The AI Mirror Edition

Noah here. There are two stories we hear a lot about AI and writing these days:

• A junior employee responded to a work inquiry by generating a bunch of AI slop, never reading it, and then handing it in as the solution to the problem with great pride for their brilliant (unread) answer.

• A college student uses AI to write their 5-page report, leaving the professor to dig through slop paper after slop paper, trying to distinguish human effort from token completion.

In both of these parables, the emphasis is on the young person handing in the AI-generated assignment. And in both of these cases, I think that is the completely wrong read of the situation, and it drives me crazy.

Why is this interesting?

A few years ago, I was asked to give a talk about AI at a very fancy high school. After the talk, I was pulled aside by an 11th-grade English teacher who had some genuine questions and worries about how she should navigate her job moving forward. I’m completely unqualified to answer questions about educating 16 and 17-year-olds, but I was a professional journalist for a period of time and have spent a good portion of my adult life writing things on the internet. I also have two kids in elementary school. I really care about writing and believe it has tremendous value.

What I said to her was only that I didn’t really think the fundamentals of her job had changed. That is to say, I don’t believe the job of a high school English teacher is to develop great writers, but rather to convince teenagers that devoting time and energy to writing is a worthwhile pursuit. That we got so far away from that and focused only on the final output is a systematic failure. I see it in the way my elementary-aged kids are being taught to write: Instead of being encouraged to find their own voice, they’re given what essentially amount to prompts and templates to follow. I don’t blame any teacher for this—navigating a room of 25 children at 25 different levels isn’t a job I envy—but I still think it’s a failure, one I’m trying to correct as a parent.

When these folks enter the workforce and people throw their hands up in the air because they’re coming back with junk, where is the accountability? If you’ve got a junior employee who comes back to you with nonsense, isn’t it exactly your job to help them understand why that’s not the right path? Similarly, as all these universities look at AI and wonder how they’re going to deal with all the cheating, isn’t the better question whether the way they’re judging performance is actually optimal if it’s so easy for AI to achieve? I don’t know about you, but I certainly handed in a bunch of pre-AI slop in college.

This is also my answer to the question people like to ask when they imagine a future where one AI writes a presentation or email, and another AI reads it. What kind of world is that to live in, they wonder. I agree! That’s incredibly dumb. But the answer, quite obviously, isn’t that we should go back to the old way, it’s that we should get rid of the performative artifacts of work that have so little value that you can hand them off to AI for both writing and reading.

AI is a mirror, not a crystal ball. It reflects ourselves back to us because it was trained on our words. When we wonder why it’s boring or corporate, it’s because we’re boring and corporate! We start teaching kids to write that way when they’re nine or ten, and then we encourage it throughout their lives into adulthood because we prize quantity over quality and allow for decisions to be made by committee. These are the problems worth solving in my view—AI is simply surfacing them with new fidelity. (NRB)