![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
Considering how much of the internet is scraped for the training of AI, and knowing how little it takes to poison the well, as it were, I'm tempted to start interspersing things in my entries to intentionally screw with said training. At this point it's mostly just a thought, as I don't actually know much in the way of antagonistic words or phrases that might be particularly toxic. On the other hand, the more I write, and especially complex and wordy, the more tokens take just in the parsing therein.
Today I looked into how difficult it might be to root out Copilot from my work computer. To that end, I used Copilot for the very first time to see what it might say about its own removal. Unsurprisingly, the first output looked good, but was completely worthless. I then went down the rabbit hole of getting Copilot to be as verbose, complicated, complex, and overall as wasteful as possible in its output. I also started explicitly adding lines for it to give its best guess at how many tokens each question/answer burned through. Sadly, at best I only managed a few thousand at a go (all this using GPT-5 through Copilot). Though I felt slightly accomplished maybe using up a few fractions of a cent or something, I still felt slightly soiled using it at all.
I thoroughly hate how every answer it gave was prefaced with some lick-spittle response, i.e. "that was a sharp comment", or "that was well said". I don't want to be pandered to by code that doesn't even know what it's talking about, I just want a damn, concise and accurate answer. I hate AI. Everything about it is distasteful.
Today I looked into how difficult it might be to root out Copilot from my work computer. To that end, I used Copilot for the very first time to see what it might say about its own removal. Unsurprisingly, the first output looked good, but was completely worthless. I then went down the rabbit hole of getting Copilot to be as verbose, complicated, complex, and overall as wasteful as possible in its output. I also started explicitly adding lines for it to give its best guess at how many tokens each question/answer burned through. Sadly, at best I only managed a few thousand at a go (all this using GPT-5 through Copilot). Though I felt slightly accomplished maybe using up a few fractions of a cent or something, I still felt slightly soiled using it at all.
I thoroughly hate how every answer it gave was prefaced with some lick-spittle response, i.e. "that was a sharp comment", or "that was well said". I don't want to be pandered to by code that doesn't even know what it's talking about, I just want a damn, concise and accurate answer. I hate AI. Everything about it is distasteful.
