In Search of a Custom AI Chatbot

July 31, 2024 by

I’ve been playing with some AI tools that let you build a custom chatbot. Matthew Vollmer, a colleague in the Department of English, mentioned back in May that he had created a custom bot that was loaded with his course documents.

I was immediately envious. Sure, students could just do a search of the course documents, but they frequently don’t. The idea of loading up a bot with my course policies manual, short guide to the course, and the group guide sounded wonderful. Students could ask the bot questions, and they’d get answers right from the course documents.

This evening, I played first with making my own GPT with ChatGPT (which requires a paid subscription). I wasn’t very successful so I looked for something that might be simpler. I landed on ChatBotKit, and I love it. I only fed it my group guide to test it. Free accounts are severely limited.

I set up my bot as a widget and tossed it up on a webpage to test. You can take a look for now. I’m already getting warning about exceeding the limits for the free account, so I’ll have to take it down in the next day. Here’s a screenshot showing part of a conversation I had with it, asking questions students asked at the beginning of the Fall and Spring Semesters:

A screenshot from ChatBotKit showing answers to two questions about student groups
A screenshot from ChatBotKit showing answers to two questions about student groups

The answers shown in the screenshot come directly from my group guide. I’m almost giddy about the possibilities. Students could ask direct questions and get immediate answers. Admittedly, I would rephrase things a bit, but overall, it’s using the text that I had in the guide. I think the changes are primarily where I’ve used first-person pronouns in the group guide. The bot isn’t the “I” from that guide, so it’s rephrasing.

There’s a problem however. Probably two, if I count Virginia Tech’s restrictions. We are only allowed to use approved software, and the only AI that’s approved is Copilot. The bigger issue for me personally is that the free version is not nearly enough. Pricing costs $25/month for the Basic package. The Pro subscription, which offers a privacy option that strips any personal data all messages, is $65/month. There’s a Team subscription that costs $365/month as well.

While ChatBotKit does exactly what I want quite easily, there’s no way I can pay $65/month to invoke the privacy option. I’m not even able to sign up for $25/month right now. So back to more searching for me. Why can’t I win the lottery?

Filed under artificial intelligence

It’s Okay to Use AI

July 27, 2024 by

A background shows black, gray, and white code with orange highlights. On top of this faded background are these words: It's okay to use AII was just thinking about the “Run the dishwasher twice” advice this early morning. I take it to also mean “It’s okay to run the dishwasher twice” when you just don’t have the energy to unload the 1/2 pile of dishes in there. It’s okay to add what’s on the counter and run it again.

I apply this general concept to EVERYTHING (or at least I try to). The general idea to me is that it’s okay to take shortcuts, even if they may not be the ideal, best choice for you, your family, or the universe. I spent a few minutes this morning making myself some “Run the dishwasher twice” mini-posters to hang up and remind myself that it’s okay to do what you need to do, even if it’s not the most perfect choice in the universe. Believing that there’s only that ONE allowable way to do things is just perfectionism, and we all know perfectionism is the enemy of getting things done.

While in this “Run the dishwasher twice” headspace, I next picked up my to-do list of AI-related ethical scenarios that I’ve been working on. I’ve been using chatGPT to help me write these scenarios. Don’t get me wrong: They are my ideas and I edit the heck out of them. That said, I let AI brainstorm options, clean things up, and even do the HTML coding for me. It’s not that I can’t do all this myself. I’m simply using AI to speed up my productivity.

Of course, there’s a problem with all this. Those negative voices immediately showed up in my head. I frequently—VERY frequently—feel as if I’m cheating when I use AI. The kerfuffle over students using AI to cheat the system and plagiarize their way to graduation doesn’t help. I know that’s a ridiculous claim, but it’s still in my head. So when I go about using AI, I have this belief that I’m cheating the system. “Real” writers would never use such artificial shortcuts. “Real” writers would never use AI to create the images for their text. They would search for Creative Commons images. “Real” teachers would never, never use AI.

As I was sitting here in my self-loathing, telling myself that I could and should do better, my eyes glanced back over the “Run the dishwasher twice” mini-posters that I just made, that I haven’t even hung up yet. My remarkable voices shifted completely: “Stop it!” they said. “You’re just running the dishwasher twice. There’s nothing wrong with using AI to get things done.’ Sometimes my voices are super-freaking amazing.

So in case you’re in the “using AI is cheating” headspace, let me say this loud and clear:

It’s NOT cheating to use artificial intelligence to get things done. You’re using the tools you have at hand to get things done. Use it! Do your best work! Be your best you!

In fact, if you’re like me, a NTT teacher, you already have fewer resources and more to do than many colleagues for whom writing and research are required and supported. If you’re a person of color, a person with disabilities, a woman, or LBGTQ+, you historically have fewer resources and more demands on you. So take advantage of the tools you have available. What if you could write that code yourself or brainstorm a list of sample topics for your class? Let AI do that work so that you have time to mentor a student or colleague, pick up your child, or just take a deep breath. Run the dishwasher twice.

Filed under artificial intelligence, Journal