As a high school teacher I use LLMs a lot. My peers do too. It’s just too easy not to have a model type out all the text in whatever format for (all the different) software programs we use.
One quirk of the LLMs is they always make the correct multiple choice answer the longest multiple choice answers. When in double thirty years ago we just chose “C” – now, students choose the longest answer.
But there’s something lost. TNSTAFL. Maybe not in the busy work of typing out questions, but maybe so. It could be a feature, not a bug that audiobooks (and books), are so long. It’s more time to sit with an idea.
In that spirit, here are two ideas and inspirations about doing the work:
Ezra Klein: “Having AI summarize a book or paper for me is a disaster.
It has no idea what I really wanted to know and wouldn’t have made the connections I would’ve made. I’m interested in the thing I will see that other people wouldn’t have seen, and I think AI typically sees what everybody else would see.
I’m not saying that AI can’t be useful, but I’m pretty against shortcuts.
And obviously, you have to limit the amount of work you’re doing. You can’t read literally everything. But in some ways, I think it’s more dangerous to think you’ve read something that you haven’t than to not read it at all.
I think the time you spend with things is pretty important.”
Daryl Morey: “They (Bill James) said, ‘Hey intern, you go work on this project, see if you can look at point differential and how it affects the NBA.’
So that was my very first analytics project, which is in the ’93 basketball book. Yeah, I remember coming up—like, literally worked on it 24/7 because I was like, ‘This is my moment to try and prove something to Bill James,’ who, by the way, didn’t care at all when I did it.
He, you know, he probably thought it was derivative because it sort of was. But it took a while to come up with the idea of modifying the exponent. That was obviously the first time Bill’s ideas were adapted to basketball.”
The Waiter's Pad 