Thing of Things AI use policy

dynomight recently wrote an article calling for bloggers to state publicly whether and how they use AI:

I bring this up because I’m worried that blogs are heading into a sort of lemon market. You’ve surely had the experience of reading an essay only to slowly become dismayed as you realize it was AI-written. What’s the equilibrium? I expect that some people have already cut back on reading essays, at least from non-established authors. Over time, I expect this will lead to fewer humans writing essays, further increasing the density of AI-generated content, driving more people to cut back on reading, et cetera. This is bad because blogs are good.

As that cycle turns, social norms are also changing. Cast your mind back to the old world, five years ago. At that time, if you had started a blog and posted AI-generated essays without telling anyone, I’m reasonably certain that would have been considered a dick move. (Future generations will marvel.) But today, the largest corporations appear to do that all the time. There’s incredible momentum towards a world where AI can be used anywhere, for any purpose, with no disclosure, and that’s fine.

But it is fine! At this point, trying to bully people into proactive disclosure is just a tax on honesty / consciousness / integrity. Instead, I suggest we agree that arbitrary usage is, by default, fine. Instead, let’s work at the other end: If you have chosen to impose limits on your AI usage, then state those limits publicly. If you’re human, tell me...

I think this is a good idea so I’m doing it.

Like dynomight, I will “guarantee that every word I post here is the product of me physically hitting keys with my fingers. The only exceptions would be quotes from other humans or something that’s clearly labeled as an AI output.”¹

The most common way I use AI when writing posts is as an advanced search engine and/or thesaurus: I ask questions like “Synonyms for ‘safe’ in the sense of being free from danger?” or “What are some events that have a 1% probability?” or “what’s the word for the thing where if a small interest group would be helped a lot by a policy, and everyone else would be harmed a tiny bit by that policy, the policy is more likely to get passed because none of the people harmed care enough to protest?” I do this probably three or four times a post.

I use AIs to check facts in blog posts: “there are definitely Chinese temples that are critically acclaimed by architects, right?”, “can you give an example of an extremely niche kind of quantitative trading someone might specialize in?”, “what’s a hemiellipsoid body? Please explain to me like I’m five.” I usually independently fact-check anything the AI tells me; I ask it for sources and then read the sources myself, or I google keywords and follow up with a reliable source. Occasionally I don’t bother—for example, if the fact is supposed to add color to an example or a joke, rather than being information I expect people to rely on.

I use AIs regularly to find papers for lit reviews, if I’m writing a post that requires a lit review, instead of digging through Google Scholar and paper citations myself. When I’m doing a literature review post, I read all the papers myself, rather than relying on an AI summary.² I may occasionally ask AIs to explain aspects of a paper I don’t understand.

I have yet to use Claude Code to analyze data, but I expect I will in the future. When using Claude Code to analyze data, I will thoroughly look over the code to make sure that I understand what all of it is doing and that I approve. One of my beta readers doubts my claim here because Claude Code produces so much code that it is impractical to check all the data analysis code manually, but I would much rather write my own data analysis code than publish data analysis I don’t understand and might not stand by.

Probably I would use AIs to generate tables and graphs, if I ever put tables and graphs in my Substack, which I don’t because I’m not a visual thinker.

I don’t use AIs for editing nonfiction. Frankly, I don’t edit my nonfiction that much at all. At least 40% of the time you guys are reading my unedited first drafts.

I don’t use AIs for brainstorming blog posts because it feels icky. I want to develop my ideas through getting in arguments on Discord forthright and rational debate with people I respect, the way Socrates intended. I want to get several different perspectives on a nascent idea, and modern AIs can only give me one perspective. Besides, “arguing with people in a way that might become a blog post” is like 15% of my social interaction. This is the stance that is least likely to change, at least before we build a Dyson sphere out of Mercury.

1. Can we call this the Dynomight Pledge?

2. Sometimes this is a heroic act of self-sacrifice.