Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Apple Scales Back Its Ambitions for a Self-Driving Car

nytimes.com

Post reply on HN
Print view View original

341–350 of 434 posts

Re: Apple Scales Back Its Ambitions for a Self-Driving Car

#341

Post by dragontamer »

piyush_soni wrote: post #245
dragontamer wrote: post #230

Earlier quoted context omitted.

Capacitive screens were less responsive than resistive screens. The difference was that you used your physical finger instead of a stylus. Personally speaking, I preferred resistive screens for their responsiveness and accuracy. The capacitive screen was needed for "pinch to zoom" and other gestures of the modern era. The delay in the UI compared to resistive screens (as well as the higher-CPU time required for proce…

Is there some more information on that? I believed so far (or the industry made me believe) that capacitive screens were much faster than the resistive ones.

I don't have like, hard tested data. But the delay is enough that experience was all I needed to convince me.

Nintendo DS / 3DS and Wii U are all resistive screens. I also have experience with resistive screens of Palm Pilot and Palm phones of the early 2000s. And all of those were more "responsive" than a 2007-era iPhone, despite the iPhone's significantly higher processing power.

Play a serious game like Monster Hunter (very timing intensive) and compare it to a serious timing-intensive game on a capacitive screen (I know I've seen people play Marvel vs Capcom 2 on an iPhone emulator...). The lag and delay is noticeable even in the modern era.

But people wanted multi-touch pinch to zoom. Capacitive sensing also requires a microcontroller to literally "charge up" the screen over and over.

When your finger is on the screen, it "charges up slower" (because your finger changes the capacitance at that location). Innately, that "charge and discharge" cycle requires a measurement over time.

And to make sure that power isn't wasted, only one part of the screen is often tested (there are grids and stuff that then calculate where the fingers or multiple-fingers are located). So innately, there's a delay as the "scan-lines" of the sensors test each point of the screen.

Resistive screens instantly tell you where the stylus is without any such delay or processing. So it just makes sense.

Re: Apple Scales Back Its Ambitions for a Self-Driving Car

#342

Post by _jal »

jacquesm wrote: post #267
ptero wrote: post #262

Earlier quoted context omitted.

Apple often tries to re-engineer the thing they develop.from the ground up. Seems it started this way on Titan, too, then scaled back and decided to focus on the main problem: safely getting around. Nothing wrong with this I think. On reinvented wheels, early ideas rejected today may get a second life later, e.g. spherical wheels for moving in very constrained environments.

It is absolutely fine to re-engineer something from the ground up once you have a working example in your hand. But before then you are simply wasting your time. the iPod was not the first music player, the iPhone not the first smartphone. Re-engineering those was the only way to get them to market in a way that might drive mass adoption. As for the spherical wheels, yes, however such a development is totally tangent…

> But before then you are simply wasting your time.

I disagree. You seem to be discounting the learning aspect. Building an institutional knowledge base of the pile of systems that make up a car is a large task by itself. Different shops approach it differently. One may obsessively reverse engineer the competition, another might tinker with longtime designs.

I fully expect the managers in question saw the spherical wheels as something with maybe a 2% chance of becoming a reality, if that. The point at that stage is not to take a particular design to market, it is to explore the problem space, develop expertise, and work out what strategy to pursue.

From this article, it sounds like they figured out the strategy.

Re: Apple Scales Back Its Ambitions for a Self-Driving Car

#343

Post by bluGill »

icanhackit wrote: post #272
hueving wrote: post #190

Earlier quoted context omitted.

Because outsourcing car manufacturing doesn't make much sense. The transportation costs eat up most of the savings. That's why Toyotas, Volkswagons, etc sold in the US are made in the US. One shipping container holds 60,000 iphones. One shipping container holds one car.

> Because outsourcing car manufacturing doesn't make much sense. #1 It allows you to focus on design as opposed to supply chain management. #2 If the supply chain is centralised where production takes place then that bit of missing trunk carpet, as Elon Musk lamented, won't take two weeks to get to you and hold up the production line - It's down the road. > One shipping container holds one car. Which makes me think -…

Electric cars stick with the sedan/truck model for the same reason non-electric cars do: it is what people actually want. 90% of the time a car/truck only carries one person to/from work with nothing else. However that other 10% means the single occupant car will not work.

I have a car that gets 40mpg, and a truck that gets 18mpg. After you account for the extra insurance, license, taxes, and maintenance on my car, I'd be money ahead getting rid of the car and just driving my truck for everything even though times when I need a truck and a car will not do account for maybe 2% of my driving. Note that my car and truck a both paid for, if I'm making payments on both the math is even more lopsided toward not having a car.

People keep pointing out that I can rent a truck when I need on. However this is something is is true in theory, but not practice. It is actually hard to find someplace that will rent you a truck that allows you to use it. Want to take your rental truck off road - not allowed. Want to tow a trailer - most don't allow that. Want to put a sheet of plywood in the back - most don't allow that soft of damage. (though in this case the store probably has the best price and will allow it)

Re: Apple Scales Back Its Ambitions for a Self-Driving Car

#344

Post by nojvek »

pedroaraujo wrote: post #300
SeeDave wrote: post #294

Earlier quoted context omitted.

>The challenge of a self driving car is to get it to drive by itself, not to try to make it look good and to solve a whole pile of engineering problems that have had everybody in engineering looking at them since as long as we have the profession. This is why I believe that comma.ai was so visionary, say whatever you want about George Hotz's personality. He truly _gets it_ that self-driving is the critical part, not…

"Visionary"? Hardly. His self driving technology was basically a big giant hack. He is a good hacker but calling him "visionary" is a bit of a stretch.

A giant hack that works and already out on the streets.

I think commaai has a very bright future.

Re: Apple Scales Back Its Ambitions for a Self-Driving Car

#345

Post by sjm-lbm »

briandear wrote: post #337
amelius wrote: post #317

Earlier quoted context omitted.

True, but in this case it's a sugar coating without any cake below it.

[citation needed] — how do you know what Apple has or doesn’t have? People are making assumptions based on leaks — leaks, that wouldn’t surprise me if they were actually coordinated and on purpose designed to mislead the media. Five random Titan engineers wouldn’t just start talking to the Times. Not if they wanted to stay working at Apple. So it makes you wonder if their “leaking” is designed to mislead.

That is, of course, at least partially true.

However, in the OSX analogy above, Apple had a very good platform (NeXTSTEP) that had been commercially available and proven in real-world use for ~a decade before they started modifying it to become OSX. There is no similarly mature self-driving car platform for them to be working from, and we can be sure of that.

Re: Apple Scales Back Its Ambitions for a Self-Driving Car

#346

Post by Eridrus »

stevenwoo wrote: post #338
MarkMc wrote: post #167

Earlier quoted context omitted.

The article you linked to indicates that places with good weather and good roads could see the technology in 3 years. For all places in all conditions it may take 30 years: "this technology is almost certainly going to come out incrementally. We imagine we are going to find places where the weather is good, where the roads are easy to drive — the technology might come there first. And then once we have confidence wit…

Even in the Bay Area, there is periodically bad to less than perfect weather - we had 30+ straight days of rain about 15 years ago in one of the El Nino years, and that would have killed Waymo's steering wheel-less cars' useability. It's almost like none of the people who made that decision were here when that happened.

The Bay Area only has a lot of self-driving cars because the engineers are there, note that WayMo is letting people take rides in Phoenix, not the Bay.

Re: Apple Scales Back Its Ambitions for a Self-Driving Car

#347

Post by bluGill »

shandor wrote: post #325
ams6110 wrote: post #288

Earlier quoted context omitted.

If you're talking about the actual vehicle motion control systems, they probably should have been looking at what languages avionics companies use, not what their app developers are familiar with. You'll need languages and operating systems that are battle-tested (literally) in hard-real-time fault tolerant systems. Swift should never have been on the whiteboard for something like that. It's never been used in that s…

So what do avionics companies use? C, C++, Ada? Anything else?

Avionics companies have very complex rules. They can use any of the above, but generally a subset.

For example you might use C, but ban malloc, and recursion. There will be static analysis tools that look through your code and calculate exactly how much memory you are using using in the worst case of function call tree, that way they know they cannot run out of memory or stack.

Re: Apple Scales Back Its Ambitions for a Self-Driving Car

#348

Post by skgoa »

ams6110 wrote: post #281
ptero wrote: post #262

Earlier quoted context omitted.

Apple often tries to re-engineer the thing they develop.from the ground up. Seems it started this way on Titan, too, then scaled back and decided to focus on the main problem: safely getting around. Nothing wrong with this I think. On reinvented wheels, early ideas rejected today may get a second life later, e.g. spherical wheels for moving in very constrained environments.

I can't see spherical wheels working well. They would have a smaller contact patch on the road surface than a standard wheel, and consequently less traction.

Plus the whole sphere and its housing is going to take up a lot of volume.

Re: Apple Scales Back Its Ambitions for a Self-Driving Car

#349

Post by bluGill »

wellboy wrote: post #219
erikpukinskis wrote: post #52

Earlier quoted context omitted.

You're also buying huge liabilities: - constrictive agreements with dealers - rapidly deprecating manufacturing assets - a workforce with sunk costs in last generation skills - a warranty burden on old vehicles - pensions and other union agreements - a beloved brand that the public may not be happy seeing chewed up and regurgitated as iTaxi Supplier 4A If you believe in an autonomous and electric near future, some of…

I don't think these points are of any relevance. Once you have stage IV autonomy, you can partner with any car manufacturer you want, you have all the leverage. However, knowledge in cars definitely helps building the technology, since there are so many little variables that you otherwise would oversee.

You can try to partner. However all major auto manufactures are looking at self driving cars. Some are doing it in house, some are having and existing trusted partner do it. That minor auto manufactures have partnerships with someone major and have reasonable confidence that their major partner will provide the technology.

Apple and Silicone valley in general is not a trusted partner, in fact they are the opposite: untrusted. Car manufactures have lost millions of dollars on safety lawsuits. The cowboy coding culture silicon valley has a reputation (not always deserved) for doesn't work, they need controlled processes where they can show the courts they made effort to check everything, starting with the design phase.

Car makers expect that they will be sued over a 10 year old car with some parts failing. They need to show the court they did everything possible to anticipate that exact failure case and either ensure it couldn't happen, or if it they handle it. When someone is dead, "we didn't think about that case" doesn't cut it.

In short, I don't think Apple or google will get anywhere in self driving cars. They have early demonstrations, but that doesn't mean anything long term. It doesn't even prove they have an early lead (though I suspect they do) since car companies might or might not say anything about where they are. In fact in this case I expect them to be very careful not to say anything: if they say to much a lawyer might argue the car should have been self driving and thus the car manufacture is fault for an accident. As such they need to set expectations that self driving cars are a future thing that isn't ready yet.

Re: Apple Scales Back Its Ambitions for a Self-Driving Car

#350

Post by skgoa »

SeeDave wrote: post #294
jacquesm wrote: post #240

Earlier quoted context omitted.

This so reminds me of the discussion about fire from the Hitch Hikers Guide to the Galaxy. The challenge of a self driving car is to get it to drive by itself, not to try to make it look good and to solve a whole pile of engineering problems that have had everybody in engineering looking at them since as long as we have the profession. Making gadgets and making a self driving car require a different mind-set. The fir…

>The challenge of a self driving car is to get it to drive by itself, not to try to make it look good and to solve a whole pile of engineering problems that have had everybody in engineering looking at them since as long as we have the profession. This is why I believe that comma.ai was so visionary, say whatever you want about George Hotz's personality. He truly _gets it_ that self-driving is the critical part, not…

comma was a scam. The car Hotz used for his "demonstration" already comes with the functionality he "demonstrated" from the factory. He just added wires and a joystick to his dashboard.
Post reply on HN