Otan: Outreach and Technical Assistance Network.

Neda Anasseri: Hello, everybody, and welcome. Thank you so much for joining us today. My name is Neda Anasseri, I'm Technology Projects Coordinator for OTAN. We are so happy that you can join us for today's webinar. We are here with Joe Marquez, the Director of Academic Innovation at CUE, and he will be doing the webinar, AI Prompt Mastery: How to Become a Prompt Genius. And so we all know, out there, many of us are using those AI tools and we need some help on, how do we ask those tools? How do we work effectively with the new AI tools? We have the best person to give that webinar today. So I'm going to hand it over to Joe. Thank you, Joe, for joining us. If you don't mind introducing yourself.

Joe Marquez: Yes, absolutely. Thank you so much for having me here today. My name is Joe Marquez. I am Director of Academic Innovation for CUE, and I'm so excited to be able to speak with you today, because generative AI tools have seemingly came out of nowhere back in November 29 of 2022 and have taken over industry, businesses, and they will be flowing into education, in the K-12 space, in the higher ed space, and in the adult ed space.

And one of the things that I talk to educators about all the time is if we don't talk to students, it doesn't matter what age, emerging students, emerging learners, K-12, high school, college, adult ed, if we don't talk to them about these tools they're going to use them and they're going to abuse them, right, because they're not going to know how to use them. And it's up to us to guide them in a proper way.

And it's not just willy-nilly typing things in and hoping you get something out. There's actually a process to it that I'm going to showcase for you as well. And I do want you to know, if you ever hear me use the word pedagogy, please flip it to andragogy because I do come from the K-12 space and I do know there's a lot of adult learners in here, teachers of adult learners, and that's andragogy. I'm not talking down to you if I use pedagogy; please just know I'm talking about a shift in how we approach our teaching and learning with these new tools.

So and there is a chance in here for you to ask questions along the way also, so please feel free as I'm talking to ask your questions if I miss it, Neda is going to be able to let me know that I missed a really good question, and I should be able to stop and answer it. But please know that at any moment, you can ask that question. Also, I have a Padlet in here that has every piece of resource that I'm going to be sharing with you. It has the slide deck in it. It has my contact information in it, and it has every single article that I've read which I've added information from into my presentations.

And the question I always get is, Joe, did you make this presentation, or did AI make this presentation? I made this presentation. I love making presentations, and so I don't want AI to make my presentations. And that's the thing that we need to understand as well, is AI is going to be able to do a lot of stuff. If there's things that I can do that you love doing, don't let AI do it. You do it. You get to decide how much or how little you get to utilize these tools.

So I'm to go ahead and share my screen so that we can definitely get started. Here we go. Sharing that screen for you right here. Excellent. There we are. And so we're going to look at how to become a prompt genius, and all that means is how can we talk to these tools and get the most out of them. There isn't really going to be a situation where you prompt these tools and you're going to get a 100% usable piece of information. You could maybe get about 70%, maybe 80%, if you're really good at prompting, maybe 90%, but you always need to look at what you get out of it, comb through it, make sure there's no hallucinations. That's the term we use when these tools make things up, we call them hallucinations. So, making sure that we are going through and looking at that draft, making changes to that draft before we do anything with them. That is absolutely key in all of this.

Again, I did already place this Padlet here into our channel. If you missed it or you need it again, we'll definitely post it in again for you as we are going through our session today. And again, feel free to ask questions in the chat as we are going through this learning period today. The biggest thing on this page is my email. That email is in the Padlet, but it's also there. You have complete access to me after today. If you have any follow-up questions you have for myself, you have an idea of how you'd like to use this but don't know where to start, please know that you can contact me and we can have those conversations via email. If we both have time, we can have those conversations via Zoom. But know that you do have access to me to ask those kinds of questions even after we are done for the day.

All right, moving on forward. Also, I am the Director of CUE, and we do have some really great conferences, so, some shameless plugs: if you are in the Northern California area, we are having a fantastic conference called Fall CUE up at Teachers College of San Joaquin in Stockton, California, and that's going to be October 21st and 22nd. Everybody says, OK Fall CUE is great, but what about Spring CUE? We all want to go to Palm Springs. That's happening March 21st to 23rd, and we are looking for a lot more higher adult education appearances and presentations. So, know that we are trying to meet those needs as we grow into all the education spaces.

And CUE is a membership organization. This presentation, I have been evolving since January, and all members get free access to these. There's no cost to them. I do them once a month with all the new updates, and they're 100% free for CUE members, and we usually charge like $20 if you're a non-CUE member. CUE membership is 40 bucks, so you go to two of these you kind of make your money back. So, being a CUE member has helped me become a better educator, and so as in this role. I want to help other educators be the best educator they could possibly be. And I always say, the first step towards doing that is becoming a CUE member today.

All right. So here we go. Yes, all of these slides are on the Padlet and available for you to take a look at and peruse at your will after this. But I want us to take a moment and just understand the moment that we are in. We are in a moment that used to only be science fiction. We are in a moment where we can actually ask a question or have a conversation with the machine, and it will respond back to us in a convincing linguistical way. It will respond nearly identically to how a human would respond. And it does a very good job at it. It is changing the way that we communicate and ask these machines to do things for us.

That's why we are now in, as Thomas L. Friedman remarks, we are in a Promethean moment. Promethean moments are moments in human history where something is discovered or invented that changes the way we think about not just one thing, but everything. Changes the way we do things, not just one thing, but everything. How we communicate, collaborate, disseminate, create. This is one of those moments. There's been a lot of Promethean moments over history, which takes a large gap in humanity and brings it closer together, like telecommunications, the telegram, the telephone, the television, not the Teletubbies, we want to forget about those guys.

But these are all Promethean moments that we've been around in. The latest Promethean moment I would say that has bridged the gap of creativity, communication, and humanity is the internet. But we have to remember, the internet was created in the '80s. It started to really gain steam in the '90s, but in the '90s, you still had to dial up the internet, and occasionally it was busy. I mean, you try telling that to kids today, that you tried to dial up the internet and you were put on hold. Right? That was the world we lived in, in the '90s.

And then in the early 2000s, once the infrastructure started to get faster, and DSL and cable modem started coming into play, we started getting really fast ways of communicating and collaborating in the early 2000s. I blame Napster for that, because so many college kids got into the internet for trading music. And then in the K-12 space and in the college space, it wasn't until like e-rate came out where we started to really seeing infrastructure getting shored up and more devices being able to be deployed on college campuses and other campuses.

And now, it seems like everybody has a device. But it took a long time. This moment that we are living in it seemed like it came out of nowhere. Unless you love reading scholarly articles, you may not have known that this is where we were going since 2014. But this, when it dropped, November 29th, 2022, it was instantly accepted, because everybody had the device to use it. Everybody had the infrastructure to use it, and you didn't necessarily need too much skill to use it.

It is the fastest adopted tool in the history of the internet. So, we are living in a moment that is just moving forward, barreling ahead, and we have to be able to take hold of it. We have to be able to control it. And that's why when we're looking at these things, we truly have to understand the difference between a disruptive technology and the discovery of a technology. Too many times, I hear people going around saying, we have to disrupt education, we have to change the way we do things, we have to start a revolution.

But that is something that's bullish. That is something that-- is you're trying to push your way forward. We're looking for discovery. How can we discover, as educators, how can we discover how this tool can help the creation of new lessons? That can help differentiate instruction for our students? To help our students who may not understand the material to utilize these tools as a tutor to help gain the understanding of their subject and topic through the means of these devices? And the only way to do that is by discovering it.

So, understand the difference between disruption and discovery is a disruption is when we keep our heads in the sand and we get shocked when something comes our way and disrupts the way we teach. Discovery is when we see it coming and instead of running, we run towards it, to discover what it's all about. That's what we're doing today. We are discoverers in this journey, and you should be empowered by that. And you will never be disrupted when you are a discoverer in this journey.

So, what is generative AI? You hear the buzzword ChatGPT being thrown around everywhere. That is just one tool in a vast amount of tools that are out there. Generative AI is any tool that uses an LLM, that stands for large language model, to allow a user to put in an ask or a prompt, and then receive an original output. And that could be text-based, that can be image-based, that can be audio-based, and pretty soon, it can be video-based. Imagine being able to watch a 20-minute TV program just by you asking this machine to make one for you, without any actors or anything. Crazy. That's why the actors and the Writers Guild went on strike, to ensure that that doesn't happen.

But that would seem so far-fetched about two years ago, that we wouldn't even think that it's possible. But now, there's programs out there that can create five-second clips of whatever you want. Now, 5 seconds is vastly off from 20 minutes, but we are at the precipice of this. We're only at the tip of the iceberg in finding out the power of these tools.

Today, we're only going to be looking at the generative AI text generators. That's what we're going to be looking at. Maybe if you are enjoying this and you want to dive into some image generators, we can have another session. Maybe Neda can set that up. But today, we're only going to be focusing on textual output from these tools, and the tools that we're going to be focusing on are like ChatGPT, Bard, and another one called Claude.AI. I'm going to tell you what each one does. I'm going to give you a quick example of them. But I want you to choose your favorite for today. A lot of people choose OpenAI, which is ChatGPT, but then I see a lot of individuals going towards Claude as well. And I'll show you the benefits of both.

But before we get started, you have to know how these tools work. You have to know what is going on over this before we even get started. OK? So this is what is called a deep learning neural network. It was built modeled off of the neural networks of our brains, on how we take external outputs from our senses, our auditorial, our visual, olfactory, and how our brain interprets that as smells and sounds and linguistic sentences. That would be the output, what we feel, what we see, how we react to it.

These machines were built exactly off of those same ideas. But instead of external outputs being senses, it is a prompt that we give it, an ask that we give it. And then once it enters the system, it bounces off nodes of neurons, just like we would have in our brain, bouncing off, looking for patterns, crafting a response, and when it finally does, it shoots it out to you. That is the output.

The thing about this is it is a completely original, new creation. You can't search it on the internet. It wouldn't be considered plagiarism because you wouldn't find an exact replica of it anywhere. And it is brand-new. But the thing about it is as well is you wouldn't be able to replicate it if you tried. If you had 40 students in your class and they all put in the exact same prompt, they are all going to receive 40 individual unique responses. They may be similar. They may have similar facts, but they'll be written all differently.

And that's a misconception people have about these things, is they think it's searching the internet, finding words or sentences, and then quilting them together. That's not what it's doing. It's like smashing a window up, mixing all these different windowpane colors together, and then grabbing individual pieces of window to craft a brand-new mosaic. That's what's happening here. It's very unique.

And the thing about it is, the beautiful and scary thing, is even the creators of these tools have no idea what's going on in the middle. They have no idea. They're conducting experiments now to try and figure it out. But they have no idea how it does what it does, which is pretty amazing. But that's the way our brain works as well. Even neuroscientists don't completely understand how our brain works. So we don't understand how these neural networks work.

But if you wanted to explain this to some of your adult learners or anybody else, I always love to showcase Plinko. Because the prompt would be the Plinko chip, and the input is where you place it. Everybody hopes to hit that $10,000 mark, but that's not always going to land there. It's all going to depend on where it hits on these nodes all along the way. And you could put this chip in the same spot 10 times in a row, and it will land in a different spot as much as possible, but definitely hit a different node on its way down.

So when I talk to adult learners, this makes sense, this is relevant to them. But when I talk to younger kids, they don't understand this because they don't watch The Price is Right, they have no idea what Plinko is, so I have to use a different term for them. And it's this image. This image is probably one of the most powerful images you can have of what generative AI is. This picture is from Richard Donner's 1977 film, Superman, and in this picture, you see Superman off to the left asking his father a question. That is a prompt. The prompt, the ask, is his question. The output, the creation, is a response from his father.

And the reason this is generative AI is because his father is dead. You see, his father and Superman lived on a planet Krypton that was about to explode. And his father wanted to save his son so he put him in a space crib, and shot him off to Earth. But before he did that, he grabbed a crystal. And in this crystal, he shoved and programmed in as much information that he could into that crystal, the entire history of Krypton, every novel, every poem, every music, everything he could possibly shove into this crystal about planet Krypton, he did.

He also put the entire history of the El family, which is the family that Superman belongs to on Krypton, and the entire history. And also his father put all of his memoirs into that crystal. When he shot Superman off to Earth, and when he became of age, the crystal was thrown and created the Fortress of Solitude. Think of this crystal as the large language model, and all the information that his father put into it, that is the data set. A data set is what all of these companies are putting together to make these machines work.

So when Superman speaks, the crystal hears it and projects his father, and it responds as if it was his father. But it's not. That is what all of these tools are doing. They are listening to our prompts, and responding as if they are a human, trying to give us an answer or give us what we're asking for. It's not far off to get to this point. I was going to do this, but my audio wasn't necessarily working, but OpenAI is now an-- ChatGPT is now an app for your phone, and they now have a voice feature where instead of typing in a prompt, you can ask it whatever you want for as long as you want, and it will respond in a very humanistic voice. It's very interesting.

So, we're getting close to this situation. In fact, we're getting so close that there are companies that are trying to replicate this for relatives who may have passed away. And before you say, oh, that's macabre, I don't even want to hear this, I want you to imagine, if you could grab every piece of history of your family, all the diaries or memoirs or anything from your family members, and create a data set for them when they pass away and be able to have a tool to ask them questions still and get a response back, would you do it?

Again, that's a little bit macabre, others would say, but I would imagine, if you were to walk through a cemetery, instead of just knowing about a person because what's on their headstone, you can now ask them questions about their life, doesn't it make burial a little bit more meaningful in that fashion? Right? Don't get me wrong, it's not like I walk around graveyards for fun or anything, but if you had to and you were able to, oh, that's an interesting tombstone, I want to ask that person a question, and you were able to do it, would you?

Now, in the age of YouTube and podcasts, I have a lot of data out there, and there could be a lot of data that can be sucked into these things and pretend to be me when I'm gone. So, it may seem otherworldly at the moment, but it's not. But let's bring thing closer to today. Imagine that you are working with some of your learners, and they are learning about the history of the United States, and you want to have them chat with some of the founding fathers. Could you do that? Right?

Imagine if there was a tool like an Alexa in the classroom, and you said to it, hey chat history, you are now going to be one of the founding fathers of the American Revolution named Benjamin Franklin. You are now going to receive questions from my class, and you will respond as Benjamin Franklin, utilizing all the information you have in your data set about Benjamin Franklin. Do you understand? And it would respond, yes, I understand. I am now Benjamin Franklin. Have your students ask me anything they'd like.

That seems futuristic, but you can do that now. I will show you a prompt where you can do that now. I will show you a prompt framework that will actually let you talk to anybody as long as they have a biography or they've been studied. You can do it. And I'll show you how to do that. And it's a great creative writing process for students to write a prompt, to talk to a historical figure. It's fascinating. We'll definitely get to that.

But before we do, we have to talk about some of these tools' limitations. All right. Previously, I would tell you that some of these limitations are they're only limited in knowledge up to September of 2021. But Google Bard is now completely up-to-date. ChatGPT, as long as you're using the pay version, it is completely up-to-date with Microsoft Bing capabilities. If you are still going to use the free version, which is still perfectly fine for our purposes, it is limited to information up to September of 2021 at the moment. And then I don't know how far Claude is. I would say between September of 2021 to current. I'd have to read some more of what they've added to their data set. So one of the limitations is how far back you want to ask the questions and which tool you're using to do that.

The other thing we have to understand is, where is it getting this data from? Well, thinking back to the Superman analogy, the idea is that you have these companies all working off of the same LLM model, the large language model. In fact, Google is the one that invented the large language model. And they let it go out as open source. Anybody can use it and play around with it and discover with it.

So the thing that separates these companies apart is the data they are feeding it. They are not all feeding it the same amount of data. But the data they are feeding it does come from the internet, which causes worry, because as we know, not everything on the internet is true. So what are they feeding these things?

Well, the only information we have at the moment is from ChatGPT 3.0, which was from 2020, a while back. ChatGPT won't let us know where 4.0 came from. Google's not necessarily telling us where they're feeding Bard from. And so where is this information coming from? So, the latest we have on this information is from this table that I have here. And that is 60% of the data set in ChatGPT 3.0 came from the Common Crawl. The Common Crawl is a nonprofit organization that scrapes the internet of metadata. And it pulls this data in and then and their data scientists then scrape that data and sift through it.

Imagine if you're at this data set, or imagine if the Common Crawl is the town dump. Just piles and piles of junk from everybody. And the data scientists sift through all the junk and pull out what they deem is quality information. That's what's happening with the Common Crawl. But it's people picking and choosing what they deem as quality information. So that is questionable.

Next is Web Text 2.0. That is 22% of the weight in this data matrix. And that is Reddit. Anything in Reddit with three or more upvotes gets pulled into this. Now, I'm not saying they're gathering facts from Reddit. In Reddit, it's learning how humans communicate. How do humans ask questions? How do humans respond to questions? How do they argue with one another? How do they showcase facts versus not facts? So, that's what Reddit is for, it's to train these things how humans communicate with one another, not necessarily facts.

And Books One and Books Two, that's a large repository of ebooks, so it pulls from that. That's so it can identify different styles of writing in different genres, romance versus horror versus comedy. That's what it does there. And then finally, Wikipedia. Wikipedia is like sea salt. Everybody's got a sprinkle a little bit at the end. That's the reason I think they use Wikipedia right there.

But this is the latest data we have, and this is nearly four years old. Why don't we know where this data is coming from? A couple reasons. One, since everybody has access to how to code an LLM, it's the fuel you put into it that really makes the difference. So they say their data sets are what set them apart. Therefore, it's proprietary information they don't want to share. That's the outward reason why they don't share.

The internal reason is there's a lot of lawsuits happening right now, of authors and contributors and artists who say that their data has been sucked into these data sets to train these models how to work, and they didn't give them permission. It's not necessarily plagiarism. It's not necessarily pirating. It's a new form of creation, and there is no laws on the books on whether it's legal or illegal. And there's a lot of lawsuits going through the motions right now to determine all of those things.

So, in the Padlet, underneath, there's a column about-- I think it's 4 columns over to the right, it has all the different articles I can find about a lot of the lawsuits that are happening and the legal implications if they find for the plaintiff or the defendant. But for at the moment, the data sets that are in there are from the internet, from humans who have written things. And we have to be aware that humans throughout time are inherently biased. They have inherent ways that they see the world. They can be biased they can be sexist, they can be racist, they can be harmful, they can be hurtful, just as humans can be honest and joyful and triumphant and cherishable.

We can see the good and the bad in humanity in these data sets. And we have to trust that these companies are doing what's right to put up guardrails to protect us from a lot of bias and hurtful information. But that's a lot of trust to put in billion-dollar companies who probably see dollar signs before anything else. So, just be aware of that, and these are things that I tell teachers and administrators and students so that they understand, always take the output with an eye to understand that sometimes it can be harmful or hurtful. And it's not always truthful, and we have to be able to talk to our learners about these things.

All right. So here we go. Let's dive in. I'm going to showcase some of these tools first. I want to show you the pluses and the minuses, and when you would possibly want to utilize some of these tools for use. First off, though, I want you to know that none of these tools were created with the educational ecosystem in mind. OK? So, I want you to be wary about posting any identifiable student information into any of these tools.

I also want you to be wary about posting any non-public documents that your school or university may have. At the moment, we don't necessarily know what they're doing with our prompts or anything we input. And we don't want some proprietary information or student privacy information to be somehow intertwined with public information. So, I'm going to say it again, when using these, tools hold back on using any identifiable student information, and hold back any non-public school or university information. Don't do it yet until they have that privacy aspect figured out for the time being.

All right. So knowing that, we're going to take a look at Bard, ChatGPT, and Claude right here. So, this is Bard. Now, depending on what Google account you're using, if you're using a school-based Google account, Bard may be blocked. So, if you want to try Bard out, I always recommend you use your personal Gmail, because it is not blocked in that case.

So, here is Bard. I like to use Bard for research and planning. In fact, Bard is truly changing the way search works. For example, let's say that I want to know the difference between whales and dolphins. Please help me understand the difference between whales and dolphins. All right, here we go. Let's take a look and see what Bard can do.

Now, this is a very simplistic prompt. I can dive deeper into this if I wanted to, but here it is. It has whales, it shows the difference in size and diet and social behavior. Some other differences. But again, I want you to understand, it is crafting a brand-new response from information that it has in its data set.

We need to make sure our students understand that some of this stuff can be hallucinations, meaning, some of this stuff can be completely made up. Google recognized this. And so they put a really nifty button down here at the bottom that actually double-checks its work based off the internet. So if I click this button, it's taking its entire creation, evaluating it against the internet, and it prepares these results right here. Anything green means it found a corroborating website that also says that information is true.

Now, I did this presentation to a school yesterday and they're like, well, just because it's on the internet doesn't mean it's true. That's true. That's why you always want to make sure the websites that it's quoting is a reliable website. And usually, it does when it does this double-check, like Noah, and it does the Marine Mammal Center. And so, it tries to actually go to some reliable sources, but you definitely want your students to double-check that.

But notice, this one is in like peach, and this means that it either couldn't verify this information, or it found opposing information to this. And it tells you right here, it says, the smallest whale, the dwarf sperm whale, is about 10 feet long and weighs about 600 pounds. Here, it's saying, the dwarf sperm whale at its smallest is around 9 feet in length and 500 pounds. So it's different. So, you might want to do a little bit more research on the accuracy of that.

I really appreciate that Bard did that with that. And you might want to play around with it a little bit more. And I'm going to show you my personal one really quick for Bard right here, because-- so here's my personal Gmail. And in personal Gmails, you actually have these extensions up here that you can now utilize. So, Google Flights, Google Hotel, Google Maps. If you wanted to plan a trip, like a backpacking trip across Europe, you can utilize this to help you find the best flight, the best hotels, and the best routes from where you want to go. OK?

It also attaches to your Google Workspace if you want. So you can ask questions about your Google Sheets, your Google documents, anything like that. So it's really starting to become ingrained into the entire Google ecosystem. But for now, it's only your personal ecosystem. And Audrey, you might want to just try a few more times to see what it does. It doesn't always work the best. But it's been getting better. And it also can give you responses in a data sheet, like an Excel-- well, it's Google, so it's going to give it to you in a Google Sheet. But it will turn it into tables for you, which is pretty neat. And you can ask that. You can ask that.

All right. So, that is Google Bard. I like it for, again, I like it for research, internet searches, and planning. That is what I like this for. ChatGPT. Notice, it's nearly the same. You have your new chat off to the left, your past chats off to the left, and your prompt entry down on the bottom. ChatGPT is the same way. You have your new chats off to the left, your past chats off to the left, and then you have your typing and your prompt here at the bottom.

Mine might look a little bit different than yours, because I added in an extension called Prompt Genius which is basically a prompt clipboard for me. So some of my favorite prompts, instead of having to retype them, I can just click them, and they'll appear down here at the bottom. It just makes things a little bit quicker for me. If you're interested in that, I have that extension right over here for you to use. It's just called Prompt Genius underneath Generative Websites on that Padlet.

All right. I love to use ChatGPT for linguistic writings. So creative writing, essay writing, article writing. Whenever I want it to sound like a human has done it, I utilize ChatGPT. I find it does a really good job at writing. But it's going to be up to you, because I've had some people say that they like Claude much better for that humanistic writing. So, what I recommend doing after today, when you want to play around with this a little bit more, is paste the prompt into ChatGPT, see what it replies with, paste that same prompt into Claude, see what it replies with, and then paste the same prompt in Bard, see what it replies with. Find which one resonates with you more. OK?

So, a lot of people say, oh, I like ChatGPT the best. Well, everybody's different. So you may have a different view of that. There is a pay version ChatGPT 4, and with pay versions, you do get a lot more features. So you have the ability to click on Browse with Bing. By clicking on Browse with Bing, it now brings in more to the data set and it brings it up-to-date. So it's no longer limited to September of 2021; it is now available to up-to-date information.

Advanced data analysis. This allows you to now have a plus button in the bottom. And it allows you to bring in PDFs, images, and ask questions, documents, and ask questions about those things. Plug-ins allows you to bring in externals into this. So if you're trying to teach students about math, you might want to bring in Wolfram Alpha. You now have the power of that math processor of Wolfram Alpha embedded now into your ChatGPT to answer questions.

And then finally, the newest one is the ability to do image generation right out of ChatGPT. In fact, I have a seven-year-old who wanted to start a book podcast, and we were able to come in here and craft her logo just by prompting and asking it to do various things. So it does a really good job with those things as well. Again, this is ChatGPT. We're going to just use 3.5 today because-- Oh, Audrey, there's a Google-- sorry, there's a plug-in for Canva in OpenAI. So, you have that all right there.

All right. So, I'm going to be using ChatGPT today to model all the different prompts, but that's because I have all the prompts built in with my Prompt Genius. You use the tool that you want to try first. OK? You use the tool that you want to try first.

Lastly is Claude. Now, Claude is a little bit different. Claude is actually-- and this may sound weird, but Claude does have some backing from Google. They've put some millions of dollars into it, but most recently, they just got largely backed by Amazon. Amazon just put in $4 billion into Anthropic, which is Claude's parent company. So, you're going to start seeing Amazon getting into this game a little bit more. I wonder if they're going to start bringing this into the Alexa platform to start actually having better answers to those questions.

Anthropic is different because they call themselves the ethical AI. They have a mission statement called constitutional AI that promises that they are definitely going through their data sets and removing harmful or hurtful information. They're removing biased, sexist, and racist information. They're trying to create the cleanest data set they possibly can. So, if you want to learn more from them, all you have to do is go to the Anthropic website and read about that a little bit more.

But what I like about Claude is that out of the box, without having to pay a dime, you can upload PDFs, text files, or CSV files, which are like Excel or sheet files when you download them as a .csv. So you can do some data analysis for free right out of Claude. I want to show you this really quick. There's a lot of teachers that utilize The Outsiders novel to talk about relationships and how do you deal with people you don't think are doing the right thing and different members of society.

But let's say that-- I'm sorry. Let's say that you wanted to craft some lessons based on The Outsiders book but you're starting from scratch. You can come in here and ask it to do that. After you've uploaded the PDF, you can say something as simple as based on the book the outsiders, I just gave you please create a lesson plan that is three weeks long, covers the book chapter by chapter, please include collaborative activities and public speaking components, and it will craft week by week, day by day, different things that you can do per chapter. And of course, this is just a quick prompt, but it shows you the very basic beginning of how you can utilize these tools to give you ideas or examples of where to go with some of your lesson planning.

So those are the three tools that we're looking at today. I want you to choose your favorite right now. OK? And we're going to do something that goes along with how these tools are being utilized today. So I just showed you how it's replacing internet search. It's also replacing how people write code. These tools are not only helping people complete the code they are writing, but people who have no idea how to code, you can say what you want, and it will give you the code for what you're asking for. A lot of different things that are happening right now with these tools.

But we're going to focus right here on children's books, because number one, it's a little bit low-hanging fruit, but it gives us some pretty big bang for our buck when looking how to use these prompts. OK? So, we're going to base our activity right now off of this gentleman, Amar. You see, Amar saw the power of ChatGPT when it first came out. And so in December he utilized ChatGPT to write a children's book about a girl who befriends a robot, and they go on adventures together. He then took the output, the story, from ChatGPT, and went into an image generator called Midjourney to create all the images.

So in under 72 hours, he not only created and illustrated a children's book, but he got it published in Kindle's Amazon digital bookstore and was able to order a paperback version sent to his house. In under 72 hours. Completely transforms the way that we think about publishing, right? But the question is, how do you do it? Well, I was inspired by Amar's vision, and I said, I'm going to write a story, or I'm going to have the ChatGPT craft a story for my girls. I have a seven-year-old girl and a four-year-old girl. And we go through all of our books. They love to read, and I love to read them stories. I wanted to create kind of an original story, kind of.

And so there was always a story I wanted to hear written, and this story was based off the movie, Elf. And in this movie, Elf, if you've never seen it, an older gentleman believes he's an elf because he was raised in the North Pole, he goes to New York City to meet his biological father, and his father ends up being the manager of a children's book publishing company. That's just a little basics of the background of the movie.

But in the movie, you had some of the creative directors who can't come up with a new story. And one of them says, well, what about a tribe of asparagus children who are ashamed about the way their pee smells? I thought that was hilarious. I'm like, I would love to hear a book about this. And so I went into ChatGPT and I said, I want you to write me a story, a children's book story, about a farm. And on this farm live children vegetables. One of the children is an asparagus child, and he's always constantly being made fun of by the other vegetables about the way his pee smells. And he goes on a journey of discovery, and is inspired that maybe his weakness is his strength, and he goes back to discover his farm is being overrun by insects and he uses his pee to scare the insects away, because the smell drives insects away.

And it actually wrote a very good story. So if you ever want to see it, it's in the slide deck it's called Andy Asparagus: Urine Trouble. My daughters thought it was hilarious. And I thought, wow, in just a short amount of time, I was able to ask this machine to write me a story, and it did, and it did it at about an 80% success rate. 80%. Of course, I went in and changed things up. I thought maybe some things should happen, and I added my own flair to it. But it was 80% done. The framework was there. The train of thought was there. And it started with an idea.

And that's how I started to learn to prompt. The more detailed I am about what I want this thing to do, the better the children's book that it writes. And so, since this was the way that I first dipped my toe into prompt engineering, since this is the way that I was first inspired to learn more about prompt engineering, this is what you are all going to do right now. But you're only going to get five minutes. OK? We're going to take five minutes here, and you're going to write a children's book with one of your favorite tools. I recommend ChatGPT, but if you're a Bard fan, go use Bard. If you're a Claude fan, go use Claude. Heck, it's a prompt, so copy and paste the prompt into all three, and you decide which one is better. OK?

But this is what you're going to do: you are going to start your prompt engineering journey by thinking about a child in your life. It could be a son, a daughter. It could be a niece, a nephew. It could be a past student. It could be a grandchild or a grandson. It could be a friend's kid. Any child that you could think about in your life. And you're going to think about something that makes that child special, something that makes that child unique. Maybe it's a special skill. Maybe it is a special talent. Maybe the child you're thinking of has something they have to overcome. Maybe they have a learning disability. Maybe they have a physical disability. Maybe it is a student in your life or a child in your life that is-- that English is their second language, and they get made fun of because they don't have a quite grasp of the English language. Kids can be cruel.

Something that you can pin your hat on to be the reason why this child deserves a story. That is your choice. And then you are going to write your prompt. And you're going to include three things. This is your very first introduction into prompt engineering. OK? You are going to include a task or an ask. That is, what do you want this machine to do? So, your task is you are going to write me a children's story. That's the ask. It doesn't have to be any more detailed than that. That's the ask: you are going to write me a children's book. You are going to write me a children's story.

Then what you're going to do is add in the context. This is the entire meat of the prompting. This is what makes a prompt a great prompt, is this context. This is where you're going to put all the information that the tool is going to use to write the story. What is the name of the child? You don't have to use the same name of the child you're thinking about. If you are, first name is fine. What is the name of the child? How old is the child? What does the child look like? What is their hair color? What is their eye color? How tall are they? Where do they live? What are their favorite activities to do? Everything you can possibly imagine, all that information, is going to be important. What season is this in? Where is the story taking place, where they live? All of these things are very important.