[Izkata]

> But to latch onto the calculator argument: if you outsource adding numbers to a calculator, you're still you.

Except calculators have been a problem for decades, it's why they're not allowed in school when you're still learning. Without doing the math yourself and internalizing how it works, you won't develop the number sense to tell if the result makes sense (broken calculator, typo, wrong equation, etc).

I still remember my physics teacher using one of the student's test answers as an example of how he should have known it was wrong and gone back over it (I think it was a pendulum on an elevator, his result had negative gravity (so gravity going upwards)).

[SkyBelow]

I see an AI as an advanced calculator, but I see even a simple calculator as a poison to education if not strongly controlled.

Number sense is such a nebulous concept, but it captures what I've seen people struggle with. A general sixth sense about where an equation should or shouldn't go, a feeling that random numbers don't appear quite random enough and might have a pattern, or even recognizing that having a pattern signifies some relationship that otherwise has no evidence.

[nicbou]

We had exams where the calculator was allowed, because we were tested on our ability to type in the right calculations. The calculator wouldn't save you.

[kurthr]

Even in the 80s HPs would do derivatives (and simple integrals). But, open book tests are the worst, nothing could save you!

I still remember giving Mathematica a relational equation of the atomic radius expectation values for it to integrate by parts and collect terms... at the time it failed to find the right integrating factor and gave gibberish. Probably works now.

[rescbr]

I had a professor who polled the class if we preferred our exams to be open book or not. First year of undergrad.

The midterm was open book, he even shared the previous year's exams. The final, as expected, closed book, thank you very much.

[kurthr]

If you can't work in order of magnitude calculations and estimations you're not going to last long in physics. Many useful models give good answers only within a certain range of inputs, while at the same time you often can tell where in a dynamic space you are by external observations. For example, estimating Martian wind speed within a factor of 3x from the size of the sand dunes, can be done having only seen pictures and with only the knowledge of a few dimensionless parameters.

That's also how you can tell if your calculator is "lying" to you (or you typed wrong). I guess I have a few similar tools in spaces where I'm more experienced to see when "hallucinations" and gibberish are being generated by LLMs. Of course having it "check sources" and evaluate its own solutions sometimes also works, if those are reliable, but you're on the hairy edge at that point.