←back to thread

A high schooler writes about AI tools in the classroom

(www.theatlantic.com)
222 points dougb5 | 1 comments | 04 Sep 25 02:37 UTC | HN request time: 0.4s | source
Show context
greenspam ◴[04 Sep 25 22:50 UTC] No.45133130[source]▶
Imagine back in the days when calculators were just invented. An 8 year old kid might have the similar complain: “my classmate finished a 4 digits number multiplication problem in 5 seconds which generally took 1mins.” People might say, in the long term, the kid who cheated would be less proficient in arithmetic, which turned out to be true. But when you think about it, it seems not the end of the world when most high schooler in US cannot do complicated arithmetic quickly and accurately without a calculator.
replies(7): >>45133182 #>>45133296 #>>45133350 #>>45133511 #>>45133804 #>>45135086 #>>45136156 #
1. Terr_ ◴[05 Sep 25 00:33 UTC] No.45133804[source]▶
The two situations are not analogous.

The kind of task is not the same. With a calculator, you are delegating a very specific, bounded, and well-defined task. Being unable to approximate non-integer square roots by hand isn't the same as not knowing what square roots mean or when they are applicable. However with LLMs, people are often (trying to) delegate their executive-function and planning.

Another way to tell that the tasks are qualitatively different is to look at what levels/kinds of errors users will tolerate. A company selling calculators that gave subtly but undeniably-wrong answers 5% of the time would rightfully go bankrupt.

If you want to compare LLMs to something of yesteryear, it's closer to hiring someone to do the work for you: That's always been considered cheating, regardless of how cheaply the accomplice works or how badly they screw up.