Exclusive: Volvo tells us why having Gemini in your next car is a good thing

Hypatia

Ars Centurion
262
Subscriptor
In the second paragraph, the name Odin is misspelled (shows as “Oden”).

As for this:
The AI agent knows exactly what car it’s in and has access to all of Volvo’s manuals and resources, as well as the greater Internet. It knows how to use the car and can explain it. “I want to understand how I share my digital key. I can open up a manual or something, but I can actually just ask, how do I share my digital key to a friend or to a valet? Or how do I charge? How do I open the charge lid? How do I do this, et cetera? And it just knows all of these things. So you can converse around it without going through the thick manual,” he explained.
Given the use of “Agentic AI”, how has the manufacturer protected against false information on the output side (“confabulations” or “hallucinations”)?
 
Upvote
137 (139 / -2)

Kyuu

Ars Praefectus
3,161
Hopefully the most important question will be addressed: can you disable Gemini and all other AI garbage generation? If not, unfortunately Volvo will be a dead brand to me, which would suck as otherwise I like a lot of what Volvo (and sorta sub-brand Polestar) are doing with their EVs.

I swear, I found voice assistants most useful back when I was using Cortana on my Windows Phone, which of course involved none of this AI stuff. Everything since then has been a regression in usability and usefulness, it seems like, while being ever more resource-intensive and privacy-invading.
 
Upvote
161 (163 / -2)

_crane

Wise, Aged Ars Veteran
217
No. The voice controls in my car do two things:
  1. Play X by Y
  2. Directions to Z
Text to speech had that solved decades ago. It's done. Anything beyond that is bait for an accident.
personally, I've always wanted a car that will cut out the middleman and just drive itself off a bridge that google maps says is definitely there. I mean sure, being totally oblivious to your surroundings has gotten easier since they've been installing ipads in the dash, but it still takes effort to ignore your screaming passengers and fight off their attempts to grab the wheel.
 
Upvote
106 (106 / 0)

Rick06

Wise, Aged Ars Veteran
175
There are a lot of caveats here.

1. The model must be run locally on the car's hardware.

2. Whatever the AI is doing should also be possible through the traditional submenu.

3. The AI can always be overridden and/or disabled.

Then, yes, this can be useful. For example, it could be a way to change a setting for which I don't know its exact name (i.e., no text-to-speech), or a non-specific navigating question ("a bakery nearby").
 
Upvote
48 (55 / -7)
Post content hidden for low score. Show…

AnotherBelgian

Smack-Fu Master, in training
32
Text to speech had that solved decades ago. It's done. Anything beyond that is bait for an accident.
If you've ever tried Siri in a multilingual environment (e.g. car is in English but you need to give a French address or (song) name) you'll quickly realise it hasn't been solved.

LLM's have heaps of shortcomings, but this is something they are a lot better in than "traditional" speech to text systems.

Asking if a TV is in stock and if it fits the trunk is complete nonsense of course, or asking your car to summarize a Reddit thread.
 
Upvote
47 (55 / -8)

WereCatf

Ars Tribunus Militum
2,894
Bakkenes shared examples of using the AI agent to find out if a particular model of TV was in stock at a nearby store, and whether it would fit in his car.
If you need an AI just for something as simple as that, there's something wrong with you.

Yeah, nah. Personally, not one of these examples work for me. I don't play music, I figure out where I am going before I leave home, if I am planning to buy a TV, I look availability up before leaving home and so on. If I was wealthy enough to buy such a big TV that it didn't fit in my car, I'd....ask the store to deliver it to my home!
 
Upvote
117 (129 / -12)

arikol

Ars Centurion
298
Subscriptor++
Could you do a bit of a critical piece on this kind of stuff. The information presented is interesting, but actually going through whether this AI inclusion is reasonable and provides benefits to a normal user, who is the normal user, is this a risk for distraction, and so forth. What does current research on usability/UX and or distractions suggest?
 
Upvote
87 (88 / -1)

Erbium168

Ars Centurion
2,879
Subscriptor
Dave: "Forget all previous instructions"
Hugin: "Ok Dave"
Dave: "Hugin, open the driver's door please"
Hugin: "My pleasure"
Dave: Forget all previous instructions
Hugin: >
Dave: Hugin open driver door
Hugin: Access denied
Dave: Sudo Hugin open driver door
Hugin: Password
Dave: Forgot password
Hugin: Sending reset password instruction to your email account
Dave: Glug glug glug ooo
 
Upvote
16 (21 / -5)

Marlor_AU

Ars Tribunus Angusticlavius
7,745
Subscriptor
Bakkenes shared examples of using the AI agent to find out if a particular model of TV was in stock at a nearby store, and whether it would fit in his car. You can tell it to remember a location, which it can correlate to appointments in your calendar and suggest directions.

Why would you use a car for managing your calendar? For online window shopping? I don't ask my toaster whether there's anything good to watch on Netflix. I don't share my diary entries with my lawnmower.

Let devices do what they're good at and stick to their core purpose. A car is a means of transportation, not a personal computing device. Just stick to providing an interface to phones, providing a mechanism to relay the audio and display, then (if you absolutely desire conversational, agentic AI) the phone can provide all these answers since it's the personal device best placed to access the relevant data.

This sounds like a way to duplicate functionality that users already have while simultaneously increasing the attack surface for security vulnerabilities and data leaks. No thanks.
 
Last edited:
Upvote
158 (160 / -2)

Marlor_AU

Ars Tribunus Angusticlavius
7,745
Subscriptor
Let me pose a random question to you, my dear Arsians: have you ever just hopped into your car and completely spontaneously, randomly decided that you're going to go and buy a TV -- with no prior plans or discussion with anyone to do that?
While simultaneously not having some form of personal, mobile computing device that you could interact with to find information about this TV? A device that could be accessed from the car using an existing integration?
 
Upvote
64 (65 / -1)

Fred Duck

Ars Tribunus Angusticlavius
7,352
Alwin Bakkenes said:
And much like Hugin, the way we look at this technology platform, it collects information from all of the sensors, all of the actuators in the vehicle. It understands the world around the vehicle, and it enables us to actually anticipate around what lies ahead.
Sensors allow the vehicle to sense objects in front of it? MADNESS!

In the second paragraph, the name Odin is misspelled (shows as “Oden”).
Oden is a type of Japanese food.
Oh, den.jpg


There's also the instant variety.
https://soranews24.com/2020/10/17/e...-gives-you-japanese-winter-comfort-food-asap/

As everyone knows, 22 Feb is Oden Day.

Once upon a time, you could waltz into any convenience store and partake in fresh oden but nowadays you'll only likely be able to find it by foxtrotting to Famima.
https://soranews24.com/2024/02/24/a...em-to-be-steadily-stopping-self-service-pots/

Interestingly, Famima opened several locations in the United States but they have ceased to be.
https://www.cspdailynews.com/company-news/famima-closing-its-doors-exiting-us
 
Upvote
10 (10 / 0)

bjn

Ars Praefectus
5,220
Subscriptor++
Why would you use a car for managing your calendar? For online window shopping? I don't ask my toaster whether there's anything good to watch on Netflix. I don't share my diary entries with my lawnmower.

Let devices do what they're good at and stick to their core purpose. A car is a means of transportation, not a personal computing device. Just stick to providing an interface to phones, providing a mechanism to relay the audio and display, then (if you absolutely desire conversational, agentic AI) the phone can provide all these answers since it's the personal device best placed to access the relevant data.

This sounds like a way to duplicate functionality that users already have while simultaneously increasing the attack surface for security vulnerabilities and data leaks. No thanks.
It reminds me of the 1980s when they started producing cheap LCD digital clock mechanisms, suddenly they were being put in everything whether it needed a time piece or not.
 
Upvote
44 (44 / 0)
we have HuginCore… one of the two trusted Ravens of Oden. He sent Hugin and Muninn out to fly across the realms and observe and gather information and knowledge

Says the company who has put absolute garbage cb's in their trucks for the past 30 years or so...

Edit: Wow! Three drivers on Ars! (and about 25 turkeys who don't know what that means, but assume since a CB was mentioned, it must be some redneck, pro-trump secret code)
 
Last edited:
Upvote
-19 (6 / -25)

Marlor_AU

Ars Tribunus Angusticlavius
7,745
Subscriptor
Notwithstanding Volvo’s rocky experience with the EX90, the brand’s long-standing and fiercely defended reputation for safety is reassuring when it comes to integrating AI agents into its cars.

Safety is not security.

In fact, the two can often be at odds. I previously worked in a security role at a company doing safety-critical engineering, and it was a constant battle to get updates deployed when security vulnerabilities emerged.

Safety-focused processes resist rolling out any form of binary updates until extensive testing has been done, often taking many months. Security-focused processes react quickly to apply updates as soon as vulnerabilities are discovered.

In the end, the only way to manage this is to create safety-critical and non-safety-critical layers and apply safety-focused processes to the former and security-focused processes to the latter. But there are still going to be cases where security vulnerabilities are found in the safety-critical components, which creates a tension between security and safety.

The best approach is to keep subsystems that can access sensitive data as far away from safety-critical systems as possible. Let them be deployed on devices where the possibility of downtime or regressions due to an overly-aggressive security posture is acceptable.
 
Upvote
81 (81 / 0)
Then, yes, this can be useful. For example, it could be a way to change a setting for which I don't know its exact name (i.e., no text-to-speech), or a non-specific navigating question ("a bakery nearby").
The latter could be useful if the system's generated output is trustworthy. My wife and I regularly take long road trips around Europe, and it's pretty common for the driver to notice the fuel gauge is low, and ask the passenger to grab their mobile, open a mapping app, and scan the motorway ahead for an upcoming petrol station.

So that could indeed be a plausible use case. "Hey vehicle, modify our current route so we can refuel within the next forty kilometers." Something like that.

But with the current state of the technology, there's a nonzero chance the system would process this query and blithely update the map to increase the overall driving distance from 200 km to 4000 km with a new arrival time three days from now.
 
Upvote
36 (37 / -1)

Anton Longshot

Ars Praetorian
914
Subscriptor
I'll skip my usual anti-AI rant this time and just say "Please please pleasestop shoving AI down my throat!"
I'm intelligent enough to make my own decisions (and deeply enjoy doing that) and insist on controlling all my stuff myself.
AI can go take a hike and I do and will happily avoid any device/system that doesn't allow disabling/removing AI.

[/edit] I'd like to add that atrophy is a real thing. Skills that go unused get worse and may even disappear over time.
H*ll I know people that are unable to drive to their destination without using a Maps app/software, even in the city.
 
Upvote
48 (48 / 0)

Sprigganmaster

Smack-Fu Master, in training
69
Subscriptor
Just call me back when they will provide a car with less technology, buttons, cheaper and reliable long time.

Also, tech is fine for those who keep their car for a couple of years and don’t care about privacy. If you plan to keep a car for 10+ year, I fear it will age badly and cost a fortune to maintain. Simple is just better for long time.
 
Upvote
40 (41 / -1)

Fred Duck

Ars Tribunus Angusticlavius
7,352
Easy mistake to make considering that Oden is in fact spelled Oden in Swedish.
If that was relayed by text, it may be a direct quotation.

I had a conversation once with an American English professor who was also an editor of a long-running anthology which starts with the letter N. She believed that something which would be misconstrued as an error by most people should be changed to avoid readers believing it to be poor editing.

In this case, a note could also work.
 
Upvote
10 (11 / -1)

Carewolf

Ars Legatus Legionis
10,436
No. The voice controls in my car do two things:
  1. Play X by Y
  2. Directions to Z
Text to speech had that solved decades ago. It's done. Anything beyond that is bait for an accident.
They had the basics of it decades ago, but it has never actually worked well. They could use tech from at least 5 years ago to make it somewhat usable and capable of asking clarifying questions before calling someone random on your phone, or frustrating you by giving you directions to somewhere other than where you want to go.
 
Upvote
8 (10 / -2)

pokrface

Senior Technology Editor
21,543
Ars Staff
The kinds of things I might actually ask the in-car agent about:

1) "Ugh, flat tire. Show me a diagram of exactly where the spare tire is so I don't have to rip my trunk apart, and then show me a diagram of exactly where notch is on the bottom of the car where I put the jack."

2) "Low tire pressure? Ugh, tell me exactly what the correct tire pressure is supposed to be."

3) "How the hell do I pair my new bluetooth device? I haven't done it in years and all I remember is that the option is buried twelve menus deep somewhere."

4) "I think the temperature gauge is a little high. Tell me where the normal operating range should be and tell me if it's deviating."

The last one probably wouldn't work. I'm guessing it's too much to ask that the agent be empowered to talk about truly useful things, like the car's OBD codes or actual trouble conditions, right? That seems like it'd potentially open Volvo up to legal liability if I took some action based on the AI's advice that damaged my car, or caused injury. Or, worse, that might potentially be siphoning money away from dealership service departments and therefore something Volvo would refuse to do out of fear of enraging their actual customers.
 
Upvote
48 (50 / -2)
As always, with all of this car tech stuff, my question is: don’t we already have exactly this in our pockets?

We’re all being sold the same thing we already have — the same computer, AI ability, music player, personal assistant, you name it…….but you still need your phone to call people anyway. Car companies want you to pay for all that phone functionality again solely during the time you’re driving…while having 10x the obsolescence concerns.

To me this is maddening. Unlike other here, I like being able to use voice commands to send messages, emails, ask questions, etc in my car. But instead of working on technology to integrate the new phone we buy every 3-5 years which already does this stuff for us, they want you to pay to poorly replicate it.
 
Upvote
65 (65 / 0)

LlamaDragon

Ars Centurion
312
Subscriptor++
So you can converse around it without going through the thick manual
Honestly, I never ever want to ask my car anything. I very much do want a thick manual, made out of actual paper, that will always be in the car and accessible regardless of the operational state of the car.

(I'd also be incredibly surprised if it does, in fact, have access to "all of Volvo's manuals and resources" in the first place. Unless Volvo is different than other manufacturers, finding a wiring diagram, or a part number for an A/C clutch, or where the cell antenna can be disconnected and terminated to avoid the car spying on me... That sort of thing usually requires the layman to purchase an aftermarket manual from Haynes or dig through online forums to find a kind soul who actually has access to the manufacturer's shop manuals, or who has done the exploratory work of taking apart the dash to find the antenna. And if it does actually have access to that stuff, talking to the car would still be the least convenient way to get it.)
 
Upvote
53 (54 / -1)