AI challenges and successes

  • 14 March 2023
  • 4 replies
  • 138 views

Userlevel 4

I'm excited to be the first to post in our new discussion category on Generative AI and ChatGPT! As you may know, we had a webinar on this, and I wanted to kick off the discussion here. For those of you who missed the webinar, we’ve shared the recording so you can watch on-demand.

I invite you all to join the conversation and share your thoughts on how AI is impacting L&D in your organization. As a starting point:

  1. What successes have you seen with AI/ChatGPT?

  2. What challenges have you faced with AI tools?

  3. What questions surrounding Generative AI and ChatGPT do you have for one another?

 

Let's explore this exciting intersection together and discover new ways to unlock the full potential of AI in the L&D field.


4 replies

Userlevel 5
Badge

Hi Jess, I’m very interested in this topic. I have had a bit of a play with Chat DPT and it is equal parts exhilarating and terrifying. I’m not sure the world is ready for this and what it means for those of us for whom content creation accounts for a large part of our bread and butter. 

I missed the webinar. Not sure how I get on the mailing list for those things? 

Do you have any materials you could share from it?

Best,

Aimee

Userlevel 4

Hey @aswartz

I wrote a six minute recap on the webinar - check it out here. You’ll also find the full recording in there too.

ChatGPT sure is both equal parts exhilarating and terrifying. I think AI still needs a healthy dose of human scrutiny over the outputs, then AI will only do what the ‘human’ programming it puts in. So on both accounts the human aspect is crucial.

 

Thanks

Jess

 

 

Userlevel 6
Badge +2

Hi

Whilst sharing the mix of terror and excitement, I’ve decided to focus on the latter until the time comes when I get ‘terminated’!

In terms of e-learning use cases, I have created a quick proof-of-concept using the ChatGPT API, whereby a user can ask for recommendations for courses based on a free text entry. The user is then given a GPT-generated list of up to three recommendations with a brief explanation behind the choices.  

It works by building a prompt for ChatGPT consisting of the user’s input concatenated with a the text of our course syllabus (titles and short descriptions). ChatGPT 3.5 turbo has a maximum prompt size of 4k tokens, so I needed to streamline the syllabus text to fit it in. 
 

The whole thing was less than 20 lines of python plus a simple HTML template, most of which was generated by ChatGPT!  I can look to share the code here when I have some time. 

My next idea is to do something like this, but to embed it into a SCORM. I’m interested to see how well it can give real-time feedback on short essays using the same approach.  

Userlevel 5
Badge

Hi @Alan that sounds amazing. I’d be interested in the code when you have time to share and also to hear more about your other experiments. My mind is racing with possible use cases. 

 

@Jess Freeland thanks for the recap! Very handy. 

Reply