tl;dr: If my name is in the byline here, on the BoundaryML blog, or anywhere else, then you're reading something that was written by me directly, not something that came out of an LLM.
I want to make basically the same pledge that my friend Anton recently made on his blog, and for basically the same reasons: I will avoid using LLMs or AI assistants for writing.
To reiterate his reasons that I agree with:
- Writing is a reflection of thinking.
- Readers want to know if they are reading my direct thinking, or some other thing possibly based on my thinking, but either shaped or even primarily invented by an LLM.
- There is more AI-assisted writing out in the world than ever before.
- So it's increasingly important to preserve and label the words that came from a human.
There are some exceptions:
- I'll lean in to using LLMs as coding assistants, because coding is not a medium where purity of original thought is the most important thing.
- Some documentation we write at work is written with the help of AI.
- I'll consult AI a lot when researching (chasing references back to the primary literature of course).
So for blogs like this one, the prose won't be as tight. But you can be sure it's me.