Post #3243830

2025-10-15 20:56 UTC

@rachaelspooky@cyberpunk.lol The text it generated is perfectly reasonable, even if longer than necessary. It also explicitely says that its there to help you write a text, not a complete "works every time perfectly" solution. Really, someone is making it easier for people to do the responsible thing of alt-text and you're decrying it because "it can make bad text". Yes, they know that, geez. @goat@fuzzies.wtf @0x57e11a@void.lgbt @40796c19@void.lgbt @micr0@wetdry.world

Replies (1)

@rachaelspooky@cyberpunk.lol 2025-10-15 21:09

@curiousicae@tech.lgbt @goat@fuzzies.wtf @0x57e11a@void.lgbt @40796c19@void.lgbt @micr0@wetdry.world thats not my point. even if altbot worked perfectly and made great descriptions every time i would still oppose it because that will lead to trust in other genAI. i dont want it normalized as a tool. and yes, all of this is just different kinds of fancy machine learning. i see the logic behind "so do you not like transcription?" but nah i love that. it's not generative ai. it's on the same level as ocr to me; it's a system presented with symbols that is asked to replicate them in another form. there is no approximation of creativity happening like with alt text generation. alt text by its nature, human or ai, introduces a framing and implicit context of a situation, and that is what separates it from things like transcription. thats what makes LLMs so bad to normalize even with well-intended or actually helpful examples. it preserves the framing and contextual choices of its training data. with the right circumstances, it can spew out harmful framing of ideas on a global scale. wanting to avoid contributing to this is why i oppose altbot.

Open ##3243831