The send-to-the-cloud thing just exists because tech companies have a pathological fetish for recording, analyzing, and storing every single little thing you say and do and then trying to sell it to advertisers. Or train AI’s with it these days, or whatever the fuck else. The only marginal benefit you might get is that they can update their algorithms server side and not have to update your car or other device. But the technology has been mature for literal decades, so I don’t think that’s terribly important.
That said, I still don’t want my car to have voice control. It’s just as stupid as a concept as making everything touchscreen.
Speech to text is one thing. Actually understanding all the intricate details and variations of language is incredibly difficult. It’s good enough for some stuff, but I’ve yet to see a system a system that’s reliable enough for day to day use, especially in a car.
Scenarios like this happens way too often:
“Set alarm for fifteen minutes”
“Ok, setting alarm fifty minutes from now”
“No! FIFTEEN minutes”
“I’m sorry, I don’t understand what you mean”
“Remove old alarm and set it to fifteen minutes instead”
Indeed and it seems attainable now, if it weren’t for the expensive hardware and massive energy required for general pre-trained transformers. Don’t want my car to call home just to run a neural network on Azure, it needs to run locally.
I think using a car tablet is equally as dangerous as texting and driving. Voice control would actually be better for adjustments while driving.
Realtime non-cloud voice control is still unreliable. Gonna be a while before that can replace physical buttons.
I don’t want to have to talk to my car. Just have buttons and knobs. This shit was figured out 30 years ago.
More like 100 years ago.
Horseshit. My Pentium 133 could do it in 1997.
The send-to-the-cloud thing just exists because tech companies have a pathological fetish for recording, analyzing, and storing every single little thing you say and do and then trying to sell it to advertisers. Or train AI’s with it these days, or whatever the fuck else. The only marginal benefit you might get is that they can update their algorithms server side and not have to update your car or other device. But the technology has been mature for literal decades, so I don’t think that’s terribly important.
That said, I still don’t want my car to have voice control. It’s just as stupid as a concept as making everything touchscreen.
Speech to text is one thing. Actually understanding all the intricate details and variations of language is incredibly difficult. It’s good enough for some stuff, but I’ve yet to see a system a system that’s reliable enough for day to day use, especially in a car.
Scenarios like this happens way too often:
“Set alarm for fifteen minutes”
“Ok, setting alarm fifty minutes from now”
“No! FIFTEEN minutes”
“I’m sorry, I don’t understand what you mean”
“Remove old alarm and set it to fifteen minutes instead”
“Playing song on Spotify…”
Indeed and it seems attainable now, if it weren’t for the expensive hardware and massive energy required for general pre-trained transformers. Don’t want my car to call home just to run a neural network on Azure, it needs to run locally.
There’s Gemini nano which will run on phones locally, so I think we can have that soon enough