Adventures in AI (Part #8)

Well, it’s been over a week since I’ve posted, and really since I’ve been able to spend any substantial time on this project. I do, after all, have a full-time (more than full-time!) job and an active volunteer role that gets really busy this time of year. I finally fell into a 3 hour block of work cleaning up things that beta-testers had asked about. So where we were:

To where we are at the moment (and it’s changing, irregularly)

So a few visual changes:

  1. I added the ability to have multiple discussions you’re building on, much like the original ChatGPT. So if you have different courses, you can keep the conversations in a distinct space.
  2. The usage-tracker bar — we set a limit to the amount you could use the service. At the moment, we’re using the most expensive version of ChatGPT4 and so we’re trying to contain costs a bit. Now, I did have a teacher hit the maximum and there was no warning; she was just told she couldn’t ask anything else. So we (it’s funny… I write “we” but it’s just “I” at the moment. I tend to say “we” because I don’t want to seem to be taking credit for things) added a status bar to give folks an idea of how many questions they have left. Because some prompts “cost” more (it’s all based on complexity/duration of processing) it is not a linear jump each time, so we’ll see how helpful it is.
  3. Not visible but I turned on “streaming”. I had complaints that chatOAME took a long time to answer. Well, what was happening was that it was constructing its response in its entirety and then presenting it. Apparently, psychologically, if it keeps putting out individual words, people don’t think it’s taking as long as it is. So that’s where we are. It. Types. Out. Words. As. It. Goes. And. People. Are. Happy.
  4. Our OAME website guy did a great job with the spacing and embedding and cleaned it up generally. It looks much nicer and is more response to different screen sizes.

I haven’t done any work on the Instructions (that which drives our responses), the Voice (see Update #7 and 7 Part 2) or any additions or cleaning of the current KnowledgeBase. That will wait until Christmas when I have some free time, and time that extends beyond just a couple of hours.

I also opened up a OneNote Notebook to start to keep track of all the things to do, the questions I still have, the “what people need to know if I get hit by a truck”, and the research I need to be able to explain decisions I’ve made. I think it’s really important to be able to justify my choices with research — it’s not just my opinion but others in the field (people who do it as their full time job) agree with my direction. This is a field with a chaotic edge to it right now, but with the speed of research publication, we have stuff coming out already at the university level (both philosophy and actual usage) that tracks what I am doing. I do have a presentation coming up in January, as well as an article to write over Christmas on this, so putting all the pieces together will only make things easier.

I’d also like to start making a comparison of responses. I have participants’ prompts and so I’d like to drop them into other services and see what they produce — what value are we bringing into the conversation? How do we have to improve? My own experience using it in my classroom teaching definitely makes it feel centred on Ontario curriculum, pedagogy and assessment and the feedback from my colleagues has been positive — but I also know we like to fool ourselves. I trust no one, myself least of all, when it comes to evaluating what one does.

And the group of beta-testers does slowly grow (1 person added each week is growth!) If you’re an OAME member and would like to try this, I’d love your feedback. You can’t sign up for ChatGPT right now, anyways, so it’s an added bonus.

calarmstrong
calarmstrong
Articles: 29

One comment

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.