end up on the front page of the paper. And this is what we learned about it. And just having that very collaborative, safe community is essential. And actually, I'm going to go off script a little bit here, Caleb. But I think Anthony Fisher is on the call. And Anthony, are you able to take yourself off mute? Yeah, I think so. Everybody can hear me. Yeah. So Anthony is like the king of governance. And I'm going to share the screen. But he helps with the GovAI Coalition. And Anthony, how about you just share a little bit of why

**Audience**

[35:30](https://youtu.be/1EeiaSsCaos?t=2130) governance matters? Oh, man. I don't know how much time I really have to go through all this. I could talk about this all day. Make complicated things simple, okay? Just make complicated things simple. Hey, folks. So yeah, I'm from the GovAI Coalition, one of the founding board members, and I work for the Colorado Department of Revenue. And the reason why governance is important is really to make sure that we're guiding our users thoughtfully through this journey of AI. It's a new time in technology for all of us, even for folks who've been in the space for 20, 30 plus years. And so without governance, it's just kind of the wild west. And so what we started doing at GovAI is really creating a lot of templates and policy guidelines and different frameworks that you can use in your own agencies. And even if you're here from the private sector in your organization, to really help guide that journey to bring AI into the organization thoughtfully and responsibly. That was wonderful.

**Jamie Kimes**

[36:30](https://youtu.be/1EeiaSsCaos?t=2190) I've never heard you say something that fast, Anthony. So I was like,

**Audience**

[36:34](https://youtu.be/1EeiaSsCaos?t=2194) I know, right? I was like, he's going to keep going for a little bit.

**Jamie Kimes**

[36:37](https://youtu.be/1EeiaSsCaos?t=2197) Oh, yeah. So let's go back through here because this is just a refresher. But yeah, governance. Thank you so much. And look, we're going to give you that link for GovAI too. All the links will be at the end. But Anthony has done a great job and GovAI is a great place to start with that. Go ahead, Caleb.

**Caleb Williams**

[36:53](https://youtu.be/1EeiaSsCaos?t=2213) Absolutely. And the last couple we want to touch on, the other one is talent investment. And that's not just finding people like Anthony or Jamie or myself and saying, OK, we're going to go bring those people in. It's making that homegrown. As Jamie said, establish literacy programs, establish skilling programs for your staff so that not only are you bringing in experts when experts are necessary, but so that everybody on your team is actually growing their skills in this area. And especially as we think about mitigating things like job loss and stuff like that, it's really, really important that we're all getting up to speed on this. I think Mark Cuban had a great quote maybe a couple of weeks ago and said, man, if I were 18 right now, the only thing I would be doing is learning everything I can about AI, because like it or not, it's not going anywhere. And the best thing you can do is really understand it, learn where it's useful, learn where it's not, learn where it's a great application and where it's really dangerous, and really help to make that investment into your talent both

**Jamie Kimes**

[37:53](https://youtu.be/1EeiaSsCaos?t=2273) internally and externally. I do want to point out with that, though, too, GovAI does a great job. They talk about finding your AI champion. So it's not necessarily the person who is that tech whiz who's just amazing. It's that person who loves this stuff. And so not everyone has to love this stuff. And I would really recommend, especially if you're an executive, even if you learn about it, don't feel like you learn so much that then you just go and make all call all the shots for your operational team. You know, it's like, we do need to have some of that distributed out. But yeah, just find the people who love it. Because it's a place where we're really growing. And Caleb, human in the loop, we cannot go through anything without that.

**Caleb Williams**

[38:37](https://youtu.be/1EeiaSsCaos?t=2317) No, we cannot. And this is a really critical concept across AI, which is we touched a little bit on earlier, don't just take what it gives you and go right with it. You are the human, you are in control. And you need to stay in the loop. I think one of the most risky AI or the riskiest AI applications you can have is when you try to remove the human entirely. And there are some applications where that's totally okay, and totally acceptable. But for a lot of things, you really still want the human sitting there making the decision. So don't don't cede our power to the robots, especially just yet. It's really, really critical that when you're looking at how to develop workflows, deploy AI strategically to help and to speed up that workflow. Don't try to remove the humans from it entirely. Which basically just means don't have

**Jamie Kimes**

[39:25](https://youtu.be/1EeiaSsCaos?t=2365) the AI making the decisions. You don't have them AI making the decisions on who to hire, or who to fire, or like who gets benefits, or who absolutely committed fraud, like, it can inform you and give you information, but not make the final, final call on that. And finally, be nice to the robots. When we talk about the Terminator fear, Anthony, you laugh at this one, right? Because I say this all the time. We talk about the Terminator fear, we don't really know where it's going. So just to hedge our bets, just say please just say thank you. Remember, it's starting to learn a lot like we do. And it's learning about us. So let's teach it that we're great. And we should we should stay around, right? Be nice to the robots. Here we are wrapping up towards the end. How do I get started?

**Caleb Williams**

[40:11](https://youtu.be/1EeiaSsCaos?t=2411) We are and keep an eye on time. I think we're doing all right. And we should have about five to 10 minutes left at the end here for q&a. But exactly right. I think the last thing here, how do you get started, and really the best way to learn is to use it. If you're just sitting around, and you're just looking at things, you're not getting the full experience. And so, you know, some of you guys may already have AI tools approved. For those of you who don't, you know, starting at home with gen AI is a great way to get your feet wet. And I'll turn it to you, Damian to

**Jamie Kimes**

[40:44](https://youtu.be/1EeiaSsCaos?t=2444) talk through some fun ways to do that. Yeah, so just start at home, no matter what, like if you don't have it, get a paid version. So we recommend not using the free ones just for security reasons and everything, get a paid version of either chat GPT, Gemini, google.com, Copilot, Claude, something. And then this is a fun one. It actually comes from Gary Wagner. He's a Google trainer with resultant. And he literally goes and he takes a picture of what's in his refrigerator. And then he puts it into like Gemini. And he's like, What should I make for dinner with this? And it'll like sit back recipes. I love it. And another great one is just next time you want to google something, just go to chat GPT or Gemini or something and use gen AI instead. And just start comparing what the results are. You can interact with it. It's a little bit more fun. Keep learning no matter what. Always learning because that's what if we don't have anything that to teach it, it's not gonna keep wanting to learn from us. And so, yep, so the first thing first, you know,

**Caleb Williams**

[41:48](https://youtu.be/1EeiaSsCaos?t=2508) we don't want anyone out here, you know, going rogue using their chat GPT or Gemini when it's not approved. As Jamie mentioned, part of the reason we encourage you guys to look at the paid options even at home is because you want to be really careful about not sharing data that should not be shared. However, many, many states, many, many local governments at this point have approved some gen AI tool. So talk with your ID department. If you don't know, see if you have an approved tool. And then assuming you do, there's a number of different things that you can use it for. And some of the great ones to start with, summarizing documents, ask it, hey, what is this saying? Drafting emails, asking it to outline a slide deck. Again, remember things that can get you started and then you can stay as the human in the loop and iterate from there. And one thing that maybe is a little different about these tools than like a normal Google search is it is that iterative part. When you ask it something like we showed, context is everything. So if you don't like the response, give it some more context. And so we even have this awesome kind of basic prompt structure that we'll walk you through real quick that helps to get really good results out of these tools. And I think we've got, we start with high,

**Jamie Kimes**

[43:04](https://youtu.be/1EeiaSsCaos?t=2584) which is the white box here. I don't know why it's the white box. Do not leave it out. Like this is to mitigate the terminator risk, right?

**Caleb Williams**

[43:13](https://youtu.be/1EeiaSsCaos?t=2593) But one of the ways that you can kind of structure these prompts, give it the instructions. So something like, hey, please generate a meeting outline for our next update. Provide the context, which is what's important to qualify those instructions. Say the audience is our new branch manager who just joined last month. Make sure you tell it what output structure you want. So as we kind of saw earlier, if you leave that open-ended, you're going to get an open-ended response. So if you would like something very specific, include that. And again, kind of structured as, hey, this is the situation, this is the context, this is how I want you to format the output. And then if you have a long list of additional information, say a document or a draft email you already wrote, kind of put that at the end. And if you follow that, and again, make sure your thank you, again, mitigate that terminator risk. If you follow this prompt structure, it'll really help you to get a lot better results out of the output. So I encourage you guys to check that out. Oh, there it is. There it is. It comes in dramatically at the end.

**Jamie Kimes**

[44:14](https://youtu.be/1EeiaSsCaos?t=2654) Yeah. So we are coming up on time. We want to leave time for questions. I'm going to put a link to a form in the chat. It'll have our contact information. It has some of these resources from the session, also some other things that Caleb threw in there. And I would love if you just fill that out, give us some feedback. I want to say really quick, if you are interested in any type of literacy for your organization, I've partnered up with Josh Martin. I'm sure some of you know him. He was the former chief data officer for Indiana. And he and I have put together something called the AI Dialogues. So it's an AI literacy program. It takes as short as eight weeks to get your executives, your operational management, and your just day-to-day users up to speed on AI. Caleb's doing all the really complicated stuff.

**Caleb Williams**

[45:03](https://youtu.be/1EeiaSsCaos?t=2703) And absolutely, guys. I love also, Jamie, the human-centered AI part there, because I think it's just so essential. Definitely, if you find yourself actually looking to kind of implement different programs, would love to hear from y'all. And as I mentioned, please check out that dashboard. It's completely free. It will always be free. It is really just meant as a way to get people more educated on how AI can be used. And if you have any questions, comments, or feedback, again, please don't hesitate to reach out. And I really hope you guys get some use out of it. And speaking of getting some use out of things, I think Jess will put up one final poll. And our hope here is that we have a little less normal distribution, and that we have pushed you guys a little bit from the left side of the scale over to the right side. And so whenever we get a second, we'll flash up that poll just to kind of see how everybody feels at the end.

**Jamie Kimes**

[45:59](https://youtu.be/1EeiaSsCaos?t=2759) And yes, Laura, I'm about to, I'm gonna stop sharing here, and I'm going to put the link to all this stuff in the chat.

**Caleb Williams**

[46:09](https://youtu.be/1EeiaSsCaos?t=2769) All right, and we do have the poll up. So again, remember, originally we were pretty centered around the five, six region. I'm seeing at the moment, no zero through three, though, if you're still there, that's totally okay. And absolutely looking forward to hearing from you guys and continuing to engage. But definitely excited to see that, yeah, our normal distribution has shifted. We're pushing up towards an average of, it looks like six or seven at the moment. So that's absolutely fantastic. We're really glad you guys enjoyed it. We're really glad you learned something. And yeah, looking forward, we've got about 10 minutes left here. So Jamie and I are more than happy to kind of take different questions and try to touch on some of the topics that you guys are still looking for.

**Jamie Kimes**

[46:51](https://youtu.be/1EeiaSsCaos?t=2811) Yeah, so that's it. Like, Jess, do you want to take questions or how does this part work?

**Audience**

[46:56](https://youtu.be/1EeiaSsCaos?t=2816) Yeah, absolutely. So I'm going to stop the recording.

* * *

Original page: https://innovate-us.org/starting-with-curiosity-a-beginner-s-guide-for-public-servants · Licence: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org