So, good afternoon, everyone. My name is Doree Carter, and I'm excited to, talk to you a little bit today about, AI and, making it funnier.And so I think this will be a, you know, a a fun topic to discuss, you know, something where it's very interesting just in the world of AI and all the different tools and things that are out there. I think learning to be more creative, with those tools that are out there is something that can be a benefit to really anyone in in any field. So I'm excited to, get started. Alright. So to start out, I asked, ChatGPT specifically as as one of my tools of choice to tell me a joke about data science, and this is the first one that it gave me. And it said, why did the data scientists bring a ladder to work? Because the models were in the cloud, which I thought was was kind of clever.

And then I ended up asking it, to tell me that joke again with a little bit more, context. So, you know, why did the data scientists pull up with a whole ladder? Because their models were literally vibing in the cloud, which was I thought was kind of funny. I was like, okay. You know, it's it's trying something a little different here. And then I asked it again, with a little bit more, you know, information, and it gave me this sort of poem. I grabbed my ladder and took a stand because my models live in Cloudland. They don't text back. They just compute. Predicting trends in a glittery suit.

I was like, okay. It's definitely getting more creative. So when you're looking at these, you know, three jokes that have been created here, I'm curious just, you know, what comes to mind of which joke do you like best? So, you know, everyone has their different styles of comedy that's funny to them, things they enjoy, things that they think are clever. And so it's kinda like, you know, which one do you like best? Now from that, which one do you think was the most creative? So when we think about how you can kind of play around with different jokes and versions of comedy and things like that, you know, which one of these do you think might have taken the most effort or maybe the most prompting? So just right off the bat, I'll tell you the very first example I gave. I just asked it to tell me a joke about data science.

The second one, I asked it to tell me the same joke, but with Gen z slang. And the third one, I asked it to tell me that same joke, but in the style of Taylor Swift and also make sure it has a rhyme in it somewhere. So, again, as you add these different kinds of context, you can kinda get creative with it. So when we think about generative AI, I think that it's definitely gonna be, you know, a a revolution in the creativity space. But what about the playfulness side of things? Like, the funny side of things. Right? I think it absolutely can be funny. I think the AI tools can get better at humor, but part of it is prompting the right way.

We talk a little bit about, prompt engineering and the importance of asking good questions, giving good descriptions. And so this is one of those spaces where if you're trying to get a certain kind of effect from your audience, right, being able to ask those right questions, I think, is something that's really important. So I think that the, being funny is a reciprocal type of thing when working with, AI. So you've got you working with the tool, asking it some funny things, and it learns from that and then is able to kind of help you with your creativity in the process as well. So just to talk just briefly, about my background. So currently, I am a senior manager of data science at John Deere. My team focuses on personalized experiences for our digital solutions and software.

I previously was also a manager at John Deere focused on advanced sensing technologies and also spent some time at Procter and Gamble as a site director for their smart lab, doing, operations and logistics for different research. And so, you know, through my experiences, you know, looking at working with a wide variety of of folks in data science and analytics fields. I got really interested in trying out new technologies, always interested in learning new ones that are out there. And my background is in crop sciences and informatics, and so I didn't start in the tech and data science space. I started in agriculture. And with that, I found that, you know, getting up to speed with the different kinds of technology that were out there at the time that I was going to school, was gonna be something that was gonna be really important for the future. And so I've kind of continued that, learning path of what are the newest tools out there, and what are the different ways we can use, some of those types of technologies as well.

So in today's discussion, I wanna make sure I cover three main things, how generative AI works, different types of humor, and how to create your own funny prompts. So the first section here is just how generative AI works. So when we think about, you know, what exactly generative AI is, so it's a system of algorithms or computer processes that can create novel output and text, images, or other media based on user prompts. So the key point of the generative space is novel, the newness, like creating something that maybe we haven't seen before, hence, the the generative part of things. Right? And there's a lot of different tools out there. So some of these, you may be familiar with, especially in the chatbot space with text generation, chat GPT, Claude, Microsoft Copilot.

And then you also have image generation, with tools like Dolly and Midjourney and Adobe Firefly. And then we also have audio generation and video generation. Right? And so if you haven't, you know, tested out some of these tools before, I highly recommend it. Go around, play with it, understand, just get a feel for the different voices as well. Especially in the text generation tools, I find that they all have a little bit different voices in kind of how they feel, how they respond, to what your inputs are. So it's you know, definitely recommend looking into some of these tools to get a feel for their differences. So when we think about the differences between traditional AI and generative AI, you know, AI in general is this, you know, ability for computer to learn and make decisions without intervention from a human.

It can do things by itself. We often see terminology like machine learning and deep learning, you know, associated with the AI space. But when it comes to generative AI, it's really about producing original content based on its inputs and creating those new kinds of media, new pieces of text. And so we often see terminology like our generative adversarial networks, variational autoencoders, diffusion models, large language models, transformers. Now I you know, in order to make sure we kind of, you know, stay on track here and don't get too derailed with a lot of the technical language, I wanna just focus a little bit more on the large language model piece and the transformer piece, which we'll talk about next. But I do encourage for folks who maybe haven't seen some of these, you know, some of these types of terminology before to look them up on your own and kind of read through how they're used in the different generative AI tools. But for today, we're focused specifically on the text side of things. So I'll talk about the large language models and transformers.

So at a very high level, how does AI generate text? So I kind of simplified it into kinda three main buckets, specifically around the text generation. So the first bucket is the text input by the user. So you ask, for example, chat g p t. I use that just because it's a very common tool, you ask ChatGPT a question, and the text is fed into its own model. So in this case, ChatGPT has a generative pre trained transformer, which is GPT. And what that means is you have the generative part, the g, so the ability for it to create something new. Then you have the pre trained portion, which is the p, and that's really its large language model.

So all of this text that it's, you know, collected across the Internet in various places is in its own model that, that's been pretrained. So it's already ready to go. Right? And then the transformer is a type of architecture that literally transforms that text into the format that it needs to be for the model to understand it and provide you an answer. So you have the generative pretrained transformer. Now the second part here is that ChatGPT then turns the text into what we call tokens or embeddings, essentially, and they're vectors, which are numbers that have a position and a direction. So So when you think about a sentence and where the different words are in a sentence, all of those different words have a position and they have, kind of, like, a weight or a number associated with them.

So this is telling the computer kind of what do these things mean in the computer terms. And that's the encoding process. And then lastly, you have the decoding process. ChattGPT then uses this to predict the most accurate text output and provide you an answer to your question. So it has all of these vectors, goes into its model, and goes, what are the things that are really closely related to this that makes sense? It understands context to a certain degree and then says, you know what? We're gonna generate this new answer based on what you have put in. When we think about why it might be hard for generative AI to to, you know, generate humor or work with humor, there's this article that's interesting here that talks about why being funny is AI's toughest test. And there's a nuance to the communication side of things when it comes to comedy and telling jokes and kind of that playfulness.

And there was also a research article that came out a couple years ago that talked, about how, large language models and the humor side of things is still a challenge. And they found that over 90% of a thousand and eight generated jokes that that, that showed up were actually the same 25 jokes over and over. And I've actually found this out myself as as well playing around with it. And that leads me to, the another quote from from the article here that says, so maybe we just need to ask better questions to get better comedy. So with that, I wanna talk about different types of humor that are out there, because this is gonna be an important piece of the prompting space. So the first one is called absurd or this is something that's extremely silly or ridiculous. And I have this funny example here.

If unicorns ran the government, traffic jams would be solved by rainbow powered teleportation, but only on Tuesday. So this is something that's just very wild, very out there, funny type of of humor. Then we also have cringe. So this is something that is embarrassing. It's kinda awkward. It kind of physically makes you recoil, when you hear some of these things. I'm so awkward. Even my shadow tries to avoid me. Right? Statements kind of in that, area. And then we also have irony. So, you know, I I I like this example of posting on social media about how useless social media is. You know, very, very ironic. It's it's always a a a fun one there. You've probably been exposed to a lot of different irony statements, over the years. We also have non sequitur.

So this is a statement that does not follow logically from or is not clearly related to anything you said before. I like my turtles, but my coffee mug just told me it prefers jazz. Those statements do not go together whatsoever. And so non sequitur is actually a really fun term. I like to feed AI a lot because you get some very creative outputs for what it thinks. You know, it has to logically find the next statement, but it has to be an illogical statement, which is fun. There's also the classic pun. So I like this one. Time flies like an arrow and fruit flies like a banana. So that was a that's a really fun one. I'm sure you've been exposed to a lot of puns as well. Those are always a a favorite of mine. And then we also have sarcasm.

So this is, a mode of satirical wit depending for its effect on bitter, caustic, and often ironic language. You know, sure. I'd love to clean the kitchen, said no one ever. So I I thought that was a a fun example there. But sarcasm also, is very interesting to see how the computer, interprets that. So we'll do a rapid fire quiz here of what type of humor is it. So the first one here we have, I tried to catch some fog. I missed. That one is a pun. So that's that's a fun one. Alright. Next one, I poured orange juice in my shoe this morning. Now I can't trust Mondays or parrots. That one is non sequitur. Okay. Our last example here, we have a sign at the gym said lose weight here, but the vending machine was right next to it. Yes. That one is irony.

So just some examples. Right? So we kind of, you know, do a recap of the kinds of humor. Alright. So let's get into how to make your own funny prompts. So three step, funny prompt writing framework. We have intent, we have tone and setup, and we have constraints or curve balls. So the first one is around the intent. So what's your funny goal? What kind of humor do you want? So this is gonna define what the AI is going to make, whether it's, you know, a joke in general or if you want a pun, if you want a parody, if you want a poem, if you want a song. Right? You have to figure out what it is you want. So, for example, we're gonna make a dad joke about coffee. The next step is tone and setup.

How do you want it to sound? So you wanna add any type of context. Do you want to make it dry? Do you wanna make it silly? Should it be sarcastic? Should it be absurd? So if we're sticking with the the dad joke example, you know, we want a cheesy dad joke style, you know, a classic groaner joke, and we wanna make sure coffee is the main subject. Then last but not least, we've got our constraints or curve balls. So we wanna add quirky or tricky rules that really pushes the limits on the AI creativity. We want it to be clever. We want it to be unexpected. So we're gonna say we want our coffee dad joke, but we wanna make it rhyme, and we also want it to have a pun about caffeine.

So if we put this all together, we're gonna ask ChatChibiti in this example, write a cheesy, rhyming dad joke about coffee that includes a pun on caffeine. Think classic groaner style. And the output is, why did the coffee file a complaint so keen? Because it felt it was grounds for less caffeine. So we had, you know, the dad joke style. It rhymes. It includes a pun. Right? So figuring out kind of how to put those things all together. So I'm gonna show another example, in a, just, you know, kind of to to show you a little bit more of that progression of what it looks like when you add more context to your prompt. So we have I wanted a funny data science username. And I just put this in there and said this, no other context, and they gave me SQLU later. I was like, okay. That's that's fairly clever, but we're we're gonna take it up a notch. I want a funny data science username that is sarcastic. Totally not overfitting.

Okay. It's kind of, you know, it's it's getting there a little bit. We're gonna add a little bit more. I want a funny data science username that is sarcastic and sounds like Shakespeare. Okay? We're really we're really thrown at a curve ball in this one. And it gave me outlier doth protest. I was like, okay. I you know, I think we've we got some some good usernames to pick from here. But you can see that progression of as you add more context, it really, you know, tries to figure out, okay, how how how am I fitting that into that context? So why I think, you know, using generative AI, you know, to to make funny statements is important. So it really helps adjust your AI's brain. As you use it more, it gets used to your voice and how you ask questions, and it'll help it be more creative, but it also helps you be more creative.

And I've actually used this to help me figure out better business presentation titles or more relatable metaphors for technical topics that might be kind of hard to explain, but they need something a little more playful, a little more lighthearted so that folks can, you know, understand it a little easier.

So that's something where I find it very helpful, for me. So the three takeaways I wanna end on here, you learned how Gen AI works. It's not magic, but it definitely feels like it. You learned about different types of humor. There are many types of humor out there, so use a range. I only showed you a few, but there's lots out there. And you learned how to create your own prompts. So we have the three step prompt writing framework that'll turn the funny prompt prompts from dull to great. So I'll leave you with one last, funny generative AI statement around the around the conference. So we're serving up sizzling solutions at women in tech where GenAI is the secret sauce and innovation is always on the menu. And that is it for me.

Thank you all for listening in. I hope you, enjoyed the presentation and and learned something new. If you have any questions, please feel free to reach out, and I put a QR code, on the screen so you can scan that, and you can find all of my, contact information there. So thank you.