To AI or not to AI?

 This is the big question in the news of late. And, if you attend to a certain sector of the media, there are a lot of people out there saying that unregulated use of AI is dangerous. Those of you that know me, know that I've tried to keep a nuanced opinion of AI. I've argued that it has its use cases, but it's not a panacea. As I work on my research here in Chile, I've found AI to be really useful in finding relevant references and in helping me to read and understand those references. As a teacher over the past couple of years, I've found AI to be useful in preparing classroom materials--especially in developing the kinds of active learning activities that are so important in a mathematics classroom.

But, there are things I'm still wrestling with. 

• As I take this year away from teaching, I'm thinking very deeply about the role of AI in learning. In other words, how should students be guided in their use of AI for the purposes of learning? Of late, I'm beginning to think that students should not use AI, at all, in their education. This is a different position than I've held previously. Why?
• In spite of the rapid development of technology all around us, our brains are the same as they were 100,000 years ago. Learning is often challenging. Learning requires effort. Learning can be frustrating. The challenge, the effort and the frustration are the learning. AI can teach, but maybe it teaches a little too well. It reduces friction in the learning process. I think that the friction is exactly where the learning happens.
• To use AI well, a person needs to have knowledge in their head and a critical eye. A person uses their knowledge to identify problems, pose solutions, develop solutions, test solutions and revise accordingly. Without innate knowledge, a person is unable to do any of these. To gain innate knowledge see the previous point.
• I've done a lot of programming in my life. I have the barest understanding of how serious computer viruses work, but even in pre-AI times we saw dreadfully damaging viruses (e.g. MyDoom). AI is particularly good at analyzing computer code because it is so highly structured and easy to test. I'm really worried that someone is already using AI to build a computer virus (or similar exploit) that will shutdown the banking system, or damage an electric grid, or scramble the cellular network, or just turn off the internet... all of it. It's hard to even imagine the impact that would have on the world.
• Even if you're not a student and you're not trying to watch the world burn, AI has a creeping seductiveness that leads a person to cede more and more of their agency. I'll use myself as a case in point. I've been using Google's Gemini AI to determine my workouts for the better part of a year. I feed it data from my fitness watch, tell it that I'd like to run a half-marathon in March 2027 and it tells me what to do every day... and I do it. Why? I don't know. Pre-AI, I would've searched the internet for training plans. Pre-internet, maybe read running magazines or go to the library. Letting the AI figure it out is easy. So that's what I do. In this small instance, I've given up some of my autonomy. I've turned it over to a computer program.
  
  A lot of people are giving up little pieces of their autonomy every day. What does that look like in 1 year, 5 years, 10 years. One thing I'm quite sure of, it will have been much easier to give it up than to get it back. 

Maybe I just need to go outside and sit on the beach.

****

Obligatory sunset photo.