ChatGPT in Education: The Pro's and Con's

0:00:01 - Announcer

You are listening to the National University Podcast. 

0:00:10 - Kimberly King

Hello, I'm Kimberly King. Welcome to the National University Podcast, where we offer a holistic approach to student support, well-being and success- the whole human education. We put passion into practice by offering accessible, achievable higher education to lifelong learners. On today's episode, we're discussing Chat GPT and how it is being used in education. And joining us is Dr. Ron Uhlig. 

Dr. Uhlig is currently the chair at the Department of Engineering at National University School of Technology and Engineering. From 2010 to 2014, he served as Dean School of Business and Management. He returned to the engineering faculty in 2014 as an academic program director for the Bachelor of Science in Computer Science program, responsible for the successful effort to secure A-BET accreditation for that program. During 2005 to 2010, he served in multiple positions, including chair of the Department of Computer Science and Information Systems and academic program director for the Master of Science in Wireless Communications, as well as principal investigator for two Hewlett-Packard Technology for Teaching Grants. Wow. He currently teaches courses in cybersecurity, engineering management, engineering economics, technology and globalization, and computer ethics. And we welcome you to the podcast, Dr. Ron Uhlig. How are you? 

0:01:34 - Dr. Ronald Uhlig

I'm great. Thank you very much. Good to be here. 

0:01:36 - Kimberly King

Yes, wow. This is impressive and how relevant with what's going on in the world. So why don't you fill our audience a little bit in on your mission and your work before we get to today's show topic? 

0:01:47 - Dr. Ronald Uhlig

Well, as chair of the Department of Engineering, I have four different degree programs. One is a Bachelor of Science in Electrical and Computer Engineering, another one in Manufacturing Design Engineering and one in Construction Management, construction Engineering and Construction Management. And then we have a Master of Science in Engineering Management. And we're looking at some other engineering degrees also to offer in the near future. I hope wonderful. 

0:02:18 - Kimberly King

Well, this is an amazing topic, and we're talking about Chat GPTt in particular, and AI text generators. And so, Dr. Uhlig, what is generative artificial intelligence? 

0:02:31 - Dr. Ronald Uhlig

Well, generative AI is a term that we use for using computer software to generate text, but also images, music, you name it. Whatever a computer can do, you can get computer software to come up with it. 

0:02:56 - Kimberly King

Can I wash dishes? Can it wash the dishes or do our laundry? 

0:03:05 - Dr. Ronald Uhlig

Not yet right. No, not quite yet. But I mean it includes audio. It can even do computer code, so that has a lot of people watching very carefully too. A lot of programmers are watching. 

0:03:20 - Kimberly King

Absolutely. 

0:03:20 - Dr. Ronald Uhlig

Carefully. 

0:03:22 - Kimberly King

Well, so we're hearing all about this Chat GPT and what are some of the ways large language models are going to be employed in business. 

0:03:30 - Dr. Ronald Uhlig

Well, people are already starting to use them for all kinds of applications. It was kind of an interesting article recently that McKinsey and Company did. They actually said to Chat GPT, tell us what the impact is going to be on business of Chat GPT. And Chat GPT came back saying the rise of generative AI has the potential to be a major game changer for businesses. They talked about the power to revolutionize industries and transform the way companies operate. So people are giving it some pretty big thoughts as to what it can do. For example, it can increase efficiency, increase productivity, reduce costs, open up new opportunities for growth. 

It's being used in marketing and sales in operations in information technology and engineering, particularly in writing, risk and legal research and development. All of those have applications of Chat GPT. 

0:04:45 - Kimberly King

It is truly amazing, and I've been reading a little bit about what is Chat GPT going to be doing in hospitals or with doctors in the medical industry as well. So businesses all around. 

0:04:56 - Dr. Ronald Uhlig

Yes, there are a lot of different kinds of businesses, you name it. It's really just a matter of providing the training data for Chat GPT to work with, to draw from. 

0:05:11 - Kimberly King

Can large language models make mistakes? 

0:05:14 - Dr. Ronald Uhlig

Absolutely. 

They can and they do. The term that's being used right now in the industry is that it can hallucinate, which is kind of a funny term, but it can come back with information that's not accurate. It can come back with information that's just plain wrong. I've seen it. I've seen people ask Chat GPT to write a report for them in the normal, like a technical paper, and it will generate references. And sometimes those references don't exist. So you really have to be rather careful with it because there aren't filters yet. Probably the most dangerous is that there aren't filters too that can remove inappropriate responses or inappropriate language. So you have to be very careful. 

It may say something that you don't want your company, for example, to say. If you're using it to generate text for your company, biases can exist. It's basically trained on the data that on everything that we have in everyday life, and whatever biases exist in contemporary culture will exist or can exist in what gets produced by Chat GPT as well. So one really has to be careful. It can also plagiarize because it's going in and looking at whatever is out there. It may not even tell you that it's plagiarizing. So you have to be rather careful about that. Students have to be particularly careful. 

0:07:00 - Kimberly King

I was just going to say that's. My next question is what should students know about AI and large language models and about that plagiarism with Chat GPT? 

0:07:09 - Dr. Ronald Uhlig

Well, there are a lot of positives about Chat GPT and these kinds of tools. I do want to say Chat GPT isn't the only one, but it's probably the one that's gotten the most press. On the positive side, Chat GPT can give students ideas on how to express thoughts. If it's used in the form of tell me how I might say this and then the student takes that and paraphrases it themselves, then it can be very helpful. It can help them think about it in ways that they may not have thought about it before. For example, kind of like working with another student, with a fellow student working together. On the other side of it. People need to be aware that there's no ethics built into this, there's no morals built into it, because computers don't have any morals. They just follow whatever their program tells them to do and they don't have any experience of the world. And so that's where the student has to come in the student has to really learn how to use this and not just kind of use it blindly. 

0:08:23 - Kimberly King

More like a guideline. And then something to prompt their imagination to continue writing about it, but not for the value that it just gave you. You don't take it for that value. You need to proofread. 

0:08:35 - Dr. Ronald Uhlig

Exactly. You have to proofread and think about what's being said. Do you really agree with what it's saying? 

0:08:43 - Kimberly King

I mean I get in trouble if I send out a talk to text on my phone and I don't proofread that. And what if I send it and it said something completely what I didn't say? So that to me is the very basic level. That's not necessarily a thought, but that's just trying to copy what I'm trying to reiterate there or send, and if I don't take a look carefully, who knows what I just said, right? 

0:09:08 - Dr. Ronald Uhlig

Yeah. “Wow, did I say that?”

0:09:10 - Kimberly King

Right, you have to be careful. And then, so students, yeah, they need to make sure that they're engaged and that they are watching everything. Can Chat GPT AI also be used to complete assignments? 

0:09:28 - Dr. Ronald Uhlig

Yes, absolutely. I've just been working with a group of fellow professors. We've prepared a paper that I will be presenting at the annual conference this year of the American Society for Engineering Education, and one of the things we did in putting this paper together was we got some of the questions that students have to answer in what we call discussion boards in national university online classes in cybersecurity, thank you. And so we, as the prompt to ChatGPT, we put in the question exactly in the form that it appears to the students in their assignment, and ChatGPT came back with some very good answers, and so, in fact, it would pretty be very difficult to know that it was written by ChatGPT instead of by the individual student, and that raises a lot of questions in itself. 

We also tested the ability of ChatGPT to write code, and it was given the instruction to write a particular kind of code for a neural network. A simple neural network. It did a great job. Not only did it a good job, but it also produced the kind of notes that programmers sometimes skip over. So it did a really super job of producing the code, but we found that when we gave it more complicated coding assignments, it made mistakes. It wasn't perfect. So, once again, you really need the human being to be in the loop. 

0:11:24 - Kimberly King

Just a question for you, since you have provided this for your classroom. What are the guidelines or the guideposts for the rules, and is every school different that allows ChatGPT, or is it allowed, or how can you tell? 

0:11:39 - Dr. Ronald Uhlig

Well, this is going to vary with the university. National University - right now we haven't come up with any general rules. Some people, some professors, don't want a lot to be used at all. It of course depends on the kind of class. I'm talking primarily about engineering, computer science, technology classes. That's a different matter than if you're taking an English class. 

0:12:03 - Kimberly King

Exactly, exactly. 

0:12:09 - Dr. Ronald Uhlig

But I believe that ChatGPT can be a great collaborator. And that's the guideline if used properly to use it that way. But you really need to acknowledge that you've used it and let people know. This is an issue by itself. Some people, some I've seen, in fact, an example where a student was suspected of using ChatGPT to write an assignment and the professor wanted to give them a zero. And we said wait a minute. Yes, I know that the detector said that there's a probability that it used it, but the problem is up until very recently, the detectors have been coming up with a very high rate of what we call false positives, in other words, about 10% of the time, even though the student hasn't used ChatGPT, it'll come back and say that they have. 

Really And so you're going to get in real trouble, then You're going to cause real harm to a student if you mark them down and they haven't used it. That's not right either. Right. 

0:13:28 - Kimberly King

Well, okay, a question for you. What if you have a classroom of 25 students, say, and you give them the same assignment and they all plugged in to ChatGPT? Would they come up with the same answers? 

0:13:41 - Dr. Ronald Uhlig

They might if they gave it exactly the same prompt. 

0:13:45 - Kimberly King

Does it take into consideration the person's cadence or beliefs, as you say? 

0:13:52 - Dr. Ronald Uhlig

I think no. What it goes by is whatever prompt you've given it. Okay, so if one student says it a little differently than another student, then they're going to get a little different answer. 

0:14:03 - Kimberly King

That would be an interesting class, just to see what comes out. It's so interesting. Can AI tools like ChatGPT pass exams? 

0:14:16 - Dr. Ronald Uhlig

Absolutely, yes. There have been some major studies done. This has changed in the last, I’m going to say, two months. I can't remember the exact date when GPT-4 came out. Gpt first hit the headlines last November, then it was GPT-3.5. A new version came out recently and this new version is incredibly good at passing exams MBA exams, bar exams, medical exams, all kinds of things. 

0:14:53 - Kimberly King

I think I'm going to go to medical school. I was just telling this story that back in the day it sounds like- I'm so long ago, but it was before cell phones. Before that, you had Google at your fingertips. I was a speech communication major and literally I would come in. The professor would say you have five minutes and you need to research this and you need to be able to argue on both sides. Think about that. Without a cell phone today, I think we're all spoiled. I had to actually do the deep dive. This is fascinating. You can pass exams. This is great information and I can talk to you for hours. 

We're going to have to take a quick break, but stay with us. We have more in just a moment. 

-break-

And now back to our interview with Dr. Ronald Uhlig and we're discussing Chat GPT and how large language models work. And Dr., this is so interesting. How are people reacting? 

0:15:50 - Dr. Ronald Uhlig

Well, as with any major development like this, you've got all the extremes. Some people are saying it's cheating, ban it. And some whole school districts have done that public school districts. Other people are saying this is the answer to all of our problems. And, as usual, neither extreme is right. And most of us are probably somewhere in the middle, leaning one way or the other, but somewhere in the middle. I like to think about it as similar to the debate when calculators were introduced into the classroom, 25, 30 years ago, you know, and some teachers said no, no, no, no, you can't do this. Well, nowadays, everybody uses a calculator. 

0:16:40 - Kimberly King

Right, it's built in, yeah. 

0:16:43 - Dr. Ronald Uhlig

And teachers wouldn't think of banning it from the classroom. At least I don't think many would. So we kind of figured out yeah, we can manage to automate that part of our tool without losing our humanity, so to speak. So the biggest impact, the biggest concern. Some people are saying well, we can't give exams anymore because people can take the exam using ChatGPT, so we can't use them to assess knowledge. One writer said it's amazing, it's better than the average MBA.

So I like to think of it as a partner. You know, there are some things that we're, some things that we really are good at as humans, that computers just can't do, and so it needs to be a partnership. What AI does well, what artificial intelligence does well, and what humans do, they're not exactly the same thing. So what kind of partnerships can we develop? How do we work together? But we do need to teach students how to use these tools, and we're just learning ourselves what are the best ways to do it. I'm sure we haven't come close to exploring all of them yet, but at the same time, business is putting them into use. For example, we've got to generate a new ad by the end of the day today. Well, what are you going to do? You're probably going to use, or you're probably going to ask ChatGPT to give you your first copy. 

0:18:26 - Kimberly King

And you said in the very beginning about being more efficient, cost saving. So, get rid of the ad agency and bring on ChatGPT, or you know… the other thing, I guess the skepticism on my behalf, on coming from me, would be, are we dumbing down… Are we dumbing down ourselves by leaning on ChatGPT? 

0:18:47 - Dr. Ronald Uhlig

Well, I mean, I would go back to the calculator example. Have we dumbed ourselves down by not insisting that everyone know their multiplication tables anymore? 

0:18:54 - Kimberly King

Right. 

0:18:57 - Dr. Ronald Uhlig

I don't think so. 

0:18:59 - Kimberly King

I like your idea of the partnership, And I think that's what we have to think of. Moving forward is. You know, we add what we get from the ChatGPT, but you have to humanize it as well. Should the use of ChatGPT by students be considered cheating? 

0:19:17 - Dr. Ronald Uhlig

If they do it without letting you know that they've used it, yes, it is cheating. Actually, I mentioned before the paper that we've written that I'll be presenting at the conference in June, and I actually did an ethical analysis there. I, when you mentioned earlier okay, I teach computer ethics sometimes, and so I use two different kinds of ethical theories. One of them is rule-based and the other one is called utilitarianism, and no matter how we looked at it, yes, it's cheating. You can look at it from different perspectives, but either the student is cheating themselves out of knowledge that they should be learning, that they're not learning, or they're cheating society out of knowledge, and either one, no matter how you look at it, yes, it is a form of cheating. If you hide the fact that you've used ChatGPT to do it, yeah, that it goes back to that plagiarism. 

0:20:21 - Kimberly King

How do you think educators would assess a student who's taking an online exam versus taking it in person? Wouldn't it be easier to use AI tools online? 

0:20:34 - Dr. Ronald Uhlig

Probably so. Yeah, I would have to say so, but you know, online is really, particularly for National University and for our working adult students- which is the largest part of our student body- they need that flexibility. So, rather than saying you can't use ChatGPT, what we need to do is figure out the best way for you to use it. 

0:21:06 - Kimberly King

And you kind of talked about this a little bit earlier, but about the AI tools by students, can they be detected by instructors? And that I know you kind of walked into that, talk about that a little bit. 

0:21:16 - Dr. Ronald Uhlig

Yes, they're getting better. I said the tool. there was about a 10% false positive rate. I just learned a couple of days ago that they've got that down to a much smaller percentage now. I think it's I don't know it's somewhere around 2% or 3%. But it can still make mistakes. And you know, I’m not sure that I'm willing to accuse a student for the sake of that 2% or 3% false positive rate. 

0:21:51 - Kimberly King

Right. It’s a slippery slope right now. Isn’t it?

0:21:53 - Dr. Ronald Uhlig

It's just dangerous. But you can more and more. You can probably detect it. Basically, the paper that I mentioned we asked at first, we asked one of the detectors was the- what did it say that the response that was generated to the discussion board question- what was the probability that it was generated by AI? And it said 98.8%. So yeah, we were pretty sure about that. And then we said, well, what about this paper? And it came back 20% probability that our paper was written by AI. And the reason that I think it was that high was because we had a lot of quotes of what GPT came back with embedded in the paper. So probably 20% of the paper was responses from GPT. 

0:22:50 - Kimberly King

So if you had, if you were a professor and you wanted to go back and see if anybody you know detect if the papers were written by AI, you had a specific question - and it kind of goes back to what I asked you earlier- Do the professors now go into AI and ask that the given question? Just to double check?

0:23:11 - Dr. Ronald Uhlig

A few of them are starting to. Okay, just starting to. People don't know exactly where to go for these tools yet. Yeah, National University uses a software tool called Turn It In for plagiarism detection. Yeah, and Turn It In has now added - it's my understanding, I haven't seen it yet- Turn It In has added a ChatGPT detection tool. I don't know how good it is. I think that's the one that was 10%, but I'm not sure of that. So 10% false positive. 

Anyway there are a lot of people working on it and I think these tools will become available, generally available to professors in the next year or so. That's my guess. 

0:23:56 - Kimberly King

Okay, if we are creating assignments that can be answered by AI, are we really helping students learn? 

0:24:06 - Dr. Ronald Uhlig

Well, I mean, that's really the fundamental question I like to think as an academic, my first job is to my job is to impart some skills or help students gain the skills that they need to do their work. But that means that I've got to give some confidence to them that they know how to use the tools that they need to do whatever we're training them to do or whatever we're educating them to do. And so we've heavily used written assignments, written papers, in the past for assessment. I think we've got to change the way we assess. We can't rely as much on papers. Writing a paper about a topic as the only way we're going to assess a student in the future. 

We have to. There are various tools like problem-based learning, case studies and even hands-on assignments and exams, working together in small groups, that kind of thing. We're going to have to broaden the way we do the assessments in the future. But, as I say, it's not just for me to say, yes, you've passed this exam, you get this degree. But equally important is imparting to the student the confidence that they know how to do what they need to do to know how to do exactly. And be able to move into the future with those tools. 

Yes. 

0:25:45 - Kimberly King

So how could AI tools be used? How else can they be used in education? 

0:25:50 - Dr. Ronald Uhlig

Well, as I think I mentioned earlier, I like to think of the AI tools as a new form of collaboration. 

We have a lot of assignments, particularly in engineering and technology in school technology and engineering, a lot of group projects, and so I like to think of the AI tool as another member of the group that I'm going to consult with. 

I may even ask it to write a section of whatever it is that I'm working on or whatever it is that we're working on as a group, as long as you acknowledge that this particular contribution came from AI. It's just like putting acknowledgement in a paper. I mean, I do that all the time, Professor So-and-So, I learned this, or I heard this from Professor So-and-So, so I learned this by using AI. And students should discuss what they learned, what kind of lesson that they got, what kind of insight they got from using the artificial intelligence tool, whether it's Chat GPT or something else. I think that higher education is going to need to focus on new methods and mechanisms for assessing student learning, as I've already mentioned, but it's clear that AI can show students different ways to express ideas and to structure their assignments. I read a wonderful article recently about a young woman who her native language was French and she was using Chat GPT with her classes to translate the assignments into French. She would write them in French and then use Chat GPT to translate them back into English. 

0:27:53 - Kimberly King

Wow. 

0:27:54 - Dr. Ronald Uhlig

And it just made a huge difference and I thought, wow, I mean, that's the kind of positive things that we're looking for, not just French, of course, but any language Right Think of the issues that we have with classrooms now multiple languages and the struggles that some of our teachers go through in classrooms with students with multiple languages. What an opportunity. 

0:28:26 - Kimberly King

Right, keeping the language barriers at bay as long as, yeah, it's a correct translation, but yeah, that's incredible. Large language models like Chat GPT are changing the world significantly. How is that? 

0:28:42 - Dr. Ronald Uhlig

Well, I think all the things that we've talked about today. I mean it has an impact on business, it has an impact on education, it has an impact on medicine, you name it. Whatever area you want to look at, you have to watch out for all of the dangers that we've mentioned. It can produce wrong answers, it can produce references that don't exist, all of those kinds of things. So we've got to be sensitive to that. But one author actually went so far as to say this is as important as the invention of the printing press. Really, that absolutely floored me because I hadn't thought of it in quite those terms, but you know, if you think about how it, how the printing press broadened access to information, I can see this has the potential to have a huge impact. It's going to very much depend on the directions that we take over the next few years. 

0:29:48 - Kimberly King

And it makes me think about this question- You may or may not know the answer to- but if this author is referencing the printing press, how long has Chat GPT been- not necessarily in existence- but in an idea that has come to fruition?

0:30:05 - Dr. Ronald Uhlig

I don't know how long Microsoft was working our open AI. My impression is it's a few years. So yeah, it's only a few years. 

0:30:18 - Kimberly King

So how fast it's gone, huh. 

0:30:20 - Dr. Ronald Uhlig

But well, I mentioned earlier, before we started talking today, that I had a friend years ago who made the comment that whatever you do, technology enables you to do more of it faster. And so once again we've come. We've got a new tool that is going to enable us to do more of what we do and to do it faster. And if we do bad things with it, we can do more bad things faster. 

And if we do good things with it, we can do more good things with it. The technology itself is not where the problem lies. 

0:31:00 - Kimberly King

I 100% agree with you on that. I wonder if these AI tools will homogenize information and that is, with one all. Will we be getting one point of view, or how can we make sure to use these tools in the best way possible? 

0:31:15 - Dr. Ronald Uhlig

That's going to be an issue because… I read a story about someone who put into, I believe it was into Chat GPT asked them for an analysis of the performance of the current and the past president. And they didn't say what the analysis said, but they commented that it reflected what's out. 

0:31:58 - Kimberly King

Bias. 

0:32:00 - Dr. Ronald Uhlig

The biases that are out in the internet. And I can imagine that it would have. So Chat GPT and tools like it are gonna come back with whatever is out there in contemporary culture. And it's gonna be important, as it always is, that we reflect all the different views. 

0:32:22 - Kimberly King

That's gonna be key. What's your best advice to users of these AI models?

0:32:31 - Dr. Ronald Uhlig

Number one- acknowledge that you're using it. Number two- read it carefully. Make sure you actually agree with what it says. If you've generated some references, check to make sure they exist (laughs). And check out what it says with some friends and some peers and other people before you publish it (laughter). 

0:32:58 - Kimberly King

Do your research. Well, this has been fascinating. I could talk to you all day. Thank you so much for your time. And if you want more information, you can visit National University's website nu.edu. And Dr. Uhlig, we look forward to your next visit. Thank you. 

0:33:15 - Dr. Ronald Uhlig

Thank you. 

0:33:19 - Kimberly King

You've been listening to the National University Podcast. For updates on future or past guests, visit us at nu.edu. You can also follow us on social media. Thanks for listening.