Perplexity.ai, a mini-rant, and a successful experiment

  In this episode, Craig has a mini-rant about misleading click-bait headlines, discusses two recent generative AI surveys, gives the rundown on Google's rebrand from Brard to Gemini and Perplexity.ai and shares a modest experiment in redesigning an assignment to prevent generative AI academic dishonesty (which is a fancy way to say cheating). More details are available at https://www.aigoestocollege.com/p/newsletter/, where you can subscribe to the AI Goes to College newsletter. Contact Craig at https://www.aigoestocollege.com/ or [email protected] --- Transcript --- Craig [00:00:10]: Welcome to episode number 2 of AI Goes to College, the podcast that helps higher ed professionals try to figure out what's going on with generative AI. I'm your host, doctor Craig Van Slyke. So this week, I give you a mini rant. It's not a full rant, but a mini rant about misleading headlines, Talk about Google's release of a new model and its big rebrand from Bard to Gemini. My favorite part is gonna be when I talk about dot AI, which is generating a lot of interest right now, and I think it's tailor made for higher ed, even though I don't think that they're restricting the audience to higher ed and some promising results from a little experiment I did in redesigning an assignment. I'm gonna hit the highlights in this episode of the podcast. But if you want the full details, go to AI goes to college.com And click on the newsletter link and subscribe to my newsletter. A lot more details, screenshots, that sort of thing there. Craig [00:01:09]: So here's my rant. Cengage, and if you're in higher ed, you know who Cengage is. They call themselves course material publishers, Just released its 2023 digital learning pulse survey. As far as I can tell, this is the 1st time the survey gathered data about AI. The results are pretty interesting. It says only 23% of faculty at 4 year schools thought that their institutions were prepared for AI related changes, and that number was only 16% for 2 year schools faculty at 2 year schools. 41% of faculty across the 2 different types of institutions thought that generative AI would bring considerable or massive amounts to change to their institutions. What bothers me about this survey, is really not the survey itself, But how it's being reported? So the headline of the article from which I kind of learned about this survey read, Survey reveals only 16% of faculty is ready for Gen AI in higher ed, which is not at all what the survey was about. Craig [00:02:22]: The survey, at least the part of it I'm talking about, asked 2 generative AI related questions. Do you think your institution is is prepared for AI related changes. And how much will AI tools change your institution over the next 5 years? So first of all, that really isn't specific to generative AI, although I think that's what most people would interpret, AI as. The title of the article that led me to the survey said that faculty aren't ready. Well, that's not what the survey asked about. It didn't ask if the faculty were ready, although that would have been a good thing to ask. It asked if they thought their institutions were ready. So I want to caution all of you to do something you already know you should be doing. Craig [00:03:09]: Read these click headlines, and there are a lot of them. Read the articles with a critical eye. If it's something that's important, if it's something that you're going to try to rely on To make any sort of a decision or to form your attitudes, take the time to look at the underlying data. Don't just look at how that particular author is putting the data. Look at the data yourself. All of that being said, I think we're probably not especially well prepared collectively for generative AI, And that's not a big surprise. It's still relatively new, and it's changing very rapidly. So we'll see. Craig [00:03:48]: Speaking of changes, Google Bard is now Google Gemini, and it's not just a rebrand. So Google also, as part of the rebrand, announced that they have some new models. So with Gemini, formerly Bard, which you can find at gemini.google.com. There are 2 versions at the moment, Gemini and Gemini advanced, and this is kind of the same as Chat GPT and Chat GPT Pro. The nomenclature is a little bit confusing. Gemini is a family of models. Ultra is the big dog high performance model. Pro is kind of the regular model, and Nano is a light version optimized for efficiency, which I think signals that Google is gonna make a push into AI on mobile devices. Craig [00:04:36]: I was pretty confused about the names and what models there were and that sort of thing. So I asked Gemini to explain it to me. The details of that conversation are in the newsletter, which is available at AI goes to college .com/newsletter. Gemini is kind of like GPT 3.5. It's fine for most things. If Gemini isn't up to the task, try Gemini advanced, which is kind of like GPT 4. So far, I've been pretty happy with, my use of Gemini advanced. It did a good job of helping me unravel the various names and models related to Gemini, And I've played with it for some course related tasks, and it's performed pretty well. Craig [00:05:17]: I'm not sure I'd give up ChatGPT Pro for Gemini advanced, but it's a nice option, and I'm playing around with all of them. So there's no big surprise. Know, your experiences may vary, but I would suggest that you try it out for yourself. If you do, I'd love to hear your impressions. You can send those to me at craig, That's [email protected]. There was another generative AI, education survey that was out in the news recently, a higher ed think tank, and I don't even know what that means. HEPI released a policy note then included some results from a survey of 1250 undergraduate students in the UK. According to that survey, 53% of students used GAI to help with their studies, but only 5% said that they were likely to use AI to cheat. Craig [00:06:13]: I don't doubt that the statistics accurately reflected the survey responses, but I'm pretty skeptical about both of these numbers. 53% seems pretty high, for students that have actually used generative AI to help them with their studies, And 5% seems pretty low for those who said that they might use AI to cheat, but I don't know. I think there are a lot more Students that are either using or going to use generative AI kind of at the edges of, ethical use. So I still think that the uptake of generative AI among students is lower than some of us might think or some people might think, even especially if we consider regular use. So they may have played around with it, but I just don't think many students are using it regularly yet. One of the reasons I like this particular little article, which is linked in the newsletter, is that it did discuss the digital divide problem. The digital divide is real, and it has real consequences for a lot of aspects of society, including higher ed. We need to keep chipping away at the digital divide if we truly wanna adjust society. Craig [00:07:27]: And generative AI is just going to widen The digital divide. More details in the newsletter, which you can can access at AI goes to college.com/newsletter. It feels like it ought to be a drinking game. How many times will they say AI Goes to College .com/newsletter? So let's get to the resource of the week. There's been a lot of online chatter about Perplexity dot AI. The gist of all of this talk is that Perplexity is becoming kind of a go to generative AI tool when you wanna uncover sources. There's a lot of hype that says this is going to Be the new Google, and I'm not so sure about that, but it is a very useful tool. Exactly what it is is a little unclear at first. Craig [00:08:16]: 1st, I'm gonna read you verbatim what the about page says. Perplexity was founded on the belief that searching for information should be a straightforward, efficient Variance free from the influence of advertising driven models. We exist because there's a clear demand for a platform that cuts through the noise of information overload delivering precise, user focused answers in an era where time is a premium. I couldn't argue with any of that, but I don't know what it means. Their overview page is a little bit better, And and it talks about some of the use cases for Perplexity Point AI, answering questions, exploring topics in-depth, Organizing your library and interacting with your data. I can personally attest that perplexity is Pretty good with the first 3, but I haven't really tried it with my own files yet. Here's a problem that we have with search. If you go into Google or some other search engine, you want information, but a search engine doesn't really give you the information you want. Craig [00:09:21]: It gives you a list of websites that may or may not include somewhere the information that you want. Perplexity is much more about giving you the actual information that you're trying to to get and telling you where it got that information from. And so that's a fundamental difference, and I think it could ultimately reshape how we search for information on the web, and I think that's a good thing. There are a number of things that set perplexity apart. It's got a copilot mode that gives you a guided search experience. That can be really helpful. So what it does is it will ask you first, do you wanna focus on particular types of Resources. So right now, it's got 6 different categories. Craig [00:10:09]: Well, 5 and then an all. So it's got all where you search across the entire Internet, Academic, and this is a big one, where it searches only in published academic papers. Writing doesn't really search the web. It just helps you. It'll...