How I use AI

I’ll get it out of the way first: none of the writing on my blog is done with or assisted by AI. It is entirely my own, and will remain that way because my blog posts are mostly intended to be expressions of thought in public. I want you to see my thinking, not whatever garbage a model spits out.

I have sometimes asked for AI for help with programming. Most of this site was made by me though. I run some experimental projects with AI that are not published anywhere, because I want to understand what it wrote first before putting it up anywhere.

I keep AI far away from my Obsidian vault. It has no place in managing my knowledge for me. AI assistance is also not used in my contributions to open data projects like Wikipedia and OpenStreetMap.

My thoughts on AI

I’m conflicted about AI. Well, large language models and other generative models that have been all the craze as of late (which aren’t really AI, that’s just a marketing term^[1]).

On the one hand, they are tools that can be used responsibly. On the other, it is not some neutral technology: LLMs are impacting a lot of people in negative ways, and have not been a universally good thing (and in my opinion, not even a good thing most of the time).

AI has made me very weary about many things I interact with online. Was this image made with AI? Is this plugin vibecoded? Few things feel like genuine effort was put in to make it good, and it seems like people are far too willing to offload thinking and learning to some model (especially proprietary, cloud-based models).

My art, writing, code, etc. have probably been getting scraped as training data for AI. That’s not a fun thought for me, to think someone could possibly replicate my likeness with so little effort and use it to churn out low effort works.

It’s also just so easy to let the model do everything for you. It seems like they’re designed to make me stop thinking critically. Its outputs can be plausible enough for me to not see the mistake before it’s too late (especially in domains I am not knowledgeable in). So I have to exercise my due diligence to use them. I don’t always do that, but that’s why I keep anything made with AI private.

Ente’s approach to AI seems pretty reasonable to me. From what I can tell, their whole schtick with AI is that it is a tool that should be controlled by you (local and open-source), not owned by big corporations (Google, Anthropic, OpenAI, etc.). AI should also be completely opt-in and not intrusive like it is in most apps these days.

What to expect

There’s probably a bunch of other things I’m forgetting, but I don’t want you to see me as like an AI fangirl who thinks they’re the coolest thing ever. It’s a tool—one with cool tech behind it, sure, but there’s a lot of things I’m really worried about with how we’re approaching it.

I hope to update this page as I refine my view on it. I will always try to be open about how AI was used and where, because I think there’s good reason for me to be honest about that when this tech is so uniquely positioned to cripple my own thinking and expression.