AI powered Text-to-Course is not there yet, but it’s been getting better over 2023…
We have been experimenting with AI powered Text-to-Course for the last 12 months, and the progress has been amazing, but there is a way to go.
In our system you can enter inputs such as:
1. Course Objectives
2. Course Audience
3. Course Prerequisites
4. Style and Tone
5. Source documents.
6. Etc.
It will create SCORM courses based on your inputs, including test questions.
The creation of text, images, and even short videos is impressive; however we have run into a few issues that make it not ready for prime time yet.
Issues that we encountered were:
• Hallucinations ( this is still an issue despite using document inputs to narrow fact checking)
• Consistency over all modules
• Coherence between modules
• Sticking to a narrow subject context
• Integration with existing educational frameworks
The improvement of these is impressive, but there is one big lesson learnt so far.
“Do not rely on generative AI to do everything!”
One of the biggest improvements we have seen is when we add a “taxonomy of knowledge” to the system. Not only does this help reduce hallucinations incredibly, but also allows a narrow definition of what a student already knows, thus providing orders better personalized learning experience.
Whilst creating full AI authored courses is not ready just yet, we have found use for it in creating isolated short content for use in Tik-Tok style personalized learning feed.
The evolution of learning continues….



Leave a comment