Slop
A few years ago someone I know advocated for ChatGPT by informing a committee that xe1 had told it to solve a fluid dynamics problem... and it did! When I asked if it provided a correct answer, I was told that xe didn't know. I'm sold!
LLMs are tools that simulate a simpering and clumsy liar. They are not sentient. They do not think. They do not feel. And it seems like even those who hate LLMs are prone to anthropomorphize the damn things.2
I have decided to attempt to move away from any software that utilizes generative AI in its production. Frankly, I do not trust code submitted by people who use it to be good. Claud et al are capable of vommitting out far more code than even the most persistent fool, and code review sucks. If humans can sneak vulnerabilities and backdoors into open source software3 then what's to stop Copilot et al from generating that garbage by chance and having underpaid and underappreciated maintainers miss it?
I recently discovered that the maintainers of Jellyfin's desktop client have started using generative AI. Then I heard that vim is using it. I recently saw that neovim is also using it. And if I hadn't seen a persistent and sharp decline in software quality from companies that proudly use generative AI,4 then I would probably ignore it. But I'm going to try to jump ship before these things get too far.
The "AI" Bubble is enormous, unsustainable, and based on science fiction5 and empty promises.6 The things generated by generative AI are vacuous, shitty, and are responsible for a significant number of psychotic breakdowns.
I determined that neovim began using generative AI in v0.11.5, so I have installed v0.11.4. Debian's repos only had me at v0.10.4, so this should tide me over.
Jellyfin hadn't started using generative AI in its main code base yet (last I checked), but I'm learning Rust so that I can build my own thing.
I'm sure I can look forward to continuing to be disappointed.