AI detectors, amazing slides with Beautiful AI and Gemini as an AI gateway

Welcome to another episode of AI Goes to College where we help higher ed professionals navigate the world of Generative AI. I'm Dr. Craig Van Slyke and I am joined by my friend and co host, Dr. Robert E. Crosler.

And so we've got three things we want to talk to you about today. The first one is AI detectors. So, Rob, let me tell you what my general statement is about AI detectors.

And I think I'm willing to soften this a little bit. They're garbage. Don't use them. So what do you think? Is that too far?

100%. I 100% agree with that. We actually at WSU have been told not to use them.

They are very hesitant to punish students when the detection tools are as imperfect as they are. The biases baked into them are problematic. And it really is a terrible situation when you are that student who is innocent.

But this tool that someone claims is infallible calls you out on using these generative AI technologies.

So you said they're biased. How are they biased?

Well, a lot of the bias I've seen is their bias towards non native English speakers. People who learn to write in a very systematic way oftentimes are falsely flagged as having used these tools to help with their writing.

The other way that I think they're biased is in the ill defined nature of what is a generative AI tool. And the one example I saw where someone was getting in trouble is they used Grammarly. And Grammarly at one point was okay.

But now Grammarly is baked in some of this generative AI technologies. And the way it's helping people to write better is landing some students into hot water.

And in a world that's as muddy as it is now with, you know, what is okay to get help, to not get help, what is an AI tool?

What isn't an AI tool, It just creates this environment that is ugly if you're using tools and accusing someone of cheating when even if they did, maybe they didn't even know that was a generative AI tool that was helping them.

Right, right. Yeah. And that sounds odd that maybe they didn't know.

But there are like, even with plagiarism, which we've kind of understood for a long time, there are like 15 different, I'm making up a number, but there are like 15 different types of plagiarism.

Right.

And the thing I've run into with plagiarism, and I think it's an interesting parallel, is I've worked with a lot of students that when they plagiarized they didn't realize they were plagiari, you know, committing plagiarism. And part of that whole process was educating them about what is and what isn't plagiarism, similar to using generative AI tools.

Well, what is a generative AI tool? And then what is and what isn't an acceptable way to use that to prove that a person didn't use it is impossible almost.

Yeah. And I don't want to be an AI cop either.

You bring up a really important and interesting point about what is and what isn't AI as AI gets embedded in more and more and more tools, you know, at some point is just clicking on a little blue squiggly underline in word going to get you flagged. And so I think it's a wrong headed approach in a lot of ways.

I don't want any of our listeners to take this the wrong way, but if you just take the output of some generative AI detector as being the truth, I think you're being a little bit lazy. I know we've all got more students in our classes than we used to have and there's, you know, we're facing a lot of stressors like that.

But I don't know, just take this tool and it says oh, this is AI and just, I'm not sure that's a good idea.

And what I would encourage people to do is instead of looking for a reason to say, oh, it's generative AI. Generative AI is bad. Look at the writing.

And oftentimes you can tell with the writing that they didn't answer the question properly or you know, it might be written beautifully.

But if you dig deeper at what they're saying, the arguments they're making, if it's not consistent with the guidelines, the directions, the conversations you've had in class, the there's ways you can anchor to some things that get it at. Did the person know what they needed to know in answering this question?

And to be honest, if it's written nicely because you use generative AI to help you write nicely, that's taken all the pain out of it for me. Because the reason why I hate grading written assignments is you have to wade through some pretty terrible writing.

So we can find that line between using a helper to make writing sound good versus the tools did the work for them on creation. I'm happy with that.

Just yesterday I was meeting with my doctoral students about their exam and they said, well, do you want these handwritten? And I thought, oh good God, no. I don't want to read my own handwriting. I certainly don't want to try to read yours. I'm with you there.

But really, you hit on the key, I think, to all of this, and what I would recommend, and I guessing you would agree with me, is you have to do something with your assignments. And we talked about this a few episodes ago.

But if you can somehow contextualize your assignment to the class, you know, relate this to class discussions or the textbook or whatever, you're going to be better off.

And then if it's beautifully written but it doesn't answer the question, give them a bad grade, not because you think it's AI generated, but because they didn't answer the question. I think that's the approach that I would take.

Yeah, absolutely. I used to have 25% of the rubric for a written assignment about proper use of grammar and to encourage good writing.

And now I found I've either eliminated that from my rubric or drastically reduced how many points you get for writing well, because now with the tools available, everybody should be able to write well.

Yeah, like you said, that's not the worst thing in the world. So I ran a couple of tests. So early on, I ran some tests with the AI detectors, and the results were terrible.

There was something that I wrote a number of years ago that it flagged as like, 65. I don't remember the numbers, but like 65% likely it was AI. And then I wrote something with AI and it came back at like 25%.

So in preparation for this episode, I ran some other things that I wrote through generative AI and two of them were pretty low down in the single digits, which was good. One of them came back 25% AI, and it flagged a bunch of things as being likely. They don't say it is AI. They say it's likely AI.

I'm not sure where it got that from. You know, how it figured that was AI. I mean, it looks at kind of sentence variation. So when we talk, we sometimes have.

Or when we write naturally, we sometimes have long sentences, sometimes short sentences, repetitive word and phrase use. Those are the kinds of things that it looks at.

But if we go back to the bias, those are also some of the things that non native English speakers often fall back on. So they learn a way to say something, and that's what they put in their writing over and over and over again.

So I think it is fundamentally unfair, and I would not want to have to defend it. Yeah. If I don't have more to go on than just an AI detector. So I think our big message is maybe use them as a data point if you want to.

I mean, maybe it cuts your 40 essays that you need to kind of double check down to six or eight. And I don't see anything wrong with that as long as you go in and make your own judgment.

But, God, I hope none of you are just, oh, you know, this is. My cutoff is 48% and you're 52%, so I'm going to give you a zero. It's a bad idea.

Agreed. Agreed. And I would encourage people to think about active learning.

Craig, I think moving away from having you write something and then me as the instructor, being able to grade that and see what, you know, create ways of students being able to demonstrate their learning and to do so in an active fashion in the classroom and maybe multiple ways.

I mean, that's always been a good idea.

Triangulate.

Yep.

I think, again, we might be being a little bit repetitive here on a prior episode, but Tom Stafford, one of my colleagues, when he teaches our undergraduate core class, he requires everybody to use generative AI and then either summarize or reinterpret or contextualize what AI came up with in their own words. And I thought that was brilliant, you know, so he's, hey, use it. Why wouldn't you use it? So use it. But don't just use it and say it's yours.

Use it for research. Use it to help you think through some things and then write up what you think about it. I thought that seems to be working out pretty well for him.

So, yeah, love it. Love it.

All right, so I think we've talked enough about AI detectors, except there was an interesting article that came out in Bloomberg.

Craig, let me ask you a question about our next topic. So you sent me a link to something called Beautiful AI and it blew my mind. Can you tell me how you found that and what it can do?

So, yeah, so Beautiful AI. It's. Beautiful AI is a slide deck creator. It does a couple of things.

So it's maybe the most aptly named piece of software I've seen in the last year or so because the slides are spectacular. They're unlike anything you can do in PowerPoint or Keynote. They just really. They stand out. I've used.

I've used it a lot for a lot of talks, and everybody is always blown away by just how beautiful the slides are.

So is it another piece of software, or does it actually create PowerPoint slides for you?

Well, so it'll do both. The way I normally use it is I use it kind of like you would use PowerPoint. It's browser based.

So I go into Chrome and pull it up and they've got some templates and that sort of thing. But I use it kind of like I would use PowerPoint.

You can download, export the slides and make them editable in PowerPoint, you've got to, sometimes you've got to download some fonts to go along with it and do a little bit of tweaking, but it's not much. You can also just download the slides as images, which you don't have to change anything on those, but you can't edit them.

The good news is you don't have to edit them. The bad news is you can't edit them.

So when I've played with some of these image creating software tools, sometimes they put words in them and they dispel them completely wrong, wrong. And you're like, wow, how is that even a word?

Does it get the words right and the spelling right and all those things that you would hope it would get?

That's a good question. So this is not an image generator, it's a slide generator.

And so the images are stock images and they'll even tell you sometimes where they came from. So it's just like using, you know, going out to Shutterstock or something like that and finding the images.

So it does not use that I know of any AI generated images. But where it does use AI is you can put in a prompt and it will crank out a set of slides for you.

Also, it actually does research for me to say if I want to do a presentation on why dogs are the greatest pet people can own, it would make those arguments. It would give that information with pictures of dogs. And here we go.

You could, you could go in and literally type into the prompt. You know, it's got a little prompt window up at the top. Presentation on why dogs are awesome. And it'll do that.

And I just to test this out, I was working on a talk I'm going to give at a Wright State University on Tuesday with our friend Rich Johnson. And it's a fairly. I'm not going to mention the topic.

It's a pretty esoteric topic that I was connecting into Human Resources management, which is what the symposium is about. And I just wanted to see how it would do. I was shocked. And those are the slides that I sent to you, Rob.

And I was just blown away by how it made these connections because I'm taking Two things that have literally, to my knowledge, never been put together before. So it's not like it's going out and finding something on some Wikipedia website or whatever and just creating slides.

I mean, it had to make synthesizing connections across these topics. And it did. And I'm not going to use the slides. I might use some of the slides, but it.

If I had to guess, I would say it may have cut my work in half because there are some slides that are pretty good and with a little bit of editing that will kind of fit my. I've got my own style with these slides, one of which is I often have pictures of my animals, dogs and cats, equal time.

What about the goats?

Yeah, I do still throw in the goats and the horses. And so, yeah, people always love those. And so I can't really just use them right out of the box.

But if my dean came to me and said, hey, I need you to give a talk on X in an hour, I could do it and people would think the slides were awesome. So if you haven't checked it out, it's beautiful AI. I think they have a free tier and really the slides are just spectacular. Yeah.

What I love about this, Craig, is I've used generative AI tools to help brainstorm talks I'm going to give. And I've used generative AI tools to take and convert maybe something I have in word into PowerPoint.

And I've got a bunch of different tools that I've leaned on to help me do some things. And it sounds like what this tool does is it does it all in one stop shopping and can not have to have you jump around to different places.

And it sounds like you're getting a much better slide deck than I get when I just have Copilot transfer stuff into PowerPoint for me.

Yeah, I mean, the copilot slides are always pretty ugly, I think. I've never really liked them. But if I mentioned that the beautiful AI slides are spectacular. I think I mentioned that before.

I think you did say that.

In fact, I'll create a presentation just based on the prompt that I put in. And you can make these public. So I'll make it public and put a link in the show notes. But what it cannot do yet is if you give it a document.

What I want it to do eventually is give it a document and then it will create a slide deck out from that document, which it doesn't do right now, but probably will at some point if you want to stand out in your presentations. Beautiful AI is The way to go. And if you want to get a head start on your presentations through AI, Beautiful AI is the way to go. So I.

And we don't make any money off of this, by the way, so I'm not. We're not pimping this. But really, it goes back to something I wrote about in the newsletter a week or so ago.

Forget about AI having to do everything for you. If it can cut your work in half, that's pretty awesome. Did we talk about that last time?

It seems like we talk about that almost every time, Craig. I think becoming more efficient and cutting down on my workload seems to be the trend of where we're going with AI now.

Hopefully I don't find ways to replace the amount of work I do, because then all of a sudden, you're getting twice as much of me for same price.

One of the things I really like about the current environment is if you. As long as you don't get carried away. There are all kinds of interesting tools. Like I mentioned Lex Page last week, which I love. Beautiful AI.

There's one. I'm kind of throwing this on you, Rob. We didn't talk about this in advance, but called Ido Ideal. No, that's not right. Ideogram.

Ideogram might be how you pronounce it. It actually gets text. Right. It's really pretty good. I've used it a few times playing around with different images for the podcast. So it's.

It actually gets the text right 100% of the time, at least when I've tried it. Cool.

I'm gonna have to play with that one.

Yeah, and they've got a free. You know, you get. Takes like 10 credits to create images, and you get like 40 credits a day for free or something like that.

But maybe after you've looked at it, we can talk about it. So. But. But those kind of tools, you know, they're out there. They have the free tiers. You play with them. If it works for you, great.

If it doesn't, well, you know, move on to the next one. It's fine to experiment with these things. You don't have to marry them.

So, anyway, I'm really looking forward to you trying out Beautiful AI and seeing what you think about it and seeing if it's. It moves you off a copilot a little bit.

Yeah. Look forward to playing with it.

One caveat. Because they use really high quality images, the file sizes can be pretty big.

You know, most of the time that doesn't matter, but they can be pretty Big. All right, ready for our last topic of the day? This is something I've been thinking about, and I'm going to call out my wife Tracy.

And in a good way, Tracy, poor thing, has to put up with me babbling about this AI stuff all the time. And she's very smart.

She's a retired nurse, has a master's degree from a just slightly less than Ivy League school, and she hears me talk about it all the time. So finally I figured out how to get her kind of more interested in it. Any guesses on how I did it?

You probably challenged her and told her she couldn't.

No, no. God, no. I'm not that brave. I introduced her to Gemini.

So anybody that uses the Internet at some point, you've had a question, you know, who wrote this song? Or what time is this place opened?

Or whatever, and you go into Google, and if it's something that's maybe a little bit more complicated, like how do I do something? Often you have to wade through multiple YouTube videos and websites and that kind of thing. And I think we've talked about this before, too.

So I use Gemini as kind of a substitute for Google search for a lot of things. And I know you and I have talked about that previously, but what hit me is that everybody knows how to do that.

I mean, I'm exaggerating, but most computer users know how to do that reasonably well. They can type. They're used to typing something into a little search area. They're used to getting the results. So it's all kind of comfortable.

They're also used to using Google. And I think this is an important point. You're not using something who is OpenAI, who is anthropic. What is perplexity? This is just Google.

It's called gemini, but it's gemini.google.com, just like it's mail.google.com or calendar.google.com so it's very familiar.

And all you do is type in, just like you would type into Google Search and you get these results, but the results are typically what you want on the first try. Not. Okay, I've got to go to the third YouTube video and fast forward six minutes in and do all of that kind of thing.

It just kind of gives you the answer.

And this is much better than the Google Search AI overview, whatever they call it, that you and I and just about everybody else hates a lot of the time. So it's not that it gives you very useful, very useful responses.

When I started thinking about this more it really plays well to the cost benefit calculus when we're trying to decide to use something new. So people know what a pain it is to find certain kinds of things through a Google or DuckDuckGo or whatever search, Bing, whatever you use.

So everybody knows that. So if I can show you a way to do that kind of thing a lot more quickly and easily, you're going to be interested. That's a pretty big benefit.

And what's the cost? There's almost no cost. You just, instead of google.com, you go gemini.google.com?

Well, one thing I've noticed, Craig, is whenever I go to a Google page anymore, the Gemini thing pops up periodically and says, why don't you try to do this using Gemini? So Google's almost pushing it on you, so that way you remember to use that instead.

Because I think one of the things that we do is we get into this habit of always doing the same sort of thing. And so the big trick is remembering, oh yeah, I need to go to Gemini.

So I would encourage people, when you look at your Google search and it says, hey, do you want to use Gemini? Say yes. You know, give it a shot.

Yeah, try it, try it. If it doesn't work for you, it doesn't work for you.

The message I wanted to get out there to our listeners is that you, if you're listening to this, you're probably very interested in AI and you're probably pretty far ahead of most other people when it comes to using AI. And that means that you have the opportunity to help others figure out how to use it effectively.

But if you get in front of them and oh, let me show you this thing and this thing and see, I prompt engineered this and see how I've got, it's going to be too much. They're going to look at that and say, wait, the cost of that for me, and the learning curve is just way too high.

And plus, you know, I don't write a lot of documents. I don't do a lot of these things. But just about everybody that uses a computer searches for stuff. And so it's really a high benefit.

It's a really, really a low cost. And I think it's a very gentle way to kind of be a generative AI gateway.

And then as they get more used to the system, you know, maybe they try writing a poem about why cats are awesome or whatever, and they start to expand their uses or they don't. Even if they don't, they're still getting Better results when they have these kinds of questions.

So that's my kind of message for those of you that are out there in the. In the digital realm. If you want to ease somebody into it, try Gemini. It's a lot easier to do so. All right.

Anything else you wanted to talk about today?

No, Craig, I think we touched on it all, and I think I love the point you made today about helping others get comfortable using AI. I still talk to people that are hugely in that fear space of what this is all about.

And anything that those of you out there can be doing to help people get past being afraid of this, strongly recommend you doing that.

All right. All right. And I'm going to spring something else on Rob. I want to close the episode with the dad joke of the episode. Are you ready?

The dad joke of the episode. Is this going to become a thing?

It might. We'll see. Or it might just get edited out and nobody will ever hear this. So what. What do you call a boomerang that doesn't come back?

I don't know, Craig.

What A stick this is. This is how I keep my 8 o'clock students at least somewhat awake. Yeah.

So when I taught a large intro to MIS class, I had my kids send me with a dad joke every day to class. So it was a research project for my kids, and I literally had one student who, as soon as I told my dad joke, would get up and leave the class.

And recognizing this, I started waiting until There were about five to 10 minutes left in class before I'd tell my joke. And that was my way of keeping this person engaged.

There you go. He must have liked them. So, yeah, it might be a thing. We'll have to see. This might be a thing. All right. You'll have to listen.

But just like you know your class, if you want to hear the dad joke, you got to listen to the end.

Perfect.

All right, that's it for this time.

I do want to remind everybody, though, that both Rob and I are available to help you or your institution out as you trying to navigate some of these changes that are brought on by generative AI. We're also very happy to go talk to different groups just in the next month. I know. Rob, you just did a keynote, right?

Yeah, I've read it. Actually, my keynote got canceled, unfortunately.

But I did just serve on Wednesday at a panel at WSU helping about over 200 people on faculty and staff at WSU learn more about AI and gain more confidence in using it in part of their jobs.

And I'M going to the other another wsu, Wright State University, on Monday to be part of their AI and HR symposium. And and then I'm going to the I think it's called ISCAP Ed Sig.

It's an IS Education conference and Pod Indy, if you're interested in podcasting, and then going to College of Charleston. So I've got a lot of travel, but we're we love doing this kind of stuff. Here's the big message.

If your faculty or your professional staff are stressed out about AI, what we're really good at is we can come in and unstress them and get them experimenting with it so they can use it to help them in their jobs. All right. Thanks, everybody. We will talk to you next time. Bye.