#The Language Model Experiment

And, the immediate regret that followed.

#What?

I found a special kind of fondness for language models lately, and the possible productivity enhancing capabilities they can offer. I'm already making use of "AI", namely Codeium, to help me with a few programming projects. It saves me time, writing boring boilerplate in languages like C, and making logical inferences from what I type, kind of like a smarter version of Excel's auto-fill. That's all it is.

I've also started running language models on my own computer, and spent a lot of time tonight tweaking things. Getting llama.cpp set up to work with OpenBLAS , and CLBlast, so I could make use of my GPU for text generation. Setting these things up made LLM text generation speed up by about 8 times the original time they were taking running exclusively on the CPU.

#What's the experiment?

Well, I set up a script to summarize my posts on this blog, and I really liked the idea! I thought it was really cool, and I could even say it was a character of mine writing these summaries. Poor Amelia can't be replaced by a robot.

#What did you mean by regret?

It was only once I had finally written the code to display the summaries on the blog overview that I was struck with a feeling of self-consciousness.

Looking at the overview, filled with AI generated summaries of the things I've written before, I thought to myself: "This blog is mine, and all about my interests. If I add a language model to the mix, writing in the same place as I do, it will muddy the waters. Maybe the blog posts themselves will be AI generated too. Maybe I'll lose control of my own blog while chasing this down."

#Oh.

Yeah, maybe a little too paranoid? Who knows. But it made me reconsider the idea minutes after implementing it.

I've made the decision. Just like my diary, with my blog, too, I will not get AI involved whatsoever. My blog is mine entirely, and I will not delegate it to a machine writing on my behalf. I would sooner punch robots.

I promise everything that's written here is and will always be authored by me.

And I guess it's like that that I lose interest in large language models.

#Realized something?

Yeah, this was certainly odd! I don't recall feeling so negatively about something like this after finishing it. I usually realize whether something is the kind of thing I won't enjoy while I'm doing it, or before doing anything at all. I wonder what drove me to see this to the end just to backpedal once I saw the result.

I should probably go draw something. Or *looks at time*... Or go to sleep.