Ai Musings

in Pages

Note: This page was originally written by Professor Michael Gleicher. The discussion and opinions are presented from his perspective.

Can’t AI just do it?

It turns out that the assignments for the class are in the “sweet spot” for current AI models. The programs are small and simple and well-specified enough that AI tools might just give you “right answers.” If nothing else, AI can cheat - enough old solutions are in the training set that it certainly knows how to do the assignments. In fact, I find that AI models tend to make the same mistakes that students make. I don’t know whether it is copying student code or making the same reasoning flaws.

There is an academic integrity issue: if you copied an assignment from a previous student without giving proper attribution, that is an academic misconduct issue. This is similar to LLMs stealing the work of newspapers and novelists (except without the lawyers and big payouts). But we’ll put that aside for the moment.

AI can do the assignments for this class (small graphics programs taken from its training set) “better” than you (or at least well enough to get a good grade). If your goal is to get the assignment done, it is not an unreasonable choice. However, if your goal is to learn what is going on so that you can use tools (including AI) effectively to make bigger things, you need to actually learn the materials.

If you (1) know the graphics content, and (2) are reasonably skilled with AI-augmented programming, you can do the assignments for this class quite quickly. I am rusty as a programmer (as a Professor, I don’t get to program as much as I would like), and a novice with AI tools.

But years of professing (including trying to help students fix their code) and a good understanding of how the programs are supposed to work mean I can do these assignments with AI tools very quickly.

Some background

“Indeed, I would argue that this is the first time in the history of higher education that we have learning outcomes that are at war with one another. For instance, we have never said we can’t teach our students written communication. That’s going to interfere with their critical thinking skills. No, they actually amplify one another… But AI as a learning outcome actually has the possibility of diminishing the achievement of these other outcomes”.