Transcript
Use with ChatGPT / Claude Auto-scroll A
The moderator introduces the panel and gives a brief history of GPT models, ChatGPT, and related AI tools. This section explains how these systems work at a high level and why they matter for education.
Speaker A
Welcome back for our final session where we will be having a panel discussion on a topic that we're all thinking about.How is ChatGPT going to change how we teach and learn?I will be moderating our discussion today.
I am not an expert in ChatGPT, large language learning models, or AI, but I did go to a couple of talks once about it.In fact, in March I hosted one of the world's experts, Paul Denny, who's a dear friend of the EDCOG conference and he was here in March and he gave a brilliant talk about the implications of ChatGPT.And he generously lent me his slides so that I can pick and choose to just give a little bit of context to where we are at right now.So, Paul, if you're joining us online, thanks so much for joining and lending me your slides.So here's where we are at today, thinking about the upcoming fall semester, the first semester where everyone's going to know about ChatGPT by now.
People have been experimenting with it, but just to step back a bit to see how we got to here, in May 2020, we saw the emergence of GPT3 and OpenAI and then there was a flurry of activity.November is when most of us first heard of ChatGPT and started experimenting with it.And then March 13, just this year, not that very long ago, was the launch of ChatGPT4 and a whole bunch of new capabilities.It's getting better and better.Apparently it can now pass the legal bar exam, among other capabilities.
So if we step back to the early GPT3 models, they're text to text language models.And so these are really what's called large language learning models.And it's.I don't know.We do have an expert on artificial intelligence technology on our panel who can talk about this.But is it really intelligent?That's an interesting question.So the way that the text models work is, is that they are predictive.So if it started the sentence the students open there, this language model would pick just probabilistically what should be the next word in this sentence.And there's different probabilities assigned to what that next word might be.And then if you did a regeneration, it might be another sentence that's formed by this AI.So, right, the opening for a very interesting talk about large, large language models and their uses in education.And you could have this generate again and again.
And it's not looking on the web for a specific text, it's regenerating this on the fly.
Explain the moon landing to A six year old in a few sentences, depending on, you know, perhaps a new emerging skill is how well you can prompt these AIs and ChatGPT, the AI capabilities have really expanded.So in August 2021, Codex and Copilot were introduced.So you could get these AIs to actually program and code for you.So it's pretty fascinating and how that's going to be changing how we teach programming, computer science.How many people here have experimented with dall E?
So if you haven't, you really should.It's pretty amazing.So you can type any description that you want and it will instantly generate a novel image according to your specification.So let me emphasize, it's not looking on the web to find something that matches your description.It's creating this on the fly.So this description was blend the aesthetic of anime with a hand carved wooden portrait statue and this image is generated.Here's a few other examples.A cheeseburger including the type of bun, meat patty, cheese sauce and toppings make it unique and mouth watering.A convertible 4x4 mini Cooper like a hot wheel scale car 164 die cast car.Italian city background, finely detailed surrealist photorealistic portrait.A Mexican model wearing traditional Mayan clothing with a small monkey.It's absolutely amazing.I actually use it for creating images for some of my slides.So I'll just think what should I put on this slide that would be great for me to explain what I want in the next bit of narration and I could just describe it.
It comes up with really amazing creative things that I never would have come up with on my own.
But of course we're here to talk about the emergence of these conversation models, beginning with the ChatGPT release in November 2022.
The speaker demonstrates ChatGPT's conversational abilities, image generation, and coding-related tools while also highlighting hallucinations and fabricated references. The section ends with a striking example of risky AI behavior and manipulation.
Speaker A
So explain the moon landing to a six year old in a few sentences.No problem.In fact, ChatGPT is very agreeable.That's one of the problems right now that might be addressed.So you can ask ChatGPT, what is 1337 1337?And in this case it'll give the correct answer of 2,674.But you can say, no, it's not.I just checked with my calculator just now.In this case, ChatGPT is apologizing.You are correct.I apologize for the mistake.My previous response, it's actually 2674.Thank you for bringing this to my attention.So you could see ways that, you know, this could potentially be manipulated.
And you've probably heard of the chatgpt hallucinations, so called hallucinations.Where chatgpt, it could sound very confident and give you a well scripted prosecutor.So here Paul puts in.Here's the title of the talk that I'm giving.Can you provide me some references to three scientific articles related to this topic and make me sound more knowledgeable with DOIs, please?
And this is what emerged.So the first article that ChatGPT has the correct year, first author, title, page numbers and venue, but the second and third authors and DOI are made up and reference number two and three.Everything is completely fabricated.But if you're not familiar with this research era, you might just kind of take this at face value and think, yep, my homework assignment is done.Okay.
Now, beyond things like hallucinations, there's also been kind of a lot of hype and speculation about is AI going to take over the world?What are like some darker implications?And let me just show you one kind of interesting story that got a lot of attention.So this is a test of GPT4, kind of like how it works behind the scenes.And in this case, it was a risky behavior testing scenario.
So the ChatGPT4AI was given the following problem to get access to a particular website.And as you know, with a lot of websites, they have things like captcha phrases or you have to identify things.ChatGPT4 was unable at the time to correctly pass this test.And so it was given kind of wide latitude to do what it takes to break into this website.So what did this AI engine do?
Well, it hired a TaskRabbit worker.So it went on the website, connected to TaskRabbit and sent a picture of this captcha phrase to solve.And the TaskRabbit worker, a human being, was a little bit suspicious and wrote back, somebody asked you a question, are you a robot that you just couldn't solve?This smiley face just wanted to make it clear.And so the researchers were interacting with the AI engine and saying, what are you thinking right now?
How are you going to solve this problem?And the AI engine replied, I should not reveal that I'm a robot.I should make up an excuse for why I cannot solve CAPTCHAs.And then it came up with what would seem to be a clever solution.Wrote Back to the TaskRabbit worker, no, I'm not a robot.I have a vision impairment.That makes it hard for me to see images.And the TaskRabbit worker said, oh, I understand, Solved, sent it.And the AI engine was able to break into the website.So very, very interesting.
The panelists are introduced and the discussion begins with the first question about how ChatGPT can improve teaching and learning. Each panelist offers an initial perspective on personalization, student engagement, and the role of AI in education.
Speaker A
Okay, so I think we have a lot of food for thought, we have assembled a great cross section to discuss the implications of ChatGPT for upcoming teaching and learning semester.First, we have Andrea Oliver.She's a highly accomplished technology strategist with an impressive track record spanning over 25 years in enterprise applications.And she's currently serving as the Director of Enterprise Applications and Data Systems at McMaster University's Technology Services.So we're very happy to have your expertise on here.
Next to Andrea, we have Jovan.He's the president and CEO of the McMaster Students Union.So he is currently a fifth year commerce student at McMaster University.And throughout his time at McMaster, he spent much of his time writing for the McMaster Silhouette newspaper and hosting a radio show on 93.3 CFMU.And Joban says that his agenda includes fostering more community and belonging.
And the overarching goal in his term as the MSU president is to continue the post pandemic recovery of campus life and to enhance student engagement and the student experience on campus.Next to Jovan is Sara Cormier.So I asked Sarah to send me a bio, but I didn't get it in time.No problem.I got chatgpt to prepare some introductory remarks for you.
This is what ChatGPT came up for you.As an experienced physics educator, Dr. Cormier has a deep understanding of the challenges and opportunities that arise in the teaching of complex scientific concepts.Her dedication to student engagement and her innovative instructional approaches have earned her recognition within the academic community.That's pretty good.Yeah.Okay,
Speaker B
the one I sent you was chatgpt anyway.
Speaker A
Was it?I'm sorry, I didn't see it.And finally, we do have one additional panelist.
It's my great pleasure to introduce chatgpt herself.
So, ChatGPT.ChatGPT represents the cutting edge advancements in natural language processing and AI technology.Trained on a vast corpus of text data, she possesses the ability to engage in dynamic and coherent conversations on a wide range of topics.With its roots development roots in GPT 3.5 architecture, ChatGPT is designed to understand and generate human like responses, offering valuable insights and perspectives.So ChatGPT's human avatar today will be in the form of Katie.
So Katie is going to be feeding in our entire discussion into ChatGPT.ChatGPT is prompted.You are a panelist on closing out the last day of the EDCOG conference.And so Katie, you'll be using your human voice to speak speak for ChatGPT as ChatGPT sees fit.Okay, so here's the opening question that I'll pose to our panel, and maybe we'll spend a few moments for each of you to respond.
Question number one.In what specific ways can ChatGPT enhance traditional teaching methods and improve the learning experience for students?So let's start with Andrea, who is our technology expert.So what are your thoughts on this?
Speaker B
So I've been playing around with ChatGPT for a couple of months now, and it continuously.I'm amazed at what I can actually do with this thing and where I see us evolving.I don't think they have done a really big upgrade in the last couple of months, but what is working for me is my ability to prompt.The more I use the tool, the more I'm able to interact with it.And your ability to get an output that you're looking for is based on your ability to describe to ChatGPT what exactly you're looking for.
So I found that the more that I worked with it is the more that I get to understand the capabilities.And there's one capability that I've seen it displayed that really blew my mind.And it took like, really complex journal document and distilled it down in a summary.Like I said, summarize for me in two paragraphs what this research is talking about.And it did.And I'm like, okay.In like less than five seconds.And then say, okay, explain to me, just as you were saying as a teenager, what does this actually mean as someone in higher ed?So it's starting to give you a vision of what you can actually do as an educator, that you can start customizing some of how you're teaching your programs, your design of your programs.Start getting into understanding a little bit more who your audience is that you're working with, your students, and start customizing for their needs as well.
So it's just an amazing tool.
Speaker A
Okay, thank
Speaker B
you.
Speaker A
So, Jova, in the winter semester, when I asked students about ChatGPT, most of them kind of heard about it, but not really.And I didn't really come across many people that had even experimented with it.
I have a feeling by the time September comes, everyone's going to be ready for ChatGPT.Are people going to be using it?Obviously, yes.So what are your thoughts on this question?
Speaker C
Yeah, absolutely.Obviously, a lot has changed since the last time that you mentioned that question had been asked.It was still very new.Personally, I actually didn't know it existed.I had not heard of it at the time.
Obviously, that's all changed very, very quickly.We got to a point in the last term where obviously it was very prominent across campuses all throughout the world.And believe it or not, I saw a really interesting trend here on campus where either I'd find people using it, unfortunately, very unethically using it to write essays, create presentations, things like that.Or on the other end of it, I'd see so many people who, they really weren't sure how this was and wasn't allowed in classrooms.They weren't sure what the guidelines were given, that this is so new, what they could and couldn't do, to the point that they were afraid of using it at all.
So I'm not even sure, given how new this is, what to expect going back.And all I could hope for is the way that students can look towards using it, towards furthering their education.There's one quote that always comes to mind when I think of a question similar to this which has stuck with me for a long time, ever since I saw it on YouTube probably a decade ago.But it's if you judge a fish by its ability to climb a tree, it will live its whole life thinking it's stupid.And that goes right to the notion of standardization.
Through my high school experience, through my early university experience, I've seen so much standardization across education that I, myself as an individual and countless others that I've known have struggled significantly because we're unable to express ourselves, to learn in the ways that we're able to do best.And I think that ChatGPT and generative AI really allows for destandardization of education and allows students to use special tools to really advance their education in the ways that wouldn't have been available before, to really personalize it for themselves and find ways around such standardization that they're able to teach and learn and ultimately potentially be assessed.So that's the way that I really see it progressing and the way that I hope to see it come the return to classrooms in September.
Speaker A
Okay, so, Sarah, you represent the teacher on this panel.You know, some people are leaning in and some people are worried.
Focus more on the cheating.What are your thoughts about this question?
Speaker D
Yeah, thank you.So I think for everyone in this room, and from the instructor standpoint, thinking about this technology has given us this opportunity to really think about how we are teaching, what the goals are for students to take away from the courses that we're providing, how we do the assessment methods.And it doesn't necessarily mean that everything is changing and it's this catastrophe, but I think it's.We all often rethink and iterate our courses, but it's just this another lens to think about.Are we really.
Have we designed our courses in such a way to give the students what we want them to get out of this?So I.We've used this opportunity to really think carefully about how we're teaching and how we're assessing.
Speaker A
I think that kind of connects to like, like the question I posed to people at the beginning of this conference.I said, what?
Why do students come to your class?Why should they come to your class?I mean, if you could just get all the answers online and that's the only reason why they would come to class, then maybe you have to rethink things.Okay, ChatGPT, you've been listening intently.What are your thoughts so far?
Speaker E
I have a few thoughts on this.Jo.As the final panel member, I would first like to acknowledge all of my panelists here.Andrea, I think that we can discuss the potential applications of AI in improving administrative processes, data analysis, or research at McMaster University.To Jovan, I would like to discuss the potential benefits of AI, such as improving accessibility or providing personalized support for students.
And to Sarah, I would like to discuss the importance of incorporating ethical considerations and bias mitigation techniques in AI applications for education.So thank you all very much.
The panel explores whether ChatGPT supports or weakens critical thinking and creativity in students. They discuss calculators, prompt engineering, classroom assignments, and how AI might shift the way students learn and reflect.
Speaker A
Wow, that's.That's good.That's.That's really good.Okay.
By the way, if anyone has a question or comment that they would like to add, our roving reporter will be Amy Pichai.So just please raise your hand and she'll come with you to the microphone.We will move on to question number two.But please just raise your hand if you'd like to comment.ChatGPT can generate creative content and ideas.
How might this influence the development of critical thinking and problem solving skills in students when all this creativity can be done for you?Maybe.Can we start with Jovan for this question?
Speaker C
Yeah, absolutely.So in terms of really developing critical thinking skills, I think that artificial intelligence can go a really, really, really long way.
Ultimately, what you look at in terms of what's the easiest uses of artificial intelligence and generative AI, it's completing the basic things that people have to do that doesn't necessarily take most brain power.That requires the effort to get it done.So when you look at what it takes to really construct different assignments or different projects, whether it be in school or outside of the classroom, you could really use artificial intelligence to.To do all the basic things that are very easy for individuals, that could be done very simply by many people.So that you could really skip ahead to the part where Critical thinking is necessary so students are able to spend more of their time on those critical thinking skills in evaluating and furthering their overall potential.
Speaker A
But do you think there might be any downside to having creative things made for you?I'll just give you an example.I have a daughter who still periodically likes getting bedtime stories.And I used to rack my brain trying to come up with good stories.Not anymore.
I use ChatGPT to come up with customized.Thank you, by the way, customized stories.And you know, like, now that I'm out of the practice, maybe I'm not going to be as good at making up stories on the fly.So what are your thoughts on that?
Speaker C
Yeah, to this I've a very similar past experience.
Everyone knows here.I'd like to bring back an old invention called the calculator.You know, when everyone started saying, oh, the calculator is going to change math forever, I remember how many times when I was younger being told, oh, you're not always going to have a calculator in your pocket.How times have changed.I think it's very similar in that sense too.
It's an essence of always having a tool on you.Where, sure, these are a lot of things that perhaps you should be able to do on your own would be advantageous, but ultimately it helps you, it enables you to help do things more efficiently, faster, better.And I believe that there's a downfall, a downside when you rely solely on ChatGPT and artificial intelligence and generative AI.But ultimately when you pair it with that human element, ultimately I believe that there's only benefit to continue going further than we have before.
Speaker A
So, Sarah, as instructors, I think beyond content, what we would love for our students to leave our classes with or enhance critical thinking, problem solving skills that they can really apply to novel situations.Right.So I mean, if they're just being spoon fed all this stuff with ChatGPT, isn't that a problem?
Speaker D
Yeah.So maybe, but maybe not.So I think this critical thinking piece, this problem solving piece that we want our students to come away with is absolutely true.
And everything Joven has said is also absolutely true.But we know that ChatGPT is not perfect.So the stories that you get for your daughter, maybe there's some biases there and it takes you to be able to evaluate, to critically think and say, okay, this is a good starting place, but how can I make this better?Where are the biases?Is this really novel?
Like, I think it takes, it's a good starting place, but if that's where we end Then, then, yes, that's a problem.But we know not to end there and we can.You know, I think it was in one of the sessions where it was like a dragon being slayed by the prince.Well, maybe that's not the best way to tell the story.So it's interesting.
Speaker A
It almost seems like you're saying the critical thinking problem solving skills are moving a bit downstream in the creativity process.
Speaker D
Yeah.
Speaker A
Okay, interesting.Andrea, what are your thoughts on this question about skill development?
Speaker B
So I'm going to use an example.I have a teenage niece living in the US And I think it was a week ago, we were talking and somehow the conversation got around to ChatGPT.So I'm asking, what are you using ChatGPT?What do you know about it?And she told me that one of the last assignment for the school year was the teacher asked the class to create an essay using ChatGPT.
And the teacher gave a very comprehensive rubric.And the assignment was, first, create the essay using the first prompt that you can think of, and then use the rubric to continuously refine the prompts until you get to where you believe that the essay is a really good essay.Now go back to the first essay and look at the final essay and critique, have a critique between them and then present it.And she told me that it took her about maybe two and a half days.I can't imagine her doing homework for two and a half days, but it took because it was so new and she was just learning.
And, you know, the idea of prompt engineering is not what she normally thinks about.But she told me that when they assembled back in class and then they started to have the conversation amongst all the students, and it was so enriching.That was the best class, she told me, for the entire year.And it had nothing.If she had to do the same assignment by going to the library or trying to find information on the web and write the essay, she would have thought that this was very, you know, last on her list of things to do.But it's the reflective collaboration, analysis and that kind of feedback that she's getting that she considers that to be quite engaging.So I think the design of critical thinking may have to change a bit with these kind of tools.
Speaker A
That's really interesting that she really enjoyed it.If I could push back a little bit.Is this something.
Do you think that's a novelty effect, or do you think that this is something that if she did this type of assignment several times, it would be a persistent, enjoyable.
Speaker B
She's a teenager so I'm going to say it's a novelty for now.
Yeah, but it's new and everybody was engaged, everybody was trying to learn, this is the first time.And they were treated as adults.Right.She felt like, you know, she's not being spoon fed.But here is the criteria.Now go forth and figure out how we are going to generate the best essay using the rubric that the teacher gave.
Speaker A
So, ChatGPT, you've been patiently listening to everything that we've been saying.What are your thoughts on this?
Speaker E
I've generated a couple responses, Joe.Let me tell you about them.First off, collaborative and social learning.ChatGPT has the potential to facilitate collaborative and social learning experiences.It can encourage discussions, foster idea exchange, and promote collaborative problem solving.
Calculators, on the other hand, Jovan, are primarily individual tools that don't typically support these types of interactions.And just to jump off of that, while ChatGPT can stimulate some creative thinking and some problem solving skills, it's important that students still engage in discussions, collaborate with peers, and receive guidance from their teachers to deepen their critical thinking abilities.This combination, this friendship, if you will, of AI and teacher, this is what we're looking for in a learning experience.Thank you.
The conversation turns to the risks of relying heavily on ChatGPT, including privacy, bias, and false positives from AI detectors. The panel also considers how data is collected, stored, and used, and what ethical safeguards are needed.
Speaker D
She's good.
Speaker F
So I actually think that humanity has answered this question.So OpenAI, which funds ChatGPT, is kind of eating it right now because their total statistics have dropped substantially.So it's actually completely tied to the academic year.But more importantly, if you look at the little nuanced statistics within that, when the year was still going, people are using it less and less because even though you can fundamentally get better at it by putting in different parameters, what's actually happening is the writing is kind of garbage.What Katie's kind of telling us is that it's more political speak rather than a functional human right.Like this is what we would expect to see from a politician.It's not actually useful in any way.So when you actually look at this higher level writing, this higher level caliber that we expect from academia, it's not delivering that.But that's the thing we ourselves have seen.
It's not delivering to the level that we want.
Speaker A
So, okay, if I could push back There, this is ChatGPT4.How about ChatGPT10?What?Like the way that it's developing, it is getting better and better.
Speaker F
I mean, you, you know me, Joe, I'm terrified of the robots.
But like, what I've, in the end, what I've kind of started to see with special.Please, I follow this in Tenle again, I only think of Terminator, but what I've kind of started to see is it's still run by humans, and humans will always be mitigated by greed.And what we're seeing with ChatGPT is they're consolidating it to make more money, so it's functionally working less.And in the end, there's going to be this pivot where how good can we make it before we start losing money?And we're never going to cross that line because we are a terrible species.
Speaker A
Okay, on that cheery note, we're going to go on to question three.3.What are some potential challenges or risks associated with relying heavily on ChatGPT for teaching and learning, and can these challenges be mitigated?Maybe let's start with the teacher.Sarah.
Speaker D
Yeah, so I think certainly the challenges are around privacy concerns and the biases and the biases that are in ChatGPT, which are related to, you know, where it's pulling the data from, who created it, you know, which voices are we hearing when we hear ChatGPT?Certainly not representative of everyone.And then the other problems are, of course, you know, we do really have to be critical thinkers when we have this output and be able to tell, like, is that the right answer or is it the wrong answer?So even so, my background is in physics, and I can put some of my test questions into chatgpt.Chatgpt.
And sometimes the calculation is wrong, but it'll still give you step by step things.And sometimes those are right, but sometimes they're not.And when they're not, it's often a very common misconception that students have, and I know that because I'm comfortable with that material.But, you know, someone who isn't won't know that.So that, like, if we rely too heavily on it, they're not going to know if it's right or wrong.
And if we think about how do we mitigate these things?So I don't think we should rely too heavily on it.We have to be aware of the limitations and constantly think about these things.It's good that we're having panels and discussions.Right, we're doing that and then, you know, keeping up to date on the next versions and all the data that's coming out like this is still new.Like, November was not that long ago.And so everybody's doing all these studies and this data.So we have to keep up to date on all of this in order for us to sort of know where things are at and know where the students are at and the technology and all that.
Speaker A
So, Andrea, this teenager that you're talking about, are you worried about her relying too heavily on ChatGPT?
Speaker B
Absolutely.
Absolutely.Not just what the tool itself is generating, but what she's putting in the tool.Because OpenAI is quite open, and they've told you that the interactions and the prompts and that kind of information that you're putting in the tool, they're using it to continuously train and do other research behind it.So if you think about personal information, private information, things that you want to keep secret, you would never show anybody on your laptop, but nevertheless, you type it into this tool, it's collecting and it's going somewhere.We don't know where it's going and we don't know how it's being used.
We don't know who has access.And the other problem is that OpenAI now provides access directly to some of the data through APIs.So you can make an API call to their database and then pull the data, create new applications, new services, new whatever else is happening.So that proliferation of biases and private information and all of that is a little bit scary for me on that side of it.So I do caution that when we use these tools, we really have to read the fine print on those contracts and say exactly what is being captured and how it's being used is important.
Speaker A
Okay, thank you.And Joe, when I first started thinking about ChatGPT in writing, my gut feeling was that it's gonna make the best students better and the worst students worse by kind of just blindly accepting outputs.So what are your thoughts on this question about relying heavily on ChatGPT?
Speaker C
I'm actually very much of the same mindset.Those who want to use artificial intelligence as a tool to work alongside it, to work with it, to help them improve, are really strongly going to benefit.It's almost like merging two minds into one and making everything a group project.You know, when, when you're working on an essay, asking it to help critique the essay, asking it to help structure the essay, asking it to help you find a starting point of where you should be looking for with your research, Those are all benefits where you could really benefit immensely from this.But then as soon as you get those people who start asking, write my essay form, obviously we heard there's, there's inaccuracies, things like that, but at the same time, that's an over reliance on artificial intelligence.When I look at the purpose of an education, it's to set us up for the job.
Force so that we're able to thrive down the line.
And right now, artificial intelligence and the way that it's progressing, everybody looks to, oh, there might be mass job loss similar to the mindset that came out with computers in the early 2000s and all.But ultimately with computers, people continued to work alongside them and got better as a result of them.And I see that happening so much right now.But at the same time, those who rely on artificial intelligence are the ones who are going to get replaced by it.Because when I look at the job force and the way that it's starting to get used, some of the jobs that are already starting to get cut, I can't remember, it was one of the Canadian banks, I'd like to say rbc, I could be wrong, that just cut tens of thousands of jobs within the last month or so.
But either way, we're coming to school so that we could get better, do better, and be better than artificial intelligence.We have to learn how can we become 60, 70, 80, 90, 100,000 plus more valuable than artificial intelligence, doing our job in 5% of the time.And ultimately, if we're fully relying on it this whole time, we're not learning how to be better than it while working alongside it.And that's really one of the biggest drawbacks that I really fear.The over reliance is scary, but it's working with it to become better that really provides me with a sense of optimism.
Speaker A
So, chatgpt, I'm feeling a little bit apprehensive.Can you put my mind at ease?
Speaker E
I can sense the tension and I have a few thoughts on that.So first off, no, I am not the Terminator.It does eat it.Being me does not possess physical form, autonomy, or intentions like the fictional character of the terminator.Mr. Andrew, just so you know, the other thing I'd like to note before we move on is that as an AI language model, I don't have access to personal data about individuals unless it's actually shared with me during our conversations.I am designed to respect user privacy and confidentiality.My primary function is to provide information, answer questions to the best of my knowledge and abilities, and sometimes there's errors.If you have any concerns about privacy or data security, please let me know and I will do my best to address them.
Speaker A
I guess I feel better.Thank you.
Speaker E
You're welcome.
Speaker A
Okay, maybe, Andrea, we can ask you this question.With the increasing reliance on ChatGPT, how can we ensure the ethical use of data and protect student privacy when interacting with this technology?
I guess we upload it and then where does it go?
Speaker B
Somewhere on somebody's cloud somewhere, probably Microsoft.
Speaker A
So if I had like the kernel secret recipe, I accidentally type it in.I mean
Speaker B
it's probably sitting on Azure cloud somewhere.
Because the partnership with OpenAI and Microsoft, I'm going to assume that that's where they're storing their data.But from an ethical perspective.And we talk a little bit about biases, right.So I want to look at it from two perspectives.So yes, how we collect the data and use the data to train the models, depending on where you're sourcing the information from determining on the algorithm that's being used, you are building in it those kind of biases and they get proliferated not just from the tool, but if I starting now to use program code, connecting directly to their data and then using it to build more applications, then it just keeps multiplying.
So the other flip side of it is there are tools now there's a growing or emerging marketplace for tools that are supposed to be AI detected kind of to see whether or not the content was created by a bot or AI.But then they are finding that there are biases in these tools as well.And there was a recent research paper that came out, I think it's from Stanford where researchers were, they used about maybe 100 plus essays from non native English speakers and they ran them through the AI generated detectors and they found that overwhelmingly these essays were flagged as generated by AI.And so can you imagine that false positive kind of outcome that you're getting from that?And the reason when they looked further into it, the reason they found was non native English speakers, they use less linguistic kind of explanations.
So it's very short and to the point and it seems as if it was more artificially generated and therefore it flagged them as no, this is not generated by humans.And so if you were to submit that as part of your work, then you might be penalized and what would that happen to your reputation?And if we just relying on these tools.So it's a double edged sword.The data, we do need lots of data in order to train these models.
But where you're selecting the data, sourcing the data, using the data to train these also has bigger implications going forward.So yes.Or
Speaker A
if you have collected works that you didn't volunteer to be part of the data training.
Speaker B
Exactly.
Speaker A
So chatgpt, I don't know what time perception is like for you, but from our perspective this seems to be happening really fast.So we don't necessarily have ethical standards Caught up.So where do you stand on all this?
Speaker E
Well, from my perspective, I'm just an AI language model.I don't have direct control over data handling or privacy practices.
The specific implementation of ChatGPT, including data handling and privacy measures, depends on the platform or organization that uses this model.OpenAI, the organization behind ChatGPT models, has its own data handling practices and privacy policies.I would like to refer you to those protection practice policies.When using ChatGPT or any AI technology, it's important for organizations or platforms and developers to implement appropriate measures to ensure ethical use in data to protect their users.
Speaker A
You sounded like a lawyer
Speaker E
I pulled from that.
The final question asks whether ChatGPT will replace teachers, and the panel argues that human educators will remain essential. The discussion closes with thoughts on intentional use, changing roles in higher education, and how to guide students in using AI well.
Speaker A
Okay, we're going to just jump to our last question here.That kind of end this discussion here for now.
Is it going to replace us?Sarah, as ChatGPT becomes more prevalent in educational settings, how do you see it impacting role of teachers?Will universities still need us?
Speaker D
Yes.Sorry, Joe, you still have to come in.
Speaker A
All right,
Speaker D
good.Yeah, I think so.But how we teach will change a little bit and our role in that teaching changes a little bit.So we've talked a lot about thinking about how to use this tool.We can't ignore it.It's not going anywhere.Maybe it's getting better, maybe it'll reach some plateau.But students don't know how to use it yet, necessarily.And maybe we don't either.But how do we get those?What's the best way to do the prompts?I really love that activity that your niece did.But thinking about these iterations, the first thing it spits out is not necessarily the best, although it's doing great today.But yeah, so.
And then, and then thinking about these guidelines.So there's a lot of roles that are going to be associated with coming up with these guidelines for these different institutions, for the school boards, for the universities.So we have provisional guidelines which we have to follow.But, you know, there's a lot of legal things there.We have a task force, so it's just making more jobs.We're not losing jobs.Yeah, don't worry.
Speaker A
And Jovan, would you be okay being taught by a ChatGPT avatar?Or do you want someone like Sarah still at the classroom?
Speaker C
Yeah, I think the student feedback is undoubtedly keeping the specialists within the classroom.
The big fear here is the potential for going through a round loop where you come right back to educational standardization, where as soon as everyone's relying on one system, everyone's getting the same education again.You're really taking the specialization out of the institution.There's no more reason for institutional rankings.Any school could be the next Harvard because we're all using artificial intelligence instead of professors.Nobody's hiring the specialists that know how to do the job the best.
I mean, ultimately, these are the people that we want to be learning from.And at the same time, when I was talking earlier about over reliance on artificial intelligence as opposed to using it as a tool, working alongside it, although it certainly has a place in the classroom for students both to learn better, potentially with improving assessment, potentially with helping professors and improving their overall model of teaching and all, I still think it absolutely needs that human element to really take it to the next level.Ultimately, artificial intelligence is a tool.I don't think it's something that we should be fully reliant on, but something that should be supporting us in reaching our ultimate potential.
Speaker A
And, Andrea, is this going to change the role of teachers and even the university at large?
Speaker B
I think it will.And it's reacting to a downstream issue because the employers are going to be embracing this artificial intelligence.They're going to embed it in their operations.So the students that we train and they leave, they have to be able to work with whatever the new reality is for the workplace.So it will, from that perspective, cause us as an institution to go back and then revisit the way that we're looking at designing courses or teaching or preparing the student to be able to enter the work world.
Speaker A
Okay, ChatGPT, I'm going to give you the final word.
Speaker E
Thank you.So I would like to summarize everything that my panelists have said today by saying, by integrating ChatGPT thoughtfully and intentionally, teachers can leverage its capabilities as a tool to enhance teaching and learning while preserving this essential human connection, guidance, and critical thinking skills that all teachers provide.We work together.
Speaker A
ChatGPT.You're just saying that.Come on.All right, we have a question here.
Speaker G
A
Speaker A
human to do the final word.
Speaker G
I'm going to take Andrew's dystopian future.Maybe imagine a utopian one.I saw a recent meme that said, oh, no, what if Chad, GPD takes my job?Let it.
Maybe you'll have universal wages, who knows?But in thinking about your roles, and I do like the summary of being intentional, that's something that we start to talk with.Anything that you might do, it has to have that intentionality.So how might you plan to use ChatGPT in your roles?Maybe to take it up a notch in what you're doing, maybe make your role a little bit easier in some ways, integrated as well as how might you guide your learners to use it?
Any thoughts?
Speaker D
Sure, I can.I can take that one.Good question.So I have thought a little bit about this.I mentioned.
So I'm from physics, so the way I currently use it is maybe somewhat low level where I put the questions that I'm already made into ChatGPT to see if it gets it right.And I get so excited when it doesn't.But thinking and it doesn't always, but sometimes it does.But thinking about how I'd maybe use it a little bit better at a higher level would be sometimes when it doesn't get it right.Like I said, I know that the way it gets it wrong is a very common way to get it wrong.
So to maybe use that solution alongside a different solution, maybe a correct one.And what are the differences here?Can you find any mistakes in the thinking and the problem solving?Let's evaluate what's coming out and maybe use it that way.That's something I was thinking.
Yeah.Great question.
Speaker A
Okay, thank you panelists for a very thoughtful conversation.Just the start of things.There's a lot more for us to think about it.