Mercedes-Benz takes in-car voice control to a new level with ChatGPT
media.mbusa.com
media.mbusa.com
Does Mercedes Benz have advanced lasers to project them into your eyeballs or something?
I wish it had Bluetooth…
I said no and they tried to make me feel guilty that I am missing out on a great deal.
That being said, we have SBC powerful enough to run advanced assistants without any network connection.
Until that is offered, I say no.
Why would anyone voluntarily pay for internet connection in their car, when they likely already have it in their phone?
At least internet advertisers has the decency to give you something for free in return. This is level of "paying for Facebook".
I have voice commands in my Volvo, they've never worked right since day 1, and now the only time I press the button is annoyingly by accident.
They keep introducing short-lifespan tech features in cars now that nobody wants rather than just making them more durable and reliable.
So many buzzwords in one sentence
This is true but it still has to be built to spec and pass testing. Manufacturers can't substitute materials and whatnot to save cost. Having said that, the military might have different priorities than you are I would, most notably, they have to have 2,000,000 of a said widget and spec accordingly.
A good example is the Thompson sub machine gun, which was a commercial weapon, beautiful and well made, but a very expensive piece of hardware. It was replaced mid WW2 with the M3 submachine gun (Grease Gun), a useful, durable, but made of stamped steel and about 1/5th of the cost.
Judging from the enterprises I have seen this should be taken as a warning.
I hope other car manufacturers will follow soon. Their text to intent are usually very frustrating.
I really thought the personal voice assistant was dead (see Amazon below) but your comment makes me wonder if maybe these generative LLMs will revive it.
Amazon seems to have taken the tack with Alexa for which a generative LLM makes sense (a broadly-scoped personal assistant) and indeed seem to have dropped the ball. They also recently cut way back on their Alexa development and team so maybe they have realized that the personal assistant path isn't that useful, at least right now.
Google I don't know...whatever they do they will abandon, of course, but I suspect they will find a way to jam a generative LLM into their voice assistant thingy just because they need the branding. This is where your observation is particularly on point.
Apple junked most of that from Siri after they bought it and, given how poorly it was received, doesn't seem to have been particularly interested in the "personal assistant" whether in a standalone device or as a built in feature in their handsets. They seem to be (slowly) pushing their models into product rather than having a "primary entrance".
I want just a stupid electric car with a stick, pedals, speed/battery meter and parking sensors.
"You are a car computer, you control various car components and driver instructs you. I will give you instructions as the driver and you tell me what you will do to which components"
Oh, and it doesn't recognize my wife saying "hey, mercedes." As if it band-pass filters her voice pitch. ML won't solve your interaction problem, Mercedes, but I guess I appreciate the effort.
I find it really exciting - did a few demos of what's possible here for home automation here: https://youtu.be/BeJVv0pL5kY
I think people underestimate just how powerful this is.
I cannot speak for other people, but I certainly see how powerful it is, yet also very unreliable. And for some reason I want everything in connection with a car to be reliable all of the times and not only most of them.
Being able to say things like "make it cooler", or "play that song that has these lyrics..." will make life a lot more convenient.
Whatever driver-controlled functionality a car offers that can't be automated with a sensor (optionally toggle-able) and which isn't directly related to improving driver awareness, traffic signalling, or vehicle handling probably shouldn't exist in the first place.
For home automation you do you, of course! I can definitely see the appeal of this in that scenario even though it isn't something I'm particularly interested in.
aside from a future where completely offline LLMs are in the car, with over the air updates being done to the full model if its too crazy, maybe we could do dual LLMs. where a dumber, slower, offline one is there when connection is bad or not available.
No fuckin' thanks.
We live in the absolute stupidest of dystopias.
this is so true. When i think of "dystopia" i think of some post apocalyptic hellscape but I think a dystopia of constant irritation from the most stupidest digital policy imaginable is just as valid.
What's next, your car/assistant/phone overhearing you talk about voting for someone contrary to the approved party du jour and refusing to start the engine/play music/sms??
Likely referencing this recent Amazon incident, where a Microsoft engineer was locked out of his Amazon accounts after an Amazon delivery driver reported racist remarks from his smart doorbell (misreported as Ring some places, apparently a Eufy tho). Trusting the engineer, that didn't happen, no one was home or responded to the driver, and the Amazon delivery driver seemingly misheard a pre-programmed response from the doorbell.
But that just shows you don't even need to be listening to a song with a slur in it to worry about this sort of thing.
I disabled the voice assistant in my Mercedes because I don't really see the point in it, but your concerns seem kinda silly.
(Disclaimer: I'm an employee and it's an employee lease)
My concern is 'silly'. Your.. naivete? nonchalance? is why people never fail to be surprised when leopards eat their faces. Whenever I come across a news article like this, I always ask myself who's going to abuse it, and how are they likely to do it, and how much harder it's going to make daily life.
Voice assistants have to process small bits of audio in order to react to their wake words. Sometimes they react to non-wake words and record anyway. Amazon filed a patent in 2019 to record more audio before the wake word, to use more "natural language". Amazon also recently locked a guy out of his smarthome equipment because a driver accused him of being racist when he wasn't. Amazon also uses Roomba to make maps of people's homes, and I'm pretty sure somehow Roomba got private photos of a woman and those ended up circulating among Amazon staff.
Insurance companies use any excuse to raise fees or hoard money. Car insurance companies have the specter of self-driving threatening to wipe out their revenue. Of course they'd tack on a $50 "uncourteous driver" fee if only it was feasible to do so.
Speaking of feasible, Facebook doesn't need to listen through the mic, yet I bet if it was, they would do it anyway, because why not. Just like the NSA "inadvertently" collects metadata on US citizens, the majority of whom are neither criminals nor even suspects, because it's feasible and why not.
We've all seen articles about Google locking people out of their accounts for "automated" reasons, or Microsoft's horrendous labyrinth of services summoning Cthulhu and breaking everything connected to them. I'm sure anyone who's used any social media at all knows that innocuous comments can get removed for being "hate speech" or "spam", but actual hate speech and spam proliferates, and appealing does nothing, because there is no human accountability. Not knowing the "rules" of moderation, those "rules" being applied unevenly, and no human oversight to appeal has far greater negative consequences to an individual than it should.
Trends converge. Long run, only difference between China and US is that your social credit score will be privatized and decentralized in the hands of the hundreds of different companies you have to stay in the good graces of.