[cluoma]

I used to get this same feeling during lectures in uni. Often the information was presented well and, along with some clear examples, everything seemed to make perfect sense.

It wasn't until working through practice problems later, on my own, did it become clear how much detail I was missing.

[Swizec]

> It wasn't until working through practice problems later, on my own, did it become clear how much detail I was missing.

This is a common problem in learning. Recognition is easier than recall and smoothness is confused for understanding.

You actually need to struggle with the concepts a bit to learn effectively. Without the struggle it feels more effective, but is not.

[bamboozled]

Detail missing and being a confidently wrong are two different things though ?

Edit: Claude told me the other day told me my entire building might have to be demolished due to a slightly bow in my newly poured stem wall, I uploaded a photo etc and it was liked, “yes this is a serious structural issue blah blah blah” , the inspector came to look at it and literally laughed that I was worried about it.

[elgenie]

Now consider what's happening to the learning process of the (rather large) subset of current college students choosing to replace that struggle for detailed understanding with LLM queries.

[brookst]

It’s the biggest crisis since math students started using graphing calculators.

[mr_toad]

LLMs are just graphing calculators for the humanities.

[moi2388]

That’s beautiful

[avs733]

you are not the only one. There was a paper covering this exact topic in the Proceedings of the National Academy of Sciences a few years back [0].

Passive learning (lecture) scored better on:

* Student Enjoyment

* Feeling of Learning

* Instructor effectiveness

* I wish all my courses where taught this way

Active Learning (i.e., not lecture) scored better on:

* Actual learning

The differences are not small.

[0] https://www.pnas.org/doi/10.1073/pnas.1821936116

[malshe]

I suspect a majority of my students this semester used LLMs to complete homework assignments. It is really depressing. I spent hours making these assignments and all they probably did was to copy and paste them into ChatGPT. The worst part is when they write to me asking for help, sharing their code, and I can see it was written by LLMs. The errors are mostly there because occasionally the assignments refer to something we did in the class. Without that context LLMs make assumptions and the code fails to generate the exact output. So now I am fixing the part of the code that some of my students didn't bother to write themselves.

Edit: Added "I suspect" in the beginning as I can't prove it.

[chmod775]

Why are you fixing their code? You're just doing the same thing as the LLM you're complaining about.

Also, as someone who attended university in Germany, the mental image of a professor helping undergrads with homework already seems strange if not funny to me. That is... at least I hope they're undergrads, because if people managed to get any sort of CS degree while having to rely on a LLM to code I might be sick.

[WingedBadger]

Just going by the last 2 years of university teaching (energy focused computer science in germany), I feel like LLMs have already had a devastating effect. There has been a large influx of students who seemingly got through their entire Bachelors degree with nothing but ChatGPT. The university is slow to adapt and ill equipped to deal with this.

This is absolutely killing my enjoyment of teaching. There is nothing more disheartening than carefully preparing materials for people to grasp concepts I find extremely interesting, just for them to hand in ChatGPT generated slop and not understanding anything at all. In stark contrast, just a couple of years prior I would have quite rewarding projects and discussions with students. I also refuse to give detailed feedback on such "solutions" anymore because the asymmetry in student effort and my effort is just completely unreasonable.

This development is something very different from the often quipped "graphing calculator in maths education". For a graphing calculator you still need to know the mathematical foundations to input the correct things to get the correct results. LLMs are mostly used by just pasting in the exercise of the day.

This is not to say LLMs can't be a useful tool for learning. They absolutely can. But that is not how the majority of students uses them... to their own detriment and the detriment of those trying to teach them.

If universities don't adapt to this quickly, then the already weak signal of "university degree implies some amount of competence" will be entirely lost.

[michaelbuckbee]

I've heard experts comment on this from the other side, that they'll give a quick layperson's soundbite about their subject of expertise that doesn't defensibly lay out all the possible exceptions and edge cases and weirdness for reasons of time and audience interest and then they'll be inundated with comments calling them a liar and accused of falsifying things or not actually understanding the subject.