Post

I was talking with a colleague the other day who said that while he’s not sure if AI will change physics research so much as to unemploy the physicists, he’s confident that teaching will always need humans.

Benjamin Riley@benjaminjriley.bsky.social

So we know AI helped a negligible number of students. But how many did it harm? I've crunched the raw data provided by the lead researcher, and what I've found is that ~9,500 students using AI studied less than even the *least* studious, non-AI using student. This should shock the conscience.

79 reposts475 likes

Did you ever play around with Foldit, the game from the 2000s where they gamified modeling protein folding? These types of things are exactly where AI has been revolutionary to research in biology; it’s specific use cases but real

Oh AI as in machine learning has been revolutionary in science for ages. Modern LLMs are a different kind of beast but I know a ton of my colleagues are getting a lot of utility from them in their research projects.

I think that’s the problem with AI discourse online, most of the fervent anti-AI people are throwing in machine learning with chatbots as if AI is a synonym for ChatGPT

We know computers in the classroom hurt educational outcomes (or, at least, provide no benefits) so unsurprising that AI is negative overall too. Of course, that doesn't mean AI is bad overall, any more than it means we should abandon computers in business and society.

Not to the graph, but to teaching needing humans. If you're self motivated now you can find one information already, but it's a lot harder to find the right order to learn those things, know if you got it or know something wrong, and let's be honest, have you met a kid excited about addition

"What would be the point of taking physics if nobody can find a job after they graduate?" That's a question for philosophers. Also, anybody getting a degree in philosophy.

Maybe at some elevated level but I wonder whether good teachers really understand how many students go through the entire system without ever having a single teacher who takes a personal interest in them. Most of it can be replaced by videos and books.

Some schools are (finally) making the argument to add cursive back to the curriculum because people need to be able to read historical documents, and the easiest way to learn to read it is by writing it.

AI for students could be very helpful to explain things and direct reading but you can't use it to answer exercise questions and do your work, the process of doing the work is the learning bit. The temptation to just do this one report is probably very high though.

Funnily enough the best use case I can come up with for AI is all of the things we STEM types do that WASN'T on the glossy signs outside and in the pamphlets when we signed up for our majors. Like making formulas for excel spreadsheets to analyze our data or statistics from an experiment.

…well. Hell. I suppose spreadsheet formulas are a use case I can ethically get behind, if my company is going to have it available for me either way. I’d never otherwise be given a budget to help me craft a spreadsheet formula and I am stuck on one section. Particularly not for multiple programs.

Sheets hurts my brain because I'm so used to excel keyboard shortcuts that I end up using them by instinct and unfortunately those are ALSO Google Sheet shortcuts, just assigned to different functions.

A few years back, I started as a part time college professor to share my many decades of experience in engineering, math, and physics. I quit this past spring. I was wasting my time because most just turned in AI slop. Same teaching 4th year engineering at university. The institutions didn't care.

This is where I land basically - LLMs might do weird shit for physics as a discipline but I really don't think there will ever be a substitute for an actual lecturer standing in front of a blackboard, no matter how good the voice models get.

I’ve come to recognize education as “reshaping the brain”; something that starts at a microscopic level and gradually builds. Very much like exercise gradually reshapes the body. Graduates are expected to *think* differently. AI often leads to students avoiding the necessary mental exercise