Brent Warner 0:00
We’ve talked about Gemini, but there are plenty of other AI tools out there that teachers are using, and high among the most common is Anthropic’s Claude. We’re looking at Claude on this episode of the Higher Ed Tech Podcast, Season Eight, Episode Four.

Tim Van Norman 0:26
Welcome to today’s HigherEdTech podcast. I’m Tim VanNorman, the Assistant Director of Technology Services at Irvine Valley College, and Adjunct Professor of Business at Cypress College,

Brent Warner 0:36
and I’m Brent Warner, Professor of ESL here at IVC. We both enjoy integrating technology into the classroom, which is what this show is all about.

Tim Van Norman 0:44
Welcome. We’re glad you’re here with us.

Brent Warner 0:48
October!

Tim Van Norman 0:50
It is October. Wow. This has been quite a semester, and now we’re in the middle. You know that it’s it’s amazing to think about all the different things that are going on. First eight week classes are ending. Second eight weeks starting. It’s been quite a semester. All kinds of different moving pieces. So

Brent Warner 1:14
Yeah, yeah. I have been a little bit separated from all those things right now. You know, as I mentioned, I’m on leave, and so it’s. I’m trying to keep track of everything, but I’m gonna have a lot to come back to as well. So, well, there we go. So today we’re gonna talk a little bit about Claude because you you know we did a big Google conversation that was more kind of the breakdowns on several individual features because Google’s so big they kind of have their different breakdowns in their parts. But you said, well, hey, hold on a second. Well, let’s talk about Claude a little bit because we know there are lots of professors out there and lots of students. But I really do kind of think that Claude has somehow positioned themselves as like the generally speaking the academic choice. There’s some people that are really into perplexity as well for that. But but today we figured let’s talk a little bit about Claude and and you said you haven’t really used Claude, right?

Tim Van Norman 2:18
No, not at all. I know about it quite a bit. One of the programmers that works with me has uses it constantly, loves it. It’s constantly comparing Claude and Gemini and and Copilot and stuff. But it’s amazing some of the things that I’m hearing about Claude. There’s a lot of positive things that I’m hearing, but you know, not all positive about the company, but about the about Claude itself. Yes.

Brent Warner 2:47
All right. Well, let’s start with a little breakdown of the company, which is Anthropic. And so, yeah, we do have to. Here’s the problem with all one of the problems with all of these AIs, Tim. You know, this is my my take for a long time. Is like, don’t trust big corporations. They are not good. They they are not people, and therefore they are not good, right? And not to say that they’re necessarily bad, but they don’t have a moral stance as a company by themselves. I get very worried about what these companies do with our data. They’re all trying to train on data all the time. Okay, so one of the big things early on-not early on, but I think last year, anyways-that Claude got sued and in trouble for stealing basically like so much written work, and they were doing like all this damage to like so many books. I think they were one of the companies that kind of got caught for essentially making bulk orders of all sorts of different books from weird independent stores, and they were chopping the books up and scanning every page, and then burning the books essentially, and in order to, you know, in order just to get the data, to scrape data, data, data, data. It turned into a weird problem, and then I’d have to go in and look at the history. Sorry, I didn’t do the the research as deep as I would have liked to, but but they did get into a lawsuit where they had to pay all of these authors some amount of money for you know essentially stealing the work of humanity that was not meant to be for these purposes. But the thing that’s a bummer about it is that they did not get charged all that much. I mean, I think the authors who got in on the class action lawsuits like $5,000 each or something, which you know, I mean, what is that, right? That it’s tricky because most authors do not really make $5,000 on their books, right? And then the do those who do make a ton, ton more, but they’re you know that balance is really weird, and it’s not what people wrote those books for, so that’s a problem. But of course, everybody has stolen data, not to justify any of it, but that is one of the problem areas. And then the other one that really scared me last week or two two weeks ago, Tim, I don’t did you. On the news when this kid came on, he quit Anthropic just before their IPO quit, right? So this is one of the reasons why it’s really interesting because it’s like he quit, and he he went on, you know, who was it, Anderson Cooper, and he was he basically said that he that all of the people inside of all of these companies deeply believe that AI might end humanity within the next 10 years.

Tim Van Norman 5:30
That’s an interesting, scary thought, and

Brent Warner 5:33
It’s horrifying, right?

Tim Van Norman 5:36
Yes.

Brent Warner 5:37
And so he went like, and the reason why it’s extra horrifying is because he quit before the IPO, meaning he’s not getting this huge cash payout for it in theory, right? I mean, assuming that this is a legitimate thing. Other people said, “Hey, this is a hoax. Like it’s it was set up by the company in order to boost their valuation. I’m not, you know, I mean, humanity is crazy. That should not boost a value valuation on a company would be my thinking, but…

Tim Van Norman 6:02
Yeah, obviously I I’m not following that thought. But you know, I guess there’s no such thing as bad bad publicity or something. I don’t know. Anyway…

Brent Warner 6:12
That’s kind of what people are saying, yeah, I don’t know. So there’s a whole thing with this, and by the way, it’s worth going and looking at what different people from different perspectives, have to say about it because some people are like, and it’s really weird. Stay away from the political conversation on it because you’ll get looped into both sides on both arguments from multiple political perspectives that weirdly agree with each other. It’s just like it’s a mess, right? So just try to go look at people who you trust that are you know generally well known or reputable in the field. It is to me. I’m like I don’t know what to do with some of this information except to try to be as careful and thoughtful about what we’re doing and interacting with our our students and the people who are in front of us as much as possible. The one thing I will say that’s nice about Anthropic is that they’re generally more willing to openly admit problems with AI? They are one of the companies, and many of them did it just now with this, which is saying essentially we need more guardrails. We need to be, you know, we need to slow down. It’s not a question of beating you know China or whatever it is as the race because the problem is, if we’re talking about a humanity extinction level event, it doesn’t matter if we get to it first or China gets to it first. It’s still a humanity level event, right? Right. And so, but anyways, they have been more willing to talk about this. This is horrifying stuff, right? We’re not trying to get too into the horrifying side of things, but they’ve been willing to say this. So, hey, we’re willing. We’d like to slow down on these things. We want to have recognition. They do generally kind of try to do a little bit more in terms of like protecting some information and being a little bit more ethical. You know, they were one of the ones that rejected the military contracts because they didn’t want AI being used for all of these military purposes without any accountability in it, and so, so there are some points inside of there that I’m like, okay, I can get behind this, not that. So, so anyways, I just want to make sure that we kind of clarify for sure before we talk about all this. There are problems with every company, and you know, these are some of the ones here.

Tim Van Norman 8:23
Absolutely. So now that we’ve talked about Anthropic for a moment, let’s talk about Claude. I’ve heard some really good things about Claude. Honestly, I haven’t heard that much bad, yeah, per se. So, but some of the stuff that I’ve heard has had to do with let’s start with document analysis, the ability to get in and maybe it’s because they stole all those books or whatever that they can actually analyze documents better, but that it seems to do a really good job of research and then analyzing documents for various reasons. Well, what have what have you seen on that?

Brent Warner 9:04
Yeah, so I use it a fair amount. I haven’t uploaded a ton of documents myself, but I have used it for you know some levels of research, checking information. You know, there are still hallucinations inside of it, as there are with all of these things, and so you still kind of have to do your due diligence. But I’ve found that it’s much better about, especially when you’re like being specific inside of it with research, is like showing where that information came from, double checking the quality of what’s going on, like that that it’s presenting the information accurately, all of those kinds of things, and so you know, any of these tools that we talk about, Tim, they’re going to be you know the horse race. Like today, this is better. Tomorrow, this is better. But over the long run, and until this point, I will say that Claude has had a pretty good reputation for being reliable-ish in research and. Me, I always come back to this conversation where we say, “Well, hey, how reliable is that information? And I kind of always say, “Well, it’s not totally reliable, but we have tricked ourselves into believing that human-made research is more reliable than this sometimes. And I’m not totally sure I believe that, right? Like a lot of times, human research goes awry, or you know, it was presented as fact, and then you know, a year later, it’s like, oh, they made up all this stuff, you know, like whatever. All these things happen, right? And so, I kind of have some feelings that are like close around the same area with those, which is, yeah, research is interesting, but it’s fallible in any aspect that you’re looking at it. So, in terms of generally getting information together, finding things, and then connecting information across multiple sources and figuring out what’s going on, what people are talking about, I think that that is where a lot of, especially research professors, are spending time using these things and trying to link these ideas. and And I’ve seen a lot of reports, you know, people on LinkedIn who are talking about these things and you know deep dive researchers and they’re like okay yes I am seeing like this is pulling out good information this is showing me things that are useful for it and so if you’re trying to do research in support of work that you’re doing Claude might be a good assistant for you in there

Tim Van Norman 11:19
Nice

Brent Warner 11:20
Yeah.

Tim Van Norman 11:21
The other thing that I’ve heard a lot about is coding. Yeah, and that’s probably the main thing I’ve seen use of Claude for. One of the guys that I work with, he uses it at least to evaluate his code constantly.

Brent Warner 11:41
Yes,

Tim Van Norman 11:41
and I think that’s that’s actually really neat. It’s something that we’ve always had. As you know, I was taught to my bachelor’s degree is in coding and stuff, and taught how to do that. And you always wanted to have somebody review your code. Well, now you can, and it doesn’t have to take Somebody a month to review the code that you did. It’s you know five minutes, and you’ve got a really deep analysis on what you wrote.

Brent Warner 12:14
Yes. So let me point out two things inside of here. So there’s using it to check your code, and there’s using it to make code, and we will talk about. I use it to make code all the time, but I want to put a big caveat on that when we talked about it. So, so hold off on that for a minute. When you’re using it to check your code, from from what I understand, super strong, right? Because ideally, someone like you who has actually properly learned how to code is going through and creating things, and there’s one tiny bug somewhere in the middle of the code, and it’s ruining everything, and it can go through in a second and find that, and that’s not really a valuable use of your coding time to spend, you know, a full day of work, just scanning every line over and over and over again, and like looking for that tiny little period where there should have been a semicolon or you know whatever else it is. So for sure, I think it’s really powerful for that level of coding, and I’m sure that you and you know the people in your department have seen a lot of that upload things, see what’s going on. I am also imagining again. This is not my area of expertise, but like really good for checking for security as well. Like saying, hey, like like is this going to be locked down? Is this going to be whatever? You know. So I think that those kinds of things are great. I don’t have the expertise to talk about being a coder, but we can talk about vibe coding, which we talked about at the end of last last season. I think it was season season seven, episode 18, maybe 17 or 18. At the end of last year, so we talked about vibe coding, and I am using it all the time for vibe coding stuff. So, so if you if you don’t remember, vibe coding is where you are asking AI to create apps or services or you know interactions or whatever else it is for you, and you just have the idea of what you want, and the AI creates that for you. So you can do that in Google Gems. You can do that in Claude CoWork, and all sorts of different things. I’ve been doing it a lot over the last week or two, playing with some just quick and silly, fun, low stakes. We talked about this before, I think, as well. Like low stakes, non-invasive, interactive online games. Right, things that are not going to be a big problem. Things that are not going to create security vulnerabilities. All of those kinds of things. Right. This is not going into and reaching across. You know, I’m not I’m not plugging it into Canvas, for example, and then having it reach into my Canvas account and being able to pull my students’ information and you know then have them interact with it and then send it into Gradebook. I don’t think that’s even possible without an API, but who knows? You might sneak sneak sneak a secret API in there, I guess. But but all all to say is like it’s its own little book, right? It’s like, hey, here’s a little notebook. Here’s a little comic book on the side. You can look at it. You can have some fun with it. You can enjoy it. It’s not our textbook. It’s not our whatever. It’s just supplemental, and it’s not going to affect anything negatively if you interact with this thing. So, Tim, I showed you pre-show a couple of activities that I’ve been building, and I’ll put links to them in the show notes as well. But I’ve been building some of these games in a platform called Teacher Hive. And have you used Have you seen anybody use Teacher Hive before, Tim?

Tim Van Norman 15:59
Other than what we looked at pre-show, no, I had never heard of it.

Brent Warner 16:03
So currently, it still feels like it’s a lot of K 12 using it, but I imagine that as it becomes more popular and more well known, then other people will use it. But basically, what it is is you vibe code in a an app like Claude or Gemini or whatever else it is, and then it gives you the HTML for the activity, or for the game, or the thing that you’re trying to make, and then you plug that code into Teacher Hive, and it will host the activity for you, right? And so you’ve just got the link inside of there. It’s kind of similar in the same way to Play Lab, right? Play Lab is like you’re building everything on the back end. This one’s kind of gives you that platform, but there’s not much to interact with except that you’re basically just uploading the the HTML, and it will say, “Hey, we’ll just do these things for you. You can also pay more to get some of the added features, but if you want like full AI interaction, which this is one of the favorite things for me about about vibe coding, is the idea that like or or any AI interactions is that it’s not just one layer, right? We’ve talked about this before, right? Okay, here’s the vocabulary words. Show me that you know what they mean. Okay, you know what they mean. Part two: Write a sentence using them. Okay. Part three: Write a sentence using them. Show like where the vocabulary words actually use the word in context, and then we will use the AI again to analyze that and give you feedback on the quality of your sentences. Right. So multiple layers that go deeper and deeper as compared to what’s the right answer A, B, C, or D? That doesn’t require any AI at all, right? That’s just the right choice or the wrong choice. And so, for this, you can basically do all of this inside of a quick link inside of Teacher Hive. I built a few games up until now. I did one. I was showing you, Tim. It’s like which one doesn’t belong, right? So it’s like four different choices, four different things: an airplane, a car, a bicycle, and a bus. Which one of these don’t belong? So for my conversation class, activities like that are really good warm-ups, right? And so the students are like, “Oh, the the airplane doesn’t belong because it has wings, and then the students are like, “No, the bicycle doesn’t belong because it has two wheels and it can’t balance by itself, right? And all the other ones can balance on the ground by themselves. So they get into these long conversations, and I built a little app that will bring that up, and then you can click on it and see what some of the suggested answers for whichever choice the students make are. Conversation back and forth. So I am really liking it for things like that. I can also build little games for my students to play on their own time, or they can just kind of review. So you know, we’re in Halloween season now. I made a little article game where the player throws brains to the zombie of you know whether they’re using the the article a an or no article. I’m not going to worry too much about the grammar, but or sorry, I worry about the grammar for our audience. You don’t need to worry too much about the grammar, but then it’s like it’s just a fun game for students to play, and it’s like just pick the right answer, and then we’ll have some fun with it. So these types of very light things are really great for this coding because one, I’m not asking for any students’ information, right? I’m not saying like I’m not having it pull any information. I’m not having it ask ask their names or whatever else it is. It’s all stuff that is just like, hey, we’re generating a little bit. We’re having some work done out of it, but it’s not going into grades. It’s not going to my grade book. All of those kinds of things, and so there’s a lot of fun and a lot of really cool things you can do. And Claude is outstanding at it. So Claude basically puts the whole game right into Claude.

Brent Warner 19:56
They call it an artifact, and then you. Still have the chats with it, and as you’re building it, you say, “Hey, this part works. I’m testing it. It’s great. This part I don’t like, or I want to change this this amount and this amount. And so then it it does all of that right inside of there, and it catches itself with coding errors. And so I’ve seen several times where it says, “Hey, as we’re as we’re loading this, there’s a problem, and and would you like me to fix the problem? And it’s like you don’t even have to think about it. It’s like yes, I want I want you to fix the problem. I don’t want it to be a broken code inside of here. So there’s a lot of cool things to do inside of there. And it’s it’s if you just start thinking about like having fun, like letting your students have fun, letting your class have some fun with some different ways of playing with things. And I’ve always wanted to do this-that vibe. Going back to that vibe coding conversation, lots and lots to do with Claude because it is really quite strong, and it does a pretty good job at making the code minimal too. So it’s not like massive, you know, overloaded and slowing everything down when you’re playing it.

Tim Van Norman 21:00
I love the fact that it really gets what you showed me gets really away from just a text-based game or a text-based world when it comes to AI. This is generating images. You had zombies, you know, little characters that were zombies, and you had little red things that were brains going out. You had a field, and you know, so you had different components that made that up. And so that was that was really interesting to to be able to see that.

Brent Warner 21:30
Yeah, and that’s all basic HTML. It’s not even taking you know. It’s like the quality of it. You’ll see it looks kind of childish because it’s basic circles and shapes and things like that. But like you can do a lot, and I’ll I’ll put a few links to some of the things that I built and some of the things that other people that have built to kind of show some of the variety, the possibilities with that as well. But but I I have found that Claude makes better code. I was trying actually today earlier I was coding with both Gemini and Claude, and Gemini kept farting out on it. It was like, oh, it’s too much, or you know, like, or it’s running out of token or usage or whatever. And Claude was pushing through and doing a great job. So, so I’ve been quite happy with it in that level.

Tim Van Norman 22:13
Nice. So you mentioned before PII and stuff like that, how you’re making sure that it doesn’t have it. One of the things that we’re seeing quite a bit in California community colleges is people wanting to have their vibe-coded apps linked into Canvas, and they maybe they’ve found the API to do it. Maybe they have, you know, their friend has done something. If you want to go through the effort of getting a type two sock or sock two type two compliance document put together, your VPA, your Hec VAT, your all of those other things, feel free to you know at least ask if you’re not willing to do that. And by the way, Claude won’t do that. Your teacher Hive probably won’t provide that. Those just understand we’ve talked about FERPA compliance before. From a security standpoint, we as developers, we as security people, we as managers, the risk that something could from your class could potentially take away information for about your students-that’s just way too high. We’re not willing to take that risk. We don’t do that with our own stuff. When I talk about you know a programmer using it. They’re either using a using a level of CLOD that is paid for such that it’s not going to learn off the data, or they’re stripping out all of the pertinent information so that they don’t get in trouble, so they don’t have These violations, and right now, this is huge. We’ve had a number of instances where people have have been asking us to include and link to their vibe coded application in Canvas. It’s one thing to have a link in Canvas that goes to that. Students don’t need to log in. There’s no student information. Fine, but we’re really against having student information, and so often you will be shut down. And that is honestly, to me, that’s like low stakes. We’re just trying to be careful with your information too.

Brent Warner 24:41
Yeah, as as these should be should be shut down for doing anything. So for me, these things are you know outbound link. It’s on a separate page. Nothing’s being collected. Have some fun. If I want to check it, I can ask you to do a screenshot of it and send me a picture of the screenshot at the end, or you. Can call me around in the classroom or whatever else it is. There will be, I think, as we move forward, there will be more ways that are more secure. And I would imagine that even an app like you know maybe maybe like Teacher Hive or something at some point might be like big enough where they’re like, hey, hold on, let’s actually get this thing so that it is trustworthy at certain levels, and we have some sort of like multi-lock system where things move through. You know, I don’t know what it what it looked like. I’m sure there are people playing with it, but just be so careful with private information, personal information. If it’s not FERPA for other things, like it’s you know, it’s not going to be FERPA in here, right, and so so be very careful with it. You know, there are just some people that are just getting a little too loosey goosey with what they can make and what they can do, and they’re kind of willing to connect AI to their banks and to their medical records, and then that means to me that’s like, well, then that means you probably don’t care about your student’s name and address that much, right? Like, and so, so please, please, please be very, very careful. Stay away from anything that is, you know, giving information to these. Even if it claims that it’s FERPA compliant, you still wouldn’t want to upload documents with your students’ names and information about them, right? Just it’s a new world, and you got to be careful with it.

Tim Van Norman 26:25
And and remember, if even if you are really careful with yours, just because you have an API and you can use Canvas, don’t please, because we don’t want. If I give you access, I may have to give somebody else access, and they’re not as careful as you are. On top of that, how am I going to prove because I’m the one responsible for it your interaction with Canvas? Okay, how am I going to prove that you didn’t make a change later and do something? So basically, probably not going to happen. Not going to allow inter integration into Canvas, for instance. I’m using Canvas as a for instance, but just understand when we say no, there’s a lot behind that. That we say no to larger companies that don’t go through the rigor that we need as well.

Brent Warner 27:23
Yeah. For sure.

Tim Van Norman 27:24
So, it’s not about you, it’s about everybody.

Brent Warner 27:28
That’s right. That’s right.

Tim Van Norman 27:30
Yes.

Brent Warner 27:30
And then Tim, just for a brief second, I’m not going to spend a long time on this. I think we should do a longer show on this in the future. But Claude has also been getting a lot of props for their agentic AI, so that means you know AI that is not just going out and responding to directly what you’re doing, but they’re kind of doing multiple steps and interacting, pulling from multiple resources, and then you know bringing that all together, whatever else it is. I haven’t played with it a lot, but what I do understand is that with Claude, you and maybe with all of these ones, but one of the ways, if you’re interested in trying this at the lowest levels, because eventually this is going to become a part of everybody’s lives at some level, which is that your you can still build a folder, like a single folder, and tell Claude you only have access to messing around with what’s inside of this folder, right? Or what’s inside of the subfolders? And so, if you had, for example, then okay, I’ve got some student documents again, names stripped out of it, and I’ve got you know some data showing me from you know my own my own research or the the information that I’ve collected, I’ve got these kind of problems going on in my classes with these types of students. How do I fix them? Then AI might really be able to help you kind of see what’s going on with some of those things, and you can get it to pull in from multiple sources and and you know do some cool stuff if you’re really into agentic AI, it’s probably most people listening are still kind of not at that step yet. I’m still kind of only at the very first levels of the step, and I would have imagined that I would have been into it starting a year ago. But some of it’s been scaring me. I just don’t want to give access to a lot of my information. But anyways, I do understand that there are some decent safety protocols in mind that you can do if you’re being again thoughtful about what you’re choosing to give it access to. So as always, you know that’s kind of what’s going on. But Claude is quite strong in a lot of those places, and if you, I feel the most comfortable typically with Claude, except that it’s currently not with our school, so I’m not really using it for school-related information. I’m only using it for my own stuff that I’m doing outside. We have a Gemini license at our school, and so that’s kind of what I use for like more school-ish related things. But again, be careful with with what’s out there. But Claude can do a lot of cool stuff that’ll be a lot of fun for you too.

Tim Van Norman 30:06
Absolutely. Thank you for listening today. For more information about this show, please visit our website at thehigheredtechpodcast.com.

Brent Warner 30:17
As always, we want your feedback, so please go to thehigheredtechpodcast.com, and let us know your thoughts

Tim Van Norman 30:23
for everyone at IVC that’s listening. If you need help with technology questions, please contact IVC technical support. If you have questions about technology in your classroom, please contact me, Tim Van Norman, at tvannorman@ivc.edu.

Brent Warner 30:35
And if you want to reach out to me about the show, you can find me on LinkedIn at @BrentGWarner.

Tim Van Norman 30:41
I’m Tim VanNorman,

Brent Warner 30:42
and I’m Brent Warner, and we hope this episode has helped you on the road from possibility to actuality. Take care, everybody.