Hacker News Reader

Comment by GhostVII | Hacker News Reader

Doesn't seem particularly notable, they got rear ended at 15 miles an hour while essentially stopped.

The important part is that it got rear-ended while waiting to merge --- I suspect a strong factor in this was the fact that the car behind it was expecting it to go and merge, but it suddenly stopped instead.

Even for a human, merging into flowing traffic is tricky and one of the more difficult maneuvers; combine that with a (not obviously?) self-driving car that doesn't behave quite like human drivers expect, and I can see how this could happen.

If I saw a clearly-marked "autonomous test vehicle" I'd probably stay far away from it, but I suspect this one wasn't. Imagine the surprise of the human driver who thought the car in front that he hit was just driven by another idiot...

I wonder if it would be more useful if these cars had more lights than normal cars (yellow, green, etc.), so that they could signal their intentions before they act.

Looks like it happened here: https://goo.gl/maps/RHiVTRSuDJ62

Pretty short merge lane, more of a surface street than a highway. Probably not atypical for a car to stop to wait for an opening to merge.

Lawrence Expressway is terrible. I frequently see cars stopped at these "merge lanes".

But that’s what you’re supposed to do here, right? There’s a yield sign.

You're not supposed to stop at a yield sign if there isn't traffic preventing you from merging. That's why it's a yield sign, not a stop sign.

Since the report mentions Apple's car was moving slowly (~1mph) when it got hit, I suspect what could've happened is it may have just released the brakes and was about to accelerate into the merge, meaning the car coming from behind would see both a clear path in front of it and its brake lights off.

There's a comment here about the car behind following too closely, but if you look at regular traffic and your own driving behaviour, you'll find that you subconsciously monitor and anticipate the behaviour of the cars around you --- meaning that almost everyone is technically "following too closely" to expect the car in front to suddenly and rapidly brake to a complete halt. Self-driving cars are far less predictable in that sense. If you come up behind a car in a merge lane with its brake lights off and no traffic obstructing it, you would not expect it to be slowing down or even stopped, because your past knowledge has taught you that a car in this condition is almost certainly going to be accelerating.

Computer controlled cars should have some sign or way of easily telling that they are computer controlled. Maybe if this was true, too many people would fuck with them, unlike those cars that say "student driver". Maybe the sign should say "student driver" when it is controlled by a computer. Definitely a true statement at with the current tech

People often fuck with student drivers too. Eg aggressively trying to overtake because they don’t want to be behind a learner. Although I suppose this is a different sort of fucking with than the kind you mean.

One could hope that the number of always recording cameras and other sensors on auto-auto would deter most of the jerks.

> You're not supposed to stop at a yield sign if there isn't traffic preventing you from merging. That's why it's a yield sign, not a stop sign.

That's what I meant; you're supposed to stop if there's oncoming traffic. That's what I assume the car was programmed to do. I have no idea what the other car was doing or what it was thinking–there's always the chance that it was someone who was distracted.

That’s a decent length (there’s more behind the camera) with a yield at the start of it. It’s basically an acceleration ramp, lengthened probably due to the speed of expressway. Stopping after the yield sign shouldn’t be happening for either autonomous vehicle or human driver; although it happens.

I admit the camera perspective can be misleading, but it looks like what, 4-5 car lengths? If traffic is heavy it’s not enough space to accelerate up to the speed of passing cars.

If traffic gets heavy, I can easily see the need to stop and wait for an opening.

I’ve been on that ramp, and it wasn’t too bad. Usually the speed of traffic isn’t that high due to the stoplight just before it, and people usually just wait at the yield until it turns red.

If traffic gets heavy, drivers should wait at the yield until it is safe or until the light turns as the sibling poster suggests.

I often get on to Lawrence at this particular intersection, using this merge lane. The traffic you're merging in to can be going at 50 MPH in the right lane with about one car length separation between cars. A human merging here has their head craned back, watching for a gap to appear, and accelerates to merge in to the gap. Who knows what exactly happened here, but this one is a particularly tricky merge under traffic.

I feel like Santa Clara county has a bunch of these dangerous merging lanes.

It’s pretty obvious when you see them - they’re covered in hardware.

That hardware doesn’t signal to other drivers.

When autonomous vehicles register problematic situations, they do tend to be unusually cautious, right? But so do some people, so hey. And when you get rear ended, you're rarely at fault. Because the other driver was obviously following too closely.

> And when you get rear ended, you're rarely at fault.

It was my understanding that in California, it is (legally) _always_ the rear-ender who is at fault, and not the rear-endee.

It can happen after a sudden lane change where the person that’s rear ended is at fault.

My impression was that was the way most if not all states were set up.

Ummm, that's what I said :)

"The rear-ender is always at fault no matter what" is largely myth. The truth is just that historically it's been very hard to prove you weren't at fault if you hit someone else in the rear. Now, with dashcams becoming more common and people becoming more aware of deliberate-crash scams and road-raging "brake check" incidents, that's starting to change.

This is a common scenario for human-on-human crashes too. The second person in line starts scanning the oncoming traffic early, assumes the first driver has gone, and accelerates into them.

I was rearended in this same situation as well, seems like a common error in human drivers too.

Human drivers are fucking awful. I don’t understand why everybody here is assuming that the self-driving car must have done something strange to cause this.

This, but also, human drivers do similarly awful things like stop at the end of a long acceleration on-ramp when they should have been accelerating to match the speed of traffic they are attempting to merge into.

It's biased because all commenters are human drivers who think they can handle that kind of situation. We often don't realise how much risk we take on every day while at the wheel.

What percent of drivers are involved in a collision while merging? I imagine it's miniscule. Rarely deadly.

I banged up my mum's car doing exactly this the first year I got my license. Was a little nervous and in a hurry, saw an opening and went for it. Car in front of me did not see the same opening. Bang.

It’s also something I would expect self-driving cars to have a huge advantage over human drivers, simply because the former can check behind and to the side while still checking the car in front of them.

The notable part is that it appears to confirm that Apple is testing self-driving cars.

Is this the result of the old, "Every company that gets big enough also becomes an investment firm." or does Apple really think autonomous cars fit into its portfolio of technology?

Every company that gets big enough becomes a VC firm, which is natural when you've got a firehose of money and don't want to miss out on the Next Big Thing.

If Apple can take over transport and wall it off for somebody then it just further cements their high ground

I fully expect that you won't be able to tether an Android (or any open specced) phone to an Apple car, like iMessage and Apple TV, and they'll use further cement their grip on existing customers

It only mattered in the last couple of years that you could buy a car that supported Android Auto. So as long as Bluetooth works, I think that gets you 80% of the way there. Maps seems to be the only real UI app allowed anyway, and hopefully Google releases a version for iOS 12.

I can’t imagine Google NOT releasing that update. Ever since CarPlay first came out people have been complaining about this and blaming Apple.

When iOS 12 was first unveiled a huge chunk of the comments I saw were people happy about this exact feature.

Now that there’s absolutely no reason Google can’t do this… it seems moronic said they would choose to hold the feature back.

What would the benefit be? They wouldn’t get the data collection because the app isn’t running, and how many people would realistically switch to Android over this single decision?

When it comes to the Apple car I guess part of the question is… can people even buy them? It’s not as much of an issue if it’s a fleet car or a cab kind of situation.

It’ll be very interesting to see what Apple does if that day comes.

I actually think cars are a good fit for Apple's strengths. Apple is good at user interface and hardware engineering.

1. User Interface

You'd be hard pressed to find an object which most people own that has a more complicated UI than a car. Steering wheel, pedals, buttons everywhere. Probably well over 100 separate features. I think there's tons of room for Apple to create a better experience here.

2. Hardware Engineering

Cars, much more than personal computing devices, rely on hardware engineering. Granted, most of the big engineering issues in cars have been handled, but there will be plenty of new hardware to be engineered as cars transition to full autonomy.

Sleek UI and hardware is nice, but for autonomous vehicles AI is essential. You can have the best car in the world in other respects, but without AI it's not self-driving.

Apple hasn't shown that it's particularly good at AI or mapping. Siri and Apple Maps are distant runner-ups. If it weren't for name brand awareness, I doubt anyone would be paying attention to Apple's self-driving project versus the field of competitors.

Apple may have enough money to brute force their way to a spot in the rankings, but it hardly seems like a natural fit for them.

We are a long way off from 100% autonomous cars. In the meantime, human-car interaction is ripe for improvement.

And regarding Apple Maps and Siri: we'll see what the next few years holds. It's always useful to keep in mind that Google had a huge head start. There's a difference between being incompetent and being behind. Apple is rolling out new maps this fall; we'll see how they do. Remember, there was once a time when Apple didn't know how to do web services. I think the last few years has shown that they figured out whatever it was they needed to figure out.

Generally speaking, there probably isn't another company that is better than Apple at engineering management. Google and Amazon may be better in some areas, but they're probably also worse in others. Remember, this is a company with the best mobile processors in the world—by a long shot—despite not having even been in the processor business 10 years ago. They also deployed a brand new file system to a billion devices with basically no issues.

We're talking Apple's self-driving car project, though. It's in the title. There's no indication Apple is trying to get into non-self-driving cars. Nobody said Level 5 (100% autonomy) is necessary, but presumably they're at least aiming for substantial autonomy, e.g. Level 4 as others are pursuing.

I'm not sure what you're referring to with regards to Apple web services. People outside the Apple ecosystem are not using iCloud, and even many Mac users prefer alternatives.

Google did have a huge head start on AI and maps, true, but we're not grading on effort here.

If push comes to shove they can always buy the AI from a third party like Bosch.

> You'd be hard pressed to find an object which most people own that has a more complicated UI than a car.

I dunno, just spitballing here but...a VCR? Heh...or a TV (remotes have lots of buttons!), or a computer perhaps.

> Steering wheel, pedals, buttons everywhere. Probably well over 100 separate features.

Of all the complicated machines humans tend to use on a daily basis, a car is probably the most intuitive of them all, and most familiar. The basic, most important functions of all cars are practically the same. I don't think you can seriously say a steering wheel and pedals are confusing to anyone, especially with automatic transmissions. The basic controls in cars are pretty well optimized for intuitive, and safe human usage.

As for "buttons everywhere", sure, they can take a little getting used to per car type but it's not rocket science, or even car science. Pretty sure most people can get a rental car and have the whole thing figured out in 5 minutes. Apple could try something novel or "different" but remember, cars have multiple physical buttons and knobs because the driver needs consistent and unambiguous access to whatever "setting" they want to change without significant distraction from paying attention to the road. God help us if we had to swipe through menus and thumb a flat screen to change the AC temp.

There's room for improvements sure, perhaps voice control for some functions (though that might get annoying with kids and idiot friend passengers), and I'd be curious to see Apple's novel take on a UI that employs physical controls (not sure the last time they did anything like that). If they do want to improve something, maybe start with working on a way to present wtf the orange indicator lights that randomly popup on the dashboard mean.

Of course I don't think Apple has much interest at this point in regular human driven cars. In a vehicle meant exclusively to be driven by computer, I guess all bets are off. But we're a long way from that. Technical concerns aside, as we could do this with planes much more easily than with cars, it will be a long time before people would be comfortable with such a situation, and for good reason.

So I think radical changes to basic driving functions, and touchscreen controls for auxiliary ones is I think a no-go as long as humans are allowed to drive (though I have no doubt there are already examples/experiments in such things out there).

You don't need to make radical changes to make a significant improvement. Laptops of today are essentially the same as laptops of 25 years ago. And yet Apple has found plenty of ways to improve them in that time.

Also, what you're missing about car UI is all the other stuff a car does. Seats are adjustable. Sometimes they're heated and/or air conditioned. Ditto for mirrors, etc. Humans interact with many different parts of a car. All of those are game for improvement.

Did you look back at what people thought the "iPod phone" was going to look like before Apple unveiled the iPhone?

Let's just say that touch wasn't a major Apple product feature until they owned it.

What if Apple's release of a car is coincident with new UI tech like the iPhone's arrival with capacitive touchscreen tech?

"I dunno, just spitballing here but...a VCR? Heh...or a TV (remotes have lots of buttons!), or a computer perhaps."

Better to consider UI in relation to the possible failure modes. VCRs can be complex, but failure generally doesn't risk death. Cars are probably the most complex widely used device use of which carries significant risk.

Critical personal medical devices, like insulin pumps, have pretty simple interfaces. Or don't have any interface, as seen in pacemakers.

It's like a humble-brag form of advertising that you've got a self-driving car program.

Somehow I don't think Tim Cook is the kind of guy who does shit just so he can brag about it.

Why waste an opportune situation?

Reply on news.ycombinator.com