I agree with the point that the content of the text matters, but most people care more about who wrote it. For example, no matter how much I write on my blog (www.makonea.com), no one reads it. But if the same content is written by someone famous, it gets cited far more often. That's the problem.

There are people who want to obtain information from text, and there are people who want "communication." In communication, the more famous the person is, the more enthusiastic the response tends to be. In other words, the purpose of reading is often different.

So I think there are two types of reading: reading for communication and reading for information.

Writing for communication elicits completely different reactions depending on who wrote it. That's one example. Writing by AI betrays that expectation. The problem is that there's quite a fuzzy boundary between what's written by AI and what's written by me.

On top of that, I want to add something about the "flatness" of AI writing. By flatness, I mean not pushing any particular argument strongly and appropriately acknowledging drawbacks as well. When I write my blog posts, I run everything through an AI red-team review. The final draft ends up very different from the initial one, but the writing loses the sharpness of my subjective bias. It becomes blunted. As a result, the sentences become flatter in places. This happens because the AI detects and smooths out my unconscious biases. (Of course, AI has its own biases too, but humans also chip away at AI's biases.)

That said, if I don't use AI at all, since I pride myself on running a technical blog, not detecting technical errors would also be a problem. So it's a dilemma.

My personal concern is that in the long run, a lot of people's writing will become flat. AI is genuinely useful as an encyclopedia, and while it can contain errors, on average it's often more accurate than I am, so it's hard not to use it. The problem is that the materials AI learned from, the writings of knowledgeable people, were shaped by a strategy of finding safe, low-cost landing points when engaging in high-cost criticism. In a social context, that's actually a fairly valid strategy. Strong claims invite strong backlash, but most people smooth things over appropriately for the sake of social harmony.

The issue is that human writing is generally rougher. But if we insist on preserving only human writing, we end up having to define humanity's roughness as mere "barbarism." Ultimately, I think human writing might even start imitating AI writing. And if that happens, what becomes of writing as "communication"? We write not only for information, but also because we want to talk to someone. Like when I write on HN.

And while AI expands the creative world, giving tools to people who want to create but didn't know how, like building a philosophical Frankenstein RAG chatbot, whether that tool is actually accurate ultimately requires knowing how to use the tool properly. In other words, liberation and accuracy exist in tension with each other.

As for cultural comparisons, I've worked with people from Korea, China, Japan, and the US. The fact that Korea and China are more lenient toward AI is partly because Korea, China, and Japan are direct beneficiaries of AI. For instance, Korean stocks are essentially a massive semiconductor leverage play, and the direct consumers of memory semiconductors are AI. China, meanwhile, sees AI as a new path to surpass the US.

I really enjoyed reading your text. Humanities scholars definitely write in a way that's fun and logically compelling, which makes it a pleasure to read. It really drove home for me that this is the kind of writing you need to produce to make money from writing. Fortunately, I'm grateful that my job is being a programmer.