@benfulton@mastodon.london the first and most dangerous problem is and always has been the mothership attached to the clanker. Privacy is non-existent even if the company says they "respect privacy". CF LG shitshow if you need proof. Also a standard stt will make mistakes in the retranscription which will be easy to recognise and because it'll be a soundalike it'll be easy to deduce with great certainty what was said. Clankers tend to rephrase when they don't recognise something because they're configured in a way to hide their nature rather than produce the best possible outcome. That's even before the automated summary. The machine has no understanding of the dialogue, it's purely based on statistical inference meaning it'll miss important details and put way too much emphasis on stuff that isn't important. So, the nomenclature is irrelevant, the very concept of the system is the issue. Let's not even talk about biases in the training data or the horrendous way the training data is collected and all the other humanitarian and ecological issues, those are just the piss icing on the shitcake. Your thinking it's just the nomenclature and that people just hate abominable idiocy because it's trendy to do so really shows your lack of understanding of the issue and sends a clear signal you either don't give a fuck and don't care about others (then you're just an egoistical hypocrite) or you didn't inform yourself which would just make you naive until you learn about those issues. Of course failure to do so would put you in the first category. I truly hope you're just naive... For your own sake. @drahardja@sfba.social