AI & eLearning design: I compared ChatGPT and Didask... the verdict is final!

A few days ago, while I was hosting a webinar to demonstrate the Didask LMS platform and when I am about to show the automatic generation of content with our educational artificial intelligence, a prospect calls out to me via the chat module: “But how is it different from ChatGPT?”

I could have expected the question, because after all, who hasn't had fun in the last few weeks with this revolutionary tool and playing with its prompt. eLearning pros are far from being an exception, and many, myself the first one, have tested ChatGPT to create modules, and more and more prompts are circulating on the internet. I started to respond orally and very quickly felt that my explanations lacked concrete content. I then decide to stop the current presentation and start: “Ok, we will do the comparison live, together, you can judge the result yourself”.

So here are the results of this experiment below. Obviously my objectivity can be called into question because I belong to Didask, which is why I will try to be as factual as possible, by illustrating my point with real captures of the test carried out.

Preparing for the test

Let me start by taking an example of a common training situation that we are often encountering lately, onboarding. Many organizations indeed have tons of information to send to their new hires, often scattered in slides, PDFs and other “passive” documents. This was also our case at Didask before, where we grouped our elements on Notion pages. For my test, I choose one of the underlying themes, which is often poorly treated, transparency, or rather how we implement it in our organization.

Here is the page in question, I am simply copying its content.

ChatGPT side

I open the online tool, I paste my text by adding the following instructions to the prompt “Design a ready-to-use online transparency training module based on the following text”, then I start the AI. The text appears as you go, and here's what you can read after a few seconds:

Let's take a look at what Chat GPT did.

In a few seconds, the solution offers me a new text that is similar to a training plan combined with advice. ChatGPT offers me a title, an overall duration of training and then outlines the various points to be addressed, in a fairly organized manner, with a recommendation of the time to be given to each one.

However, if we take into consideration our initial need (designing a training course), we can express 2 regrets:

• If the training plan is interesting... It's still a plan and there is still time to feed each of the parties. Moreover, it essentially repeats the chronology of my original text and does not seem to question or enrich it.
• the rendering is far from being usable for learners and there is still work to be done to make it a real module: choose teaching methods (types of exercises, practical cases, formats, etc.), create them and then format them...

What ChatGPT offers me here is not without added value but only meets part of my need. Interesting for educational engineers who are proficient in training design, but may still be limited for other users such as business experts.

No doubt I could improve the rendering of ChatGPT By working harder my prompts, or by relaunching it with additional instructions. But here too, time can quickly spread out and I would not escape a text → module conversion step either.

‍

Didask side

I connect to my LMS account, access my creation studio and open a new module. I then proceed in exactly the same way and copy and paste the same text directly into the window. I specify a title of my choice.

The LMS then offers me to select a goal for my learners. Here I choose “understand these concepts” which seems to me to be the most appropriate among the items. Finally, I specify that my request concerns an onboarding device, then I launch educational artificial intelligence.

Very quickly, the machine starts to mill and I see gradually appearing Granule objectives. This is the first step in the process: the AI integrated into the LMS scanned my content and granularized it, i.e. it identified the essential concepts and provided a granule for each.

You might think that this granularization suggestion is a kind of improved plan, not that far from what ChatGPT offered me, but if you look closely you also see that the tool already has selected pedagogical modalities. In fact, we see information granules (we explain a concept, below 1, 3 and 5) but also exercise granules (with a scenario, below 2, 4 and 6).

If I “go” into one of the granules, for example 2, I access all its details and I notice that the LMS AI really created the practical case. I am thus offered a scenario that highlights the essential concept of this granule, in this case the fact that transparency is not an end but a means. So I have a statement, a royalty-free photo that illustrates it, a question, my proposed answers with a Corrective feedback for each, the key message to remember.

If this content proposal does not suit me, I can then ask the AI to generate a new one. If I decide not to follow the educational recommendation, I can also ask him to change the type of exercise if, for example, I want to increase or decrease the level of difficulty.

(If you want to know more about the pedagogical functioning of the LMS Didask AI, I invite you to consult the white paper). How to create truly effective courses 10x faster with educational AI written by my colleague Philip).

After reviewing all my pellets, I can see the real results in real life. No formatting or anything else to do, the LMS takes care of it for me. I only click on the “Learner Preview” button and I find myself in the skin of the person who is browsing my course.
‍

Here is what it looks like if I repeat my granule on transparency-what-is-not-a-finality:

I would stop there for this test, as the publishing or administration functionalities of LMS are not too closely related to our problem of content generation today. In terms of timing, all the tasks done on the LMS so far have taken me by less than 10 minutes. This time saving is likely to improve even further with the recent developments of 2024, which will make it possible to create courses directly from materials such as your ppts, slides, pdfs...

What verdict?

The same initial raw content, two very different results, but only one that can be used quickly in training.

But when you think about it, it seems pretty logical. Indeed, if we tend to want to compare everything related to AI with Chat GPT, However, the tools are very different and are not designed for the same purposes.. So the comparison has its limits.

If ChatGPT is a great AI tool for generating universal content, Didask is a eLearning platform LMS that uses AI to create training content, but integrated into an educational architecture. The latter owes its relevance not to AI, but to the recommendations of cognitive science research that guided its development. This is precisely what makes it possible to divide, granularize, and then choose the right teaching methods according to the expected objectives.

Because as Son Ly* often tells us “If you want AI to help you learn better, you already need to know how to learn better”:)

If you want to see in more detail how the Didask LMS and its educational AI works, you can join us on the occasion of an upcoming demo webinar, or even ask directly An appointment.
‍

‍

To continue reading on this subject, I recommend our articles:

• The great history of e-learning platforms
• Adaptive learning and Generative AI: the winning combination for online training!
• LMS: 3 prejudices to bury once and for all
• Beyond ChatGPT: Generative AI at the service of pedagogy
• E-learning: generative AI at the service of learner engagement

‍

*CEO and Co-founder of Didask