2026-09-20 11:48 UTC
A good friend who's vibe coding now tells me I'm talking like a blind person about colours when arguing to reject generative AI without "really" trying it out first myself.
The issue isn't if generative AI can be useful for highly specific work and operated by experts for that task. Generative AI is a genuinely impressive and fun tool. Especially in prototyping and security fuzzing. But it still causes a deep feeling of disgust, at least for me.
Replies (2)
-
@hzulla@infosec.exchange 2026-09-20 11:49
The visceral disgust for the technology doesn't come from its obvious niche usefulness. But from its externalised costs we're told to ignore to embrace it. What do we have here? We have combined a massive problem space fuzzer with an addictive Clever Hans Effect machine. Tech bros are successfully lying to the public about it having magical capabilities.
-
@tasket@infosec.exchange 2026-09-20 23:09
@hzulla@infosec.exchange I see lots of reasons to reject what is being marketed as AI, but there is one reason that stands above them all: LLMs amplify bias, and the effect seems inescapable. The issue could be identity politics, but also things like coding styles, system architectures, threat models, etc.