I dunno that I'd agree with that either (for this specific type of example at least).

An adult wanting to find "how many hours are in 3.228 years?" when some task is looking for the answer aligns a bit with what you're saying. In this kind of scenario it has nothing with the adult not understanding how to do the conversion and everything to do with what the most accurate way to quickly complete the conversion itself is. I.e. the task is "produce an answer so we can use it". Whether that be done via calculator, spreadsheet, timespan conversion tool, or LLM (though the latter can be a bit of a poor choice for this type of task to be honest) the use of the tool is about best achieving the original goal accurately and quickly, not about sidestepping it.

A student wanting to find "how many hours are there in 1 day and 7 hours" on their homework aligns very differently in my opinion. In this kind of scenario the goal was that the student become more educated on time conversions by practicing them. The actual answer provided is just a way to be able to provide feedback on how accurately that learning process went. Learning these kinds of basic things through experience allows them to tackle learning ever more complex things as they advance through their education. Faking gaining that education by using a tool to get an answer to write down isn't just the lazier way to do the conversion, it does leave them without the education the task was actually about.

Not only that, learning to do such basic timespan conversions is going to be the lazy way for such tasks over anything but the briefest timespan anyways!

If LLMs were being used as part of the learning process instead of a replacement of the learning process I'd hold different opinions of the post. I think tools (be it calculators or LLMs) can be extremely powerful learning aids... but extremely weak ones when used as learning replacements.