Post by @mawhrin@circumstances.run

2026-09-20 08:19 UTC

RE: https://circumstances.run/@mawhrin/117302336573323503 think i'm exaggerating? the money put into the the autoconfabulator investments goes into fucking multiple trillions (order of magnitude: 10¹²). divide that by a million, and you get the order of magnitude of fucking millions. and instead we're getting things of the quality of the whisper model, famously trained on subtitle databases, and the extremely costly pervert glasses, neither of which is actually targeted at or half-useful for disabled people.

Replies (2)

@JigmeDatse@social.openpsychology.net 2026-09-20 10:38

@mawhrin@circumstances.run I'm not sure if video captioning is a case where it's not appropriate or not. I'd very much prefer to have it done by a human, but if I do it myself, it's something like 10 times the length of the video to do the manual captioning. Would it be better to just not caption, or is this a case that to you is justified?

Open ##5390252
@wronglang@bayes.club 2026-09-20 10:52

@mawhrin@circumstances.run is there an alternative to whisper for speech to text? I looked at the open source engines for a while and they look like a mess to created a dev env for, not even thinking about packaging? I wouldn't mind running training locally but just setting up for it seems like a nightmare edit: I agree disabled people are being used, I just don't articulate outrage very well and I want something to do about it

Open ##5390255