AI and me
This post is well overdue, because I’ve been struggling lately with how I feel about AI. I have 9 months of drafts of this blog post!
First, let’s clarify that I’m talking about Generative AI, including LLMs (Large Language Model) and Image Generation. I’m still pretty happy with Machine Learning, a subject that I’ve worked with extensively, and teach about.
I’ve started using these tools, slowly, but as is inevitable, more and more as I realized the possibility. I have a pro ChatGPT account, as well as a Copilot account through BCIT. I’ve looked at Claude, and that’s probably what I’ll pay for next, once my last month of ChatGPT expires. I’ve started using it more, and I now have thoughts.
The Good
First, my regular reader(s) will recognize that I updated the CSS of this website, or rather ChatGPT did it for me. It also added in an RSS feed, so now you can read this blog in your RSS reader of choice. (I use Feedly). AI has helped me research backpacks, and write grant proposals.
I fed ChatGPT this image of my living room wall, and asked it to show me where to hang art:
It didn’t get everything correct, but I do enjoy this image:
As well as this one:
I especially like how it took the art I already had, and created new art, based on it. Pretty interesting!
I’ve been using AI as a “coding buddy”. I’m not going to write in all of my projects, but maybe I will? Needless to say, it’s fun and I’ve completed some projects that I’ve meant to get to for years, in a matter of minutes.
The Bad.
I’m not alone in seeing an explosion of AI cheating among my students. The tool exists, so it will be used. But thanks to tools like Google Lens, all of the work I’ve done creating randomized exams in D2L is now useless - with a click of a button, an AI will work out an exam and solve it, earning a score between 45% and 80%. (It’s still pretty bad at math, and that is what I mostly teach)
As a result, I’m now back to mostly paper exams. Even then, students smuggle in phones and AI pens to feed my questions to an AI and get solutions back. Right now, it feels like playing whack-a-mole against new cheating items. A ruler that has a screen and sensor to communicate with AI! Smart Glasses that have built in AI! Smart watches! It’s a lot, not to mention good old “copy off of someone else”
I find that reports and emails, from my colleagues and students are becoming “smoothed out”. Everything has the same diction, the same voice. I know you can fix this, but creativity seems lacking.
I was assured that the enterprise CoPilot will not produce images that violate copyrightlaw, but this image is clearly a copyright violation:
I like the idea though!
Conclusion
I’m not sure what my conclusion is! Genarative AI is here to stay. It’s what my job has become, and all of us need to learn how to use it as a responsible tool. We need to change how we do things, but how?
I’ll leave you with an image of what AI does best: take ownership of mistakes!
