If you wanted a vision of the future of autocomplete, imagine a computer failing at predicting what you’re gonna write but absolutely burning through kilowatts trying to, forever.
The whisper model has always been pretty crappy at these things: I use a speech to text system as an assistive input method when my RSI gets bad and it has support for whisper (because that supports more languages than the developer could train on their own infrastructure/time) since maybe 2022 or so: every time someone tries to use it, they run into hallucinated inputs in pauses - even with very good silence detection and noise filtering.
This is just not a use case of interest to the people making whisper, imagine that.
Ooooh that would explain a similarly weird interaction I had on a ticket-selling website, buying a streaming ticket to a live show for the German retro game discussion podcast Stay Forever: they translated the title of the event as “Bleib für immer am Leben”, guess they named it “Stay Forever Live”? No way to know for sure, of course.
It’s distressingly pervasive: autocorrect, speech recognition (not just in voice assistants, in accessibility tools), image correction in mobile cameras, so many things that are on by default and “helpful”
Now I’m curious how a protected class question% speedrun of one of these interviews would look. Get the bot to ask you about your age, number of children, sexual orientation, etc
It’s the risk:benefit tradeoff always. Sure you can hold on to all your illiquid stock in a private company with transfer restrictions, but will that pay for a house or even a banana? Does it ever not. The “take a loan for liquidity until you can sell some stock” trick worked for a little bit but companies are wise to that too now, and don’t allow it.
So people can pick between real money they can spend on stuff (or invest in a slightly less illiquid way) or being paper multimillionaires with no actual liquidity; not a hard choice imho.
I didn’t see this article here yet, but I just saw it elsewhere and it’s pretty good: Potemkin Understanding in Large Language Models