Creating Interactive eLearning using ChatGPT w/Garima Gupta
Episode Overview
In today's episode, our host Matt Gjertsen sits down with guest Garima Gupta to discuss the fascinating world of creating interactive eLearning using Chat GPT. They dive into the machine learning algorithm behind the ChatGPT OpenAI API, the importance of prompt engineering, and how to have deep exchanges with AI. Join us as we uncover the potential of ChatGPT in revolutionizing the eLearning landscape. Garima sheds light on topics utilizing the transformative power of ChatGPT.
Let's dive in!
About Garima Gupta
Garima Gupta is the Founder and CEO of the award-winning Learning design firm Artha Learning. Garima brings over 18 years of diverse experience to the field of Learning & Development and is passionate about the intersection of technology and education. Currently, she is doing some amazing work integrating ChatGPT into Storyline courses which more people need to know about.
Full Transcript
-
Garima Gupta [00:00:00]:
The most basic standard thing is you have your learner ask any question within the Elearning module and Chat GPT replies. Now it can reply as an expert on the topic that is being presented, it can reply as their manager, it can reply as their peer. So those kind of roles, but you could create that live conversation where somebody is asking something and getting a response.
Matt Gjertsen [00:00:25]:
Welcome to Making Better, a podcast from Better Everyday studios devoted to helping small learning teams have a big impact. Today I am talking to Garima Gupta, the founder and CEO of the award winning learning design firm Arthur Learning. Garima brings over 18 years of diverse experience to the field of learning and development and is passionate about the intersection of technology and education. Currently she's doing some amazing work integrating Chat GPT into storyline courses, which more people need to know about. So let's dive in. Garima, how are you doing today? Welcome to the making. Better podcast.
Garima Gupta [00:00:59]:
Thank you. Happy to be here. And thank you for inviting me to talk to you and your listeners.
Matt Gjertsen [00:01:04]:
Oh, absolutely. I mean, I think this could be a really great and important conversation for our listeners because for me, as somebody who's been making Elearning, worked in instructional design, worked in all things L and D, I think one of the critical things that always comes up around Elearning is the lack of real learner interaction, right? Where when it comes to we want to test whether or not people actually understand what we're talking about. And the best we often get left with is some form of multiple choice question and it leaves so much to be desired.
Garima Gupta [00:01:47]:
Absolutely. And we can disguise those multiple choice questions in fancy garb. They look like scenarios and whatnot, but they're all pre programmed. You have to select pre made choices and you get pre made responses. And there was nothing we could do about it till now. But now game is changing. I am super excited about it, just like you are. And I think that we have future of Elearning standing right in front of us.
Garima Gupta [00:02:19]:
It's time for us to grab it.
Matt Gjertsen [00:02:23]:
Yes, exactly. And I think that's 100% right. And you're going to talk us through what I love about what you're going to talk to us about today. In a minute we'll do introductions, but just to give a preview, is it's not some new fancy app that people need to go get right? It's something that's built in to the most popular learning development tools that we use every day. So it's super exciting. Before we get there, you have a different background than anybody that I've talked to in terms of your pathway into Elearning or into learning and development. You kind of came from software, if I'm right. So tell us a little bit about your background and how you transition into learning development because I think it's really relevant to how you're able to do what you're doing right now
Garima Gupta [00:03:06]:
My pleasure. And it's a long story, but I'm going to try and keep it short. Yeah. I trained as a software engineer. In fact, I was part of the team that coded and built the very first short message service center in India. Long time back. I'm not going to date myself, but what I found always was I liked talking about and demonstrating my code rather than coding itself. And I liked to see that immediate impact.
Garima Gupta [00:03:34]:
And I missed that in the coding world because your code is a little part of a bigger machine that goes in a bigger machine that goes in another bigger machine. You never get to see what your code really did. And I always found that I was a storyteller. So if you would ask me, at the bottom of my heart, I feel like I'm a storyteller. So that was it. Well, I was doing okay in my coding job, but as I started a family and had kids and spent more time with them, I was fascinated by the act of learning in children. When kids learn young, young kids, it is so pure, so natural. And I just loved it.
Garima Gupta [00:04:20]:
And I thought that my natural fit was to live and work as a catalyst to help someone learn something.
Matt Gjertsen [00:04:27]:
I love that phrase.
Garima Gupta [00:04:29]:
Exactly. That sort of made me move career. So I went into K 212 because I thought I was good with kids, but I did not sustain that. I did not have the infinite patience that it needs. So I moved to adult learning. And that was beautiful. I think I felt like everything was preparing me over my life to get into adult learning facilitation, then moved to Elearning, which was, as a coder, a natural shift for me. I saw the scale of digital learning, the impact that good, well designed digital learning can have on scale, on motivated learners.
Garima Gupta [00:05:11]:
And then six years ago, started my own firm because I found that this really is my jam, I guess.
Matt Gjertsen [00:05:23]:
Excellent. I love it. Yeah. And that's the thing that I love. That's such a great way to put it. This is our jam. And it's kind of quirky and a lot of people don't understand it, but those of us in the industry, it's our jam. I just love it
Matt Gjertsen [00:05:37]:
It's great. That's awesome. Thank you for that background, that backstory. I think that was great. And I love that you highlighted storytelling, because at the heart of all of this, I think that's such a critical thing. Really, what we're doing is telling stories to help people learn. Like, that's really what we're trying to do.
Garima Gupta [00:05:59]:
And you know what in the middle of it, if you lose sight of that, then your stuff comes out dry. Then it is not impactful. So, yeah, you're looking at your tools and your triggers and variables and slide design and graphics and all of that, but the soul is in the story you're telling.
Matt Gjertsen [00:06:19]:
Absolutely. I think that's totally right. Awesome. Okay, so then six years ago, you started your own firm. You had that background in software. We're going to be talking today a lot about how you're utilizing well, first, how about what is the solution that you created that got us talking?
Garima Gupta [00:06:38]:
What did you do for the learners? Matt and I got talking because he heard about something that we've been doing in this space and there is a lot we can do in Elearning. There is amazing, well made impactful Elearning out there. But people who are in the field, they feel that there are limitations that naturally are imposed on us. We cannot have live content, we cannot give proper feedback on open text responses. Those kind of we can't have really authentic assessments without the help of a human instructor. So when Chat GPT came out, me and my team started looking into it. I naturally, somehow my inner software engineer woke up. So we sort of started there you go, doing a lot of experiments and that became just so exciting for us.
Garima Gupta [00:07:34]:
It sort of sorry, consumed me in some sense. We created different use cases on how this would change Elearning and there are other people too in the field who are doing work on this. That was very exciting. But I realized that there is a pretty deep level of technical expertise that's needed that usually you will not know a typical LND department or LND professional might not have that because we are instructional designers, we're not really coders and you need a little bit of that. So we then came up with a solution so that anybody using industry standard tools like articulate storyline or captivator electora or easy generator could connect their modules with Chat GPT and do anything that Chat GPT can do within that context of an Elearning module. So yes, that has been keeping us
Matt Gjertsen [00:08:35]:
Very busy over the last few months, I'm sure. So, just a little bit of backstory. Before that. Had you been doing any work with AI before Chat GPT or did this like for so many of us it was like, oh, this is the killer app. It's just like such a thing to utilize. So was this your first thing or had you worked?
Garima Gupta [00:08:55]:
Definitely. Well, Chat GPT was a game changer, but we had worked on AI powered platforms before, so we used Area Nine Lichm for a project on Amazon AWS Basics for a Japanese materials firm. And that is where the content is present to learner. Depending on what AI decides based on how learner behaves, what are their response patterns. It is powerful stuff. Even when I first saw it, it ensures comprehension, retention. It is basically adaptive learning, but it's like a specific purpose knife. It is great for that use case, but it is rather unreasonable or expensive for most organizations to move to a completely new platform for part of their training needs and generative AI changes.
Garima Gupta [00:09:45]:
That right.
Matt Gjertsen [00:09:47]:
Yeah, I think that's totally right. And being able to build this into our own systems is so critical. And so the solution that you made is something it's a module that it's inside of a storyline course, just like anything else. I know you've done a lot of build out with this. So just to give people an idea of some of the things that they could potentially do with this, what are some of the use cases that you've built out already and how do they absolutely.
Garima Gupta [00:10:17]:
So it's very difficult to explain this in a podcast without showing it, but I'm going to try my best.
Matt Gjertsen [00:10:24]:
Yes.
Garima Gupta [00:10:27]:
The most basic standard thing is you have your learner ask any question within the Elearning module and Chat GPT replies. Now, it can reply as an expert on the topic that is being presented, it can reply as their manager, it can reply as their peer. So those kind of roles, but you could create that live conversation where somebody is asking something and getting a response in in person training we always end trainings from. So any questions? Right. And then you have time for that Q A. That Q A is missing in Elearning that now can be now. The same thing can be used for other things. We have scenarios which are essentially infinite scenarios because instead of choosing from ABC predetermined options, learners could write something and then get feedback on that.
Garima Gupta [00:11:19]:
So you could write something else, you'll get some other feedback, so it becomes an infinite scenario. Right. We've also used it for pre and post assessment and then giving learners a pretty comprehensive sort of report on how they did in the course, which is very personalized. We've used it to do assessment on assay type questions based on a rubric. So we send the rubric, we send the answer and then users actually get a mark that can be then sent to your Scarma LMS or whatever. Right. So you can Pneumatically grade an open text asset type question without use of any live instructor. So many different we've also done another experiment because one of the questions I kept getting was, well, it doesn't know about my policy, my company's policy, right.
Garima Gupta [00:12:20]:
If we are teaching code of conduct yes. If we are saying, okay, you cannot accept gifts over x dollars of value, how will Chad GPT know that? So we've done that experiment as well, where it compares the answer to a given policy or a document and then provides its response. So a lot of different exciting stuff there. The beautiful thing with generative AI is that it is so quite general purpose, right? It's like a multifunction Swiss knife. You could use it in many different ways based on what is your requirement. And the power stays with the instructional designer and the learning specialist, not with somebody who created a specific purpose platform and now you can only use it for that.
Matt Gjertsen [00:13:10]:
Yeah, and I think there's a lot to break down there because I think you're right in that it's tough to see to think about this without a lot of visuals. It can get kind of technical, but correct me if I'm wrong to kind of walk through it a little bit. So when most people use chat GPT, if a random person just uses it, and that's what people are thinking when they're listening to this conversation, they're just seeing, here's a text box that I type into, and how can I have an exchange that's very deep on something that's specific to my organization? And I think there's several layers to how you do that. And one, something that I know you've done is essentially what you're doing is that when the user in the course clicks submit, you're sending a prompt to chat GPT. And one neat thing about chat GPT is prompts can get very long. I have actually written prompts. I didn't write them, I copied and pasted them from somewhere else that were like six pages long, because it just contained all it was basically how to train chat GPT to make really good mid journey, right? And so in order to do that, you have to train chat GPT what mid journey is because it doesn't know what it is, because mid journey is very new. And for anybody who's listening doesn't know, mid journey is one of the more popular new AI image generation tools.
Matt Gjertsen [00:14:51]:
So you have to teach it what mid journey is and what a good prompt is. And in those prompts you can say the type of camera you want to use, the type of exposure, the lighting, all kinds of stuff. And so somebody out there had written this massive document that was just like giving all of these different use cases. And so inside the prompt, you trained chat GPT about this new thing that had never seen before, and now you could get it to act on that information. And so I think how much of your solution involves kind of really good prompt engineering versus is there other stuff.
Garima Gupta [00:15:29]:
Beyond well, prompt engineering, as you correctly said, is the biggest thing, right? Because as powerful as AI is, it's still answering your question. So the better question you ask, the more information and context you give it, the better it's able to answer. So in our experiment, putting chat GPT within e learning modules, we figured that it's not only enough to give Learner question, let's say in A-Q-A sort of application to chat GPT, because honestly, learners will try to break it, they will write funny stuff. So you got to put context around it in the prompt, right? You got to say to Chad GPT, okay, this is the question Learner is asking. It should be about XYZ topic. If it is not, give this answer. If it is, this act as this expert, an answer within these many parameters, maybe this answer in five sentences and answer like a coach or whatever. So prompt is a big piece.
Garima Gupta [00:16:27]:
There are other technical items as well. You could using parameters within Chat GPT define how, for example, creative it should be or not. So those can be adjusted as well. But I think the most important thing definitely is how you're writing your question and then doing a lot of audit. I think the beautiful thing right now is all of us across the board are early adopters, right? So there is a lot of testing going on. One thing that we have put together with our solution is sort of a learner analytics report where you get to see what people are asking and what it is responding with, because you need to know that you're right, what people are asking is important. Well, that's amazing data for any learning professional is what their learners are asking in an Elearning module, but also what AI is answering. And is it in line with what you wanted to answer? Because that is where it becomes a little tricky.
Garima Gupta [00:17:27]:
We are used to having 100% control on content and you are sacrificing that in some ways to bring in more authenticity, more real time conversational generative work within your Elearning module.
Matt Gjertsen [00:17:47]:
Yeah, that makes sense. I think the key thing there is just one to connect the two things on the prompt engineering. So when the learner is in there asking their question or giving their answer, whatever is and they hit submit the tool that's sitting within storyline is sending to Chat GPT a whole bunch of other stuff beyond that they're saying like answer this question or judge this response based on this context. And that's kind of the context that you're sending as the prompt engineering, if you will.
Garima Gupta [00:18:25]:
Totally, yes.
Matt Gjertsen [00:18:27]:
Okay, that makes sense. Awesome. One question that I have that I think a lot of people would have is security, because there's been a lot of different discussion and I think your solution maybe relies because you're actually kind of directly interfacing with Chatgbt via API versus through the GUI, the user interface. Where are we at in terms of security of that stuff that's getting sent to Chat?
Garima Gupta [00:18:57]:
That's a great question because that's something that automatically we're thinking about. It is a machine learning, continuously learning algorithm. It is learning based on all the stuff we are feeding it in the graphical user interface that GOE however, in this case, because we're calling the API Chat GPT OpenAI came up with its policy in March of this year 2023 people listening it to later. So data questions, queries and answers given via API, those are not recorded, those are not used for training and they are deleted after 30 days maximum from their system. So that way using the API is way more secure. Now, we've added another layer of security on that because we have our servers on AWS and they are calling the Chat GPT API, right? So you're not even calling OpenAI API directly because that would expose your API key if somebody was techy enough to use and look into your code backend. So we've sort of put another layer of security on top of it. But anybody who wants to try it, you could try it directly and OpenAI itself does not suggest that you directly use its API key in a JavaScript call, for example.
Garima Gupta [00:20:27]:
And I might be talking gibberish to most of the people here. So my apologies, but yes, to summarize, it is as secure as we can have it at this point of time because we're calling it through an API system which has completely different paid secure setup.
Matt Gjertsen [00:20:49]:
Yeah. And so kind of to put I was going to say as soon as somebody dropped API in JavaScript, a bunch of people were like, well no, I don't know. But I think the point just for the learners or for the listeners and I'm sure there are other people doing this, that's a big thing. So essentially what you're trying to help people with is you know how to do this. You have these modules that can be dropped into Storyline that can call Chat GPT and you can help people, help companies, help individuals do all that prompt engineering, figure out all the stuff that they need to build around and make it more user friendly for them. So they can kind of be if they want to be more technical, they can be, but they can also be super untechnical and just know like, okay, I need to drop in our policy on X here and that's all we need to do. And now we have more than options because honestly, I think even the really simple questions of just like to your point earlier, of what's our gifting policy? Like what's the dollar limit? Simply going from needing to say have options a B and C 5100, 150, versus just asking the learner what's the dollar limit for policy and leave it at that, because even now you could have an exact match answer. But what if somebody types out dollars instead of puts a dollar? There's just too many ways to mess it up.
Matt Gjertsen [00:22:27]:
And so simply moving from specific responses to just an open answer like that right there is huge. So anyway, even if you're not technical.
Garima Gupta [00:22:40]:
It can be done. Exactly. My reasoning and logic behind all of this is if you're technical, get inspired and try it out, absolutely, you can do it. If you aren't, you don't have to be. The beautiful thing with these generative AI is that it is so easy. You just ask in simple English and you don't have to prompt engineering is really like being a manager with a very smart sort of intern and you have to explain them all the context so that they can give you a good answer. So that is what instructional designers and learning specialists anyway, specialize in. They are amazing at it.
Garima Gupta [00:23:27]:
So that is the key skill. You don't have to worry about the technical because my thought process here is just because it is a little bit technical, people will not try it. And that will be a shame because if you ask me, I think every single Elearning module should end at least with one slide saying, what are the questions you still have in your mind? Can we help you with that? And give some sort of answer? Because if you don't answer, people don't want to write a question. They know it's a useless exercise. Right. And just that little thing can now change the human computer interaction. Right. So I feel like it is such a low hanging fruit that we should not as an industry, miss it.
Matt Gjertsen [00:24:20]:
Yeah, no, I definitely agree with that. So stepping back or stepping forward, I guess a little bit from where we're at today. I'd love to ask a quick question as we get to the end here of the future I keep waiting for. I'm trying to think of what I think is probably the most likely way that this is going to happen, but it feels like we're not that far from every company being able to have their own chat GPT in a way where it's a pretrained model that's just sitting there waiting for you to go, and you just kind of dump your company repository into it. And it's now an expert in your company. And it's contained. Right. You're not giving it's an AWS s three container or something.
Matt Gjertsen [00:25:11]:
It's just like sitting there and how far away are we from something like that?
Garima Gupta [00:25:18]:
I could probably pull up a horoscope more easily and it will be better accurate. There is no way to say, but the pace is so fast that I won't be surprised. I do think, though, I believe that AI will very soon become so pervasive, just like Googling is right now, right. We won't even think of it as novel because it is just so easy right now. In my webinars I find 20% to 30% of webinar attendees haven't even used chat GPT. And these are people who are coming to the webinar, so they are actively looking to upskill themselves. So overall I would think maybe 50% to 60% of population haven't tried it. I think that will change quickly and yeah, it'll be the new Google.
Matt Gjertsen [00:26:07]:
Yeah, no, I think that's totally, yeah. I would highly encourage anybody. It's still free. So getting on the basic, just kind of getting on the website and I would just encourage anytime you're going to Google, just like ask chat GPT first, just see what it says. Some things, it doesn't make any sense to do it that way, but sometimes it does. And it's only by using it that you start to develop a mental model for what makes sense for using this for and what doesn't make sense to use this for. And I think your analogy was so spot on of the manager and an intern, because I have a Tesla, and so I'm constantly getting to experiment with self driving and experienced how it's always 18 months away. But when I'm letting it drive, I think it is so the perfect analogy of it's like you're driving with a student driver, it's like a teenager, it's hesitating, it's just making decisions a little bit slower than you would.
Matt Gjertsen [00:27:23]:
And so I think when it comes to AI, we often think about like, oh, it's a super intelligent Joe. It's able to do all this stuff and it's like it can, but it just doesn't have the experience. It's the most skilled person with no experience.
Garima Gupta [00:27:38]:
I like to tell people that when you're using AI. Absolutely. Try it out. Like you said before going to Google, just try it out. Just get used to it. It is going to be an important skill in the future. So get on the bandwagon. But think of this.
Garima Gupta [00:27:52]:
I think the biggest mistake people do is they will use the Chat GPT's output as is. Right. Think of it as a burger. Right? It starts with the bottom bun. That is your imagination, your prompt, your skill, your questioning. Right? Then, yes, 80% of your burger meat patty is your AI response. But then that editing that we do on it is, again, very crucial because now you're looking at it, it's like, yeah, it gave me all of this, but it missed out these important things that it couldn't have known or didn't know or whatever. So, yeah, it's important to balance it with our own experience and knowledge of learning science.
Matt Gjertsen [00:28:40]:
Yeah, exactly. Awesome. Well, thank you so much for this quick walkthrough here. Any particular way that people should reach out to you if they heard this? And like, OOH, I need to know more. How should people reach out?
Garima Gupta [00:28:54]:
Absolutely connect with me on LinkedIn. My name is not very common, so you should be able to find me Darima Gupta. My company, Artha learning. Artha is actually a Sanskrit word for meaning, so hopefully that will help you remember this a little bit more. Arthur or write to me at Garima@arthalearning.com. Hopefully you'll write that somewhere in the description. Matt. There is no way people can figure because it's so difficult for Matt and I to explain what we are so excited about.
Garima Gupta [00:29:31]:
If you'd like to see a demo, drop me a line and I will send you a link to our AI demo. Happy?
Matt Gjertsen [00:29:39]:
Awesome. Perfect. Well, this was a great discussion. I had a lot of fun. Thank you so much. Thank you so much for tuning in today. If you liked the discussion, make sure to hit like and subscribe so you never miss an episode. As a reminder, if your team is struggling keeping up with the training development demands of your organization.
Matt Gjertsen [00:29:57]:
We want to help Better Everyday Studios is a full service instructional design team that can help you with everything from ideation to actual content creation and delivery. Please reach out to us using the link in the episode notes below. Have a great day.
Thanks for Listening!
It means so much to me and the guests that you chose to spend your time with us. If you enjoyed listening, make sure you subscribe using your favorite player using the links below.
