No surprise really, those that don’t offload their thinking understand more and thus can do better at the set tasks and exams etc.
No surprise really, those that don’t offload their thinking understand more and thus can do better at the set tasks and exams etc.
I’m doing a CompSci bachelor program right now, and I’m almost certain there aren’t any students who don’t use AI. The difference is rather between students who use AI responsibly, for learning and own benefit, and students who cheat every assignment and could barely write a line of code by themselves after 2 years of study. I estimate that between 1/2 and 2/3 are the latter kind.
So training the LLMs only on their own data and only running them locally?
That’s the only way to run them responsibly, all the rest is using other’s words, art etc without their consent and using too many resources.
Using it as a search (eg: how do I drop a column in polars?) is not good, but is less bad than “write a function for me to <assignment description>”. Given how Google and others have pushed AI to be the search method, I have a little bit of slack for people who use it like search.
It’s still bad. I’d rather have good documents and a method to find them.
I don’t use AI 🤷♀️
I had to in my internship but I don’t in college
By responsibly I meant using it for improving and augmenting their learning experience, rather than for cheating.
How would it augment and improve the learning experience? Most of the current data on AI is that the only improvement is to homework speed and quality. You don’t learn from the homework, but you get it done fast.
I never learned from homework anyway so what’s the point?
You aren’t incorrect. Homework is not a good system for actually learning anything.
It one of the reason finnish students do better in actually understanding things, because they don’t have homework and thus are not tired out and stressed all the time when they should be playing and exploring things they enjoy.