[periheli0n]

Computer science, and academia in general, has always adapted to technical progress, although slowly.

In the case of ChatGPT and similar LLMs, these should become part of the toolbox that students are being taught. I.e., how does it work, what can it do, what are its limits, how can it be used to help solve a problem or complete a task.

An exam question could then be e.g. to ask ChatGPT for an essay on a topic, critically discuss its shortcomings, and improve the essay e.g. by adding references and deeper discussion, which will be graded.

Alternatively, use ChatGPT to iteratively discuss and improve the essay. The whole chat transcript should be included and will be graded.

[ckemere]

Plus one to this. The reason we make students read and write is to teach them what good communication is. ChatGPT currently produces (and likely will continue to produce for a while) mediocre prose. How do I know that? Because I've read good prose and I've attempted to write it and understand the editing process. At some level it represents a huge pedagogical opportunity.

(Grading IS moderately harder, but not that much.)

[908B64B197]

The slide rule didn't render arithmetic obsolete. Neither did the calculator. Being able to do math on paper is a side effect of understanding the base 10 representation of number most of the world has settled on. And the first thing someone in college for EE does is re-learn this arithmetic in base 2. The calculator is just an extra tool to get to numerical results faster.

ChatGPT is an extra tool as well. Smart students will realize they have an extra tool to get an upper edge when competing in the zero-sum game they've been thrown in. They'll have ChatGPT generate the essay from an outline they wrote and will then trim and re-formulate as needed, spending most time fact checking and reading from source material to make sure ChatGPT isn't making stuff up (what will get some of their classmates caught).

[russdill]

ChatGPT is just another mile marker that we are whizzing past.