I 90% agree, but also, there's a ton of informational & sensors & systems that need control. With physical knobs/dials/switches/buttons disappearing, the main console now has a lot more obligations than just infotainment like it used to, and it's not clear that letting a phone run the show is really viable with what's afoot, given the scope of systems the display has to control. Wait, sorry. It's not clear how we'd get sufficient data to the phone to let it try. We just don't have clear starting places to let the phone act as a good puppet-master over such an expansive complex multi-screened roving multi-functional device.
This is actually a very interesting ubiquitous & pervasive computing challenge. If we do want to let the phone be the main thing in control, it has to access & orchestrate a lot more systems than it has.
Back in 2012, the BMW/Land Rover/Jaguar folk (I think they were one entity then?) car-maker had started making interesting demos based on an already longstanding very interesting user-first user-sovereign ubicomp project, Webinos (https://en.wikipedia.org/wiki/Webinos, still a halcyon model of what connected-computing might have been (but with almost no uptake)), & were making really interesting API-based integrations, over effectively VPN networks directly to their cars. I don't remember whether things like HVAC control or lighting were integrated (I suspect so), but there were definitely a lot of examples of radar/lidar integration, engine information (tach, fuel remaining, battery voltages, et cetera). This >10 year old example is by far the most pro-user most open-possibility system we've ever done, by a country mile. https://www.wired.com/2012/10/bmw-webinos/
One of the most interesting things to me was just a couple short years latter, 2015, with the Jeep hack. The emphasis was that someone could gain access to your car network & do bad things, but this was the first time we'd ever gotten a real peak into a car network & it was fascinating. The often-underlying QNX OS it turns out- even though it is not Linux- runs a bog-standard-ish FreeDesktop DBus service bus, and all car systems are exposed over DBus. ALL car services. So like, one could fully automate & script their own car, via the Jeep hack (which IIRC typically required some physical access to break in via). It would be utterly trivial to script a bunch of nice lighting & sound effects, to roll back the moon-roof & dynamically set a mild thumping VU meter lighting, to dynamically roll the EQ low & high,... the possibilities were so open, over such a common well known easy to control system. The Jeep Hack was the most exciting look at what life could be like, but it was mostly used to sell Fear Uncertainty & Doubt, to insure even less people had access to cars.
Somewhere someplace sometime this trend needs to turn around. Somewhere someplace sometime we need to start figuring out how to pipe relevant systems to the user's agent. Rather than forever letting the car take-over more and more, forever reducing agency, forever shrinking what is possible.
It does suck that physical interfaces are going away, but I feel like there's still a real & interesting challenge to let people use their own user agent, rather than just become reliant on whatever premade jumble of systems a car happens to cook up for itself.
This is actually a very interesting ubiquitous & pervasive computing challenge. If we do want to let the phone be the main thing in control, it has to access & orchestrate a lot more systems than it has.
Back in 2012, the BMW/Land Rover/Jaguar folk (I think they were one entity then?) car-maker had started making interesting demos based on an already longstanding very interesting user-first user-sovereign ubicomp project, Webinos (https://en.wikipedia.org/wiki/Webinos, still a halcyon model of what connected-computing might have been (but with almost no uptake)), & were making really interesting API-based integrations, over effectively VPN networks directly to their cars. I don't remember whether things like HVAC control or lighting were integrated (I suspect so), but there were definitely a lot of examples of radar/lidar integration, engine information (tach, fuel remaining, battery voltages, et cetera). This >10 year old example is by far the most pro-user most open-possibility system we've ever done, by a country mile. https://www.wired.com/2012/10/bmw-webinos/
One of the most interesting things to me was just a couple short years latter, 2015, with the Jeep hack. The emphasis was that someone could gain access to your car network & do bad things, but this was the first time we'd ever gotten a real peak into a car network & it was fascinating. The often-underlying QNX OS it turns out- even though it is not Linux- runs a bog-standard-ish FreeDesktop DBus service bus, and all car systems are exposed over DBus. ALL car services. So like, one could fully automate & script their own car, via the Jeep hack (which IIRC typically required some physical access to break in via). It would be utterly trivial to script a bunch of nice lighting & sound effects, to roll back the moon-roof & dynamically set a mild thumping VU meter lighting, to dynamically roll the EQ low & high,... the possibilities were so open, over such a common well known easy to control system. The Jeep Hack was the most exciting look at what life could be like, but it was mostly used to sell Fear Uncertainty & Doubt, to insure even less people had access to cars.
Somewhere someplace sometime this trend needs to turn around. Somewhere someplace sometime we need to start figuring out how to pipe relevant systems to the user's agent. Rather than forever letting the car take-over more and more, forever reducing agency, forever shrinking what is possible.