Wharton's Ethan Mollick argues AI use proves homework is essential for human learning, prompting comparisons to model training

See, for example

@emollick Because AIs learned by doing homework?

@emollick Felt this. When I let AI write the first draft I retain almost nothing. When I force myself through it first and then have it critique me, it actually sticks. Turns out the friction was never the obstacle, it was the point.

@emollick Obvious but it also follows that writing is part of learning. And while that is obvious to many of us, ppl seem not to notice

@emollick Didn’t we know that before AI?

@emollick I don't necessarily agree I think understanding matters, and homework and repitition can improve understanding, but if understanding comes before homework then they are unnecessary Homework were important when I didn't really understand a topic, not when I did

@emollick I think it’s too early to judge. It will probably take decades to tell whether younger generations have really lost it or not

@emollick 😂😂 L'AI ha dimostrato che saltare i compiti è come saltare il giorno delle gambe — il deficit si vede sempre dopo?

@emollick AI shows the gap between copying answers and actually understanding. Homework still builds the real muscle.

@emollick We’re the luckiest generation to have something so interactive and powerful as our tutor or learning companion.

@emollick The tech everyone thought would make doing the work optional just proved that doing the work is the point. That's either the best joke or the most expensive lesson in education research history.

@emollick Somehow humans learn without trillions of examples. So... I'm not sure.

@emollick The detractors are all wrong. If you practice something you get better at it.

In my undergrad circuit classes I spent evenings practicing circuit problems. I was always one of the first to finish the final exam and got near perfect scores.

@emollick this take is interesting bc it proves the act of doing something yourself actually locks it in.

AI shortcuts expose how much we leaned on completion over comprehension.

@emollick Learning requires effort, but not necessarily the effort the teacher expected. Learning can be much faster due to better motivation and explanations and an infinitive amount of exercises to try out newfound understanding on

@emollick And …

@emollick Absolutely. AI expands access to knowledge, but we still learn through effort, discipline and critical thinking. Technology changes the tools, not the value of the work.

@emollick depends on what you are trying to "learn"

@emollick Same in building. An agent shipped a working invoice app from one prompt, but I only knew what to fix because I'd done the manual version first. AI does the homework faster; you still have to have done it once to recognize when the output is wrong.