> Mr. Ive and his team of designers drew concepts for a car that would look like a European minivan such as the Fiat Multipla 600, which has a half-dozen windows and a curving roof. It had no steering wheel and would be controlled using Apple’s virtual assistant, Siri.
> One day, in the fall of 2015, Mr. Ive and Mr. Cook met at the project’s headquarters in Sunnyvale, Calif., for a demonstration of how the car might work. The two men sank into the seats of a cabinlike interior. Outside, a voice actor read from a script of what Siri would say as the men zoomed down the road in the imaginary car. Mr. Ive asked Siri what restaurant they passed and the actor read an answer, said two people familiar with the demonstration.
Tackling one of the hardest technical problems ever attempted, where it was not clear at all whether a solution is even possible (it is still not clear) they seriously discussed the script Siri would read.
I was going to say exactly the same thing. This is precisely the issue with UX-led processes. When you're dealing with very hard engineering problems, you really don't yet know enough to be designing a product.
I think it's very interesting that Apple went back and forth on the purpose of the project: were they building an electric car that people could use? Or a self-driving dream car? This is a critical question. Depending on which one you want to build, the process has to be completely different, no?
My outsider impression: most Apple groups excel as "productizers" of technology that has already been developed, but perhaps not as developers of technology themselves. I guess the chip making shop that produced the M1 and M2 would not be this type of group.
It depends where you draw the line in the tech stack. On the hardware side they've always worked at least at the same hardware level as any workstation or PC vendor, and often arguably lower than that working with component manufacturers to custom spec particular components.
They're an aggressive early adopter of new tech, putting in the work to adapt it to their product categories, sometimes directly funding tech development with partners to make that happen. They jumped on the tech for laser cutting aluminium laptop and mobile device bodies from aluminium blocks, gorilla glass for mobile devices, helped develop the specs and tech for retina and 5K displays, were actively involved in the development of open technologies like Wifi, USB, Firewire, etc. They've never been content to just buy parts off the shelf, pushing to promote the development of and get early access to new stuff other competitors didn't have.
On the software side they hired the best, including about half the former staff of Xerox PARC, the developer of the MACH microkernel, the developer of LLVM. You could argue that Apple didn't invent these technologies, but people at Apple did.
> This is precisely the issue with UX-led processes. When you're dealing with very hard engineering problems, you really don't yet know enough to be designing a product.
I suppose we can get into the weeds on how you define "designing a product," but under the right circumstances targeting a specific experience and working backwards to the technology has pretty clear benefits.
Ignoring Apple - because this kind of over-the-top process is just how they're wired - look at something like a modern automatic defibrillator. Literally any human being can use one with zero training because the goal was "make a defibrillator that someone can operate with one button" from the beginning.
If your goal is instead e.g. "make this existing defibrillator easier to use," you might still end up with a really good product but you probably don't have something that my mom could reasonably be expected to deploy if she runs across someone having a heart attack on the street.
>This is precisely the issue with UX-led processes
As opposed to engineer-led processes, which as we all know are infallible and always produce products that go to market 100% of the time :)
>I think it's very interesting that Apple went back and forth on the purpose of the project
This is the real reason it failed. The fact that nobody at Apple could say "this is what we're aiming to do" and then stick to that vision is more telling about Apple culture than it is about any UX processes.
> I was going to say exactly the same thing. This is precisely the issue with UX-led processes. When you're dealing with very hard engineering problems, you really don't yet know enough to be designing a product.
It's like the Wright brothers, before working on the Kitty Hawk, worked on mocks up of an online ticket ordering system.
That actually seems like a reasonable thing to do. Early-on going "assuming that we solve the technical issues with self-driving, can we make this into a product that works for us?" makes sense to me.
After all, if they found that they couldn't come up with a good experience for it, then they could stop throwing all that effort into solving the hard technical problems.
While I want autonomous cars to become available if for no other reason than to allow my elderly parents to continue using a car, I have serious doubts that it’s going to come about during their lifetime.
Having said that, I’m fully prepared to become their chauffeur at some point in the future. I’ve started looking at bigger vehicles I previously would not have been interested in so that I can drive them around in comfort and luxury.
Also curious about what are the Chinese EV manufacturers doing in this space? Are they putting much effort into autonomous driving, are they mainly focused on just making a plain old good electric car?
Chinese OEMs are putting a moderate amount of work into the space at L2-L3, as those offerings are rapidly becoming mandatory features for flagship product lines. There are Chinese teams working on L4, but it's difficult to get an good sense of their maturity right now.
I could be wrong, but seems only American and European (or just the Germans?) are aggressively pursuing L4+ while Asia (China/Japan/Korea) is focusing more on L2/3 for now?
No, there are a large number of Chinese companies that have active L4 efforts. Xpeng, Baidu, weride, pony.ai, etc. They tend to also have partial autonomy (L2-L3) efforts because that's what the Chinese OEMs want to invest in.
I knew a few people who worked for one the traditional car manufacturers back when Apple made the first big car announcement and apparently 'everybody' there sent in their CVs to Apple. The only people who heard back were the people from either design departments or who worked with electronics or software. People who actually work with engineering and building the actual cars, as in chassis, suspension, engines and other stuff made out of metal heard nothing. That was when it became clear to me that Apple had no real interest in ever actually building a car.
> Apple’s dead car project will be survived by its underlying technologies. The company plans to take what it has learned about artificial intelligence and automation and apply it to other technologies that are being researched, including A.I.-powered AirPods with cameras, robot assistants and augmented reality, according to three people briefed on the projects.
It's difficult to get a sense for what actually came out of the car project while it was running. But you can certainly imagine a case where a $1b/year R&D project, at Apple's scale, was giving them acceptable returns to feed into their other products.
e.g. a self-driving electric car presumably involved a lot of work on battery tech, augmented reality (at least at the world-parsing level), and interesting display tech for the windshield, all of which is heavily relevant for many Apple products.
An actual insider gossiping to tell us what directly came from the project would be interesting to see!
> The car project’s demise was a testament to the way Apple has struggled to develop new products in the years since Steve Jobs’s death in 2011.
One failed product is a testament?
I never really understood the Apple car, but I guess you'd consider it a high-risk, high-reward moonshot project. If that's how Apple saw it, then an unsuccessful outcome isn't even necessarily a failure since it's part of the strategy. In this case, being unsuccessful in developing an autonomous car wasn't the failure -- the road to that turned out to be much longer and bumpier than people thought at the time.
The failure was probably in letting the project live too long after realizing the project vision wasn't going to be achievable in a reasonable timeframe. The pivoting wasn't the cause of the problems but a symptom. Those probably should have been recognized as death throes, and the project ended sooner.
Yeah, worth considering that Musk has been regularly saying that full autonomous driving will be coming "in 6 months" since 2016-or-so.
Assuming that he and Apple were both equally confident in this being a problem solvable in the near-term, the main difference between them is that Apple didn't make any public statements about it... and then eventually gave up when that confidence didn't bear out.
|If it ever came to market, an Apple car was likely to cost at least $100,000 and still generate razor-thin profit compared with smartphones and earbuds.
I simply dont get this point. i believe they tried to go with a traditional automaker to get over the traditional design issues and manufacturing issues, even with that, a $100000 car would be a lot harder to chew without a steering.
I knew Tesaa went this way because , they didn't have the bank to break, but Apple had like $150 billion. even if they are just scratching the surface of ultra premium market, Did they tried to do all at once and failed?
Cant they simply buy a EV maker (like rivian as munster suggested) and do what they do best?!
I want a self-driving car. I'd love to never have to steer a car myself again.
I'd need to be very impressed by the technology to go along with not having a fallback steering wheel for edge cases, but I would absolutely want a car that could just do my normal city driving for me without me needing to engage with it.
I'd particularly like it if I could get such a car for my teenage child. Because let's face it -- you don't have to be that good as a self-driving system to be better than the average new teen driver.
It's not about the final outcome(which is hard to argue with, obviously everyone would want that), but I understand the sentiment in that I actively don't want a car equipped with all the tech necessary to enable self driving - because it's going to be extremely complicated and costly, not to mention a whole range of privacy concerns that I'm sure a lot of people here share. If we're talking about some ideal imaginary car that's 20k euro and can do self driving with all processing done locally with nothing uploaded to the manufacturer - sure, who wouldn't want that? But we all know it's not at all what the tech is going to be, not for a very very long time, if ever.
I actually completely agree with that. Self driving level 1 is useful, and self driving level 5 will be amazing (if we ever get it). Self driving level 2-4 I'm rather uninterested in.
I've long thought Apple's best play was to provide OEMs some kind of "VehicleOS" stack, enabling them to compete with Tesla. I doubt most OEMs can build the necessary stack in-house.
Then Apple sells some hardware, some software, takes a cut of some services bundles. A fresh new revenue source, without taking on the all risks of making actual cars.
"AI" doesn't necessarily imply general AI. Besides, it's well known that many of the things we find easy as humans are extremely difficult for computers and vice versa. That observation even has a name: moravec's paradox.
That’s what everyone is trying to develop. It’s early days for AI/ML/whateveryoucallit. We’re not at the 3D TV stage of things, where the tech works but no one wants it.
More interesting to me is that Apple is making noises about going all in on AI/ML but they decided it wouldn’t help their car efforts. And by car I mean self-driving.
With no training, actually. Anyone can "figure out" how to drive a car. Me and a friend started stealing his mom's car and joy riding at night at 13. No one taught us how to drive.
Thankfully, shockingly, no one got hurt and we never got caught.
Well no, not quite - you've had "training" in that you must have seen cars driving around and been driven in one, so you understood that it's a machine that moves, that is controlled with the steering wheel, and that you can make it move by turning the key in the ignition and moving that weird stick in the middle.
All of that is training and you'd need to teach an AI all of those steps too. Just because you didn't have proper driving lessons doesn't mean you didn't come to this with a set of assumptions learned by observation.
Well, you had 13 years of seeing your parents drive. Even if you aren't paying attention, it's probably enough to get the basic idea. The AI has never been in a car or seen someone drive before.
To people downvoting this comment... to be fair, this might be a response to some of the breathless hype surrounding LLMs these days.
There are people making wild claims about the capabilities of AI right now, and it's maybe worth pointing out some concrete examples of AI limitations.
The biggest problem with the word intelligence is it's not actually defined. Yep, we bandy about the I word quite a lot, but consider it as charged as the phrase 'freedom fighter', someone is going to take it much differently than you.
If we look in the animal world we can see 'intelligence' encoded in structures everywhere from single celled creatures all the way up to the brightest human being. This creates an obvious problem in definitions. It's like saying "I'm rich because I have money" and when someone asks how rich you say "between one penny and a billion dollars".
Any claim that says modern AI isn't "intelligent" needs to specifically claim "Human level intelligence at at common tasks" or it really is false, as again, intelligent has no commonly accepted agreed upon definition.
