Hacker News Reader

Comment by twblalock | Hacker News Reader

The problem with self-driving cars is you can get most of the way to a working solution but that’s not good enough.

If the solution is flawed even a little bit, people will die. That will give self-driving cars a bad reputation and people will be unwilling to ride in them.

If you are going to geofence self-driving cars it is questionable whether they are better than trains and buses.

I think the best we can hope for in the next 10 years is better driver assistance systems, like GM Super Cruise or Tesla’s Autopilot (if they can make it stop crashing into things).

View on HN

I recently used a rental that had GMs "adaptive cruise" and it was mostly awesome even in dense traffic.

There were two main issues that had to do with following distance.

1: I had it set to 80mph and the other cars in the hov lanes on I-80 were going 75-80 and I was impressed how the car maintained a relatively safe following distance. This seems to irritate the other drivers and I would have tailgaters and eventually people would pass me on the right to occupy the space in front of me causing the car to slow down to get back to a safe following distance, causing the tailgaters to receive an autonomous brake check.

2: In slower traffic around 30mph (still set to 80) that same safe following distance allowed other drivers to cut me off, and several times the brakes engaged hard when the collision detector was activated. The first time that happened I was not prepared and I barely got my foot to the brake pedal in time. I decided after the second time that happened to turn it off as the time it took initialize my foot press the brake was too long, that when I was actively pressing the gas/brakes I could react much quicker to traffic.

I wonder if we can't take all these advanced sensors for self-driving and use them to help LEO fairly detect unsafe driving in order to increase mean driver safety.

Ironically the parent was probably technically going over the speed limit, though everyone goes over the speed limit there. Whoever builds automated traffic violation software would need to be very fairhanded in order to not upset the status quo.

There is an Israeli startup that does this (using machine vision and freely issued iPhones to rideshare drivers; the name escapes me at the moment, I apologize) to provide data to insurance companies for better pricing of auto insurance rates.

I'm not sure I want to live in a future where my Tesla is a mobile sensor platform for someone else's benefit besides myself and Tesla, although I can see the allure of negating the need for LEO patrols on road infrastructure once a critical mass of sensor platform vehicles is reached in an area. This would require a herculean amount of governance and transparency to be okay with me, and I would expect more sanity in road legislation (higher reasonable top speeds, for example) as well.

Is it nexar? https://www.getnexar.com

Yes, thanks.

I'm in Hawaii RN and the speed limit on the interstate here is 45 - 55! It's bizarre world compared to bay area interstate driving.

The speed limit is irrelevant. What's the community accepted reasonable speed? I would wager that it's not much different than the same highway conditions elsewhere.

people drive the speed limit here. its a much different driving culture than the mainland and it helps that speeding is aggressively ticketed.

There's also the fact that there are only 4 Interstate highways in Hawaii totaling 54.9 miles, while in California, there are 25 Interstates totaling 2,455.7 miles. So, we've got more space to get up to speed ;) and, of course, longer distances to travel...

I had the same issues in a Tesla. I adjusted the following distance down to dissuade people from jumping in front of me, but that felt even more unsafe. Seemed like there was no good solution.

The solution is to get people to stop being assholes on the road. Fine tailgaters. I care about the safety issues caused by them a lot more then I do about people going 6 over the limit.

Following any closer then 2 seconds of traveling distance on a clear, dry day, 3 seconds at night, or in the rain, 4 seconds on a rainy night should be considered reckless driving.

I feel like driving cars turns most people into idiots, myself included at times.

I don't have the citation, but I remember this quote:

"The easiest way to make people stupid is to limit their degrees of freedom"

Cars really only have 2 degrees of freedom: linear acceleration and radial acceleration. In traffic, these DoFs go down by a lot. You ability to de/accelerate linearly is very limited by the cars behind you and in front of you. Your turning also goes down to nearly a binary choice of 1 lane left/right.

Get people 4-wheeling and the stupid is very much still there, but just to get back to a main road, you have to use your noggin a fair bit more. Folks tend to think things through and weight choices more.

Are the horn and middle finger(s) not also degrees of freedom?

Actually, you make a good point.

The social dynamics are themselves a dimension. Social network graphs can behave as fractals. In a gymnasium filled with people, there is a significant 'fractallyness' to the social network. In traffic, the fractallyness is very limited to just about all the people in the road near you. The social graph, though more fluid in a traffic jam than a pep-rally, is still fairly small.

3Blue1Brown has a great video on the dimensionality of fractals [0] that is useful here. I'm just going to spitball, but I'd bet that the dimensionality of the social graph of a traffic jam is really close to 1.

So in the end, the DoFs of a traffic jam are probably very close to 3, but just a bit over. Say, 3.14-ish. ;)

[0] https://www.youtube.com/watch?v=gB9n2gHsHN4

The easiest way to solve this problem is to have serious standards from driving license tests. Have would be drivers perform reliably in a variety of conditions and in longer trips and mercilessly fail people who do not meet the high bar.

Rather than apply the brakes, barring the person jumping in front of you and braking, it could just stop applying the accelerator. Nice easy slowdown to create the gap without brake-jobbing somebody behind you.

Haven't driven a Tesla but I'm quite sure the safe braking distance is a lot shorter than it _feels_ like for a human and this is actually one of the tasks I'd rather leave to a computer than a human.

I'd assume the Tesla to have a lot better (quicker) reaction skills than you and you should probably just trust it (in this particular scenario) and relax, even though the distance might _feel_ too short.

This is true, the car can panic stop very quicky. The problem is I want to leave enough space to dodge pot holes that the car in front of me just drove over, since Autopilot can't really do anything about that (nor do I expect it to)

Driving a Tesla for almost a year now and when I had mine for 2,5 weeks it crashed into a car in front of me on AutoPilot in slow moving traffic. Stop and go. After 15 stops, it moved again, but this time didn't stop. I just sat there expecting it would brake. It didn't. There was no major damage, just about $1500 due to a small dent in the hood. AutoPilot is beta. Don't trust it to be quicker than you. It will fail in ways humans do not ever expect. Always pay attention and take control if it feels off.

I stand corrected.

It's weird because this really feels like one of the areas where technology could really excel in.

This is terrible. This is the exact reason why I would want to buy a Tesla. Just for the AutoPilot to drive the car in bumper-to-bumper freeway traffic. If it does not work reliably in this situation, then the AutoPilot is essentially worthless to me.

On my regular commute, the freeway interchanges that I must take, are too frequent, that I don't think the AutoPilot is worth the trouble to switch on. Especially when the freeway is empty, since I would have to take control of the car every few minutes.

The only scenario where I think AutoPilot is worth the effort to switch on, is if I must drive long distances on the same freeway, where you don't have to switch freeways. Like the drive from LA to SF. Or LA to Vegas. Or even just staying on the same freeway for 20 to 30 minutes.

I hope they can fix this deficiency. This AutoPilot technology is definitely beta, if they can't perfect simple stop-and-go freeway traffic.

The solution would be to detect attempts at cutting you off and automatically close the gap momentarily to discourage this behavior.

Ideally your autonomous car would also take part in the community of tailgaters and communicate forward that the current speed is unacceptable if this is the consensus of several tailgating vehicles, and gently reduce distance in a controlled way until the car in front gives in and speeds up or yields.

This is basically the opposite of the ideal solution.

The solution is to remove dangerous drivers from having control of 2-ton deathtraps, not contributing to the same behavior that is causing the problem.

Opportunistic, unsafe behavior is exactly the reason we need autonomous vehicles, rather than human-controlled vehicles.

What you're describing is a car that eschews defensive driving in favor of offensive driving. Speeding up to keep people from cutting in front of you only works until the other driver doesn't see (or doesn't care) and now you're in a crash. Or traffic suddenly slows as you're speeding up to close the gap, and now you can't respond quickly enough and now you're in a crash. Letting someone cut you off and slowing to return to a safe following distance is by far the safer option.

There's a reason it's called a safe following distance. Anything closer and you cannot stop in time. Even temporarily. Especially when autonomous cars in front of you may be suddenly braking as well.

Safer from the minimizing personal liability perspective. If cutting people off is a valid tactic because they bend over and take it then more people will do it more of the time and that makes everyone less safe.

We don't negotiate with terrorists for a reason.

Just give the user an up/down button and let them dial in the following distance themselves.

Or detect their preferred following distance for any given speed when the system isn't running and copy that.

Then you'll have AI engaging in unsafe driving habits because of their drivers and people will blame AI for every single accident rather than take responsibility.

Good way to get self-driving cars banned before they take off.

If the engineers allow the system to follow closer than is safe they should be personally liable for any accidents.

I used a Honda Accord with a similar system and even had a similar experience to scenario 1 on the interstate (though I was set at 69). After I realized that the following distance control was the reason the car was going slower, I was initially fine with it, but then someone cut me off which then forced me to driving slower resulting in me turning off the system.

I have a feeling these systems won't really be great until all cars are using them and considering I don't see that happening, I don't see them gaining a ton of traction, unfortunately.

Don't cruise in the passing lane, that's bad form.

I wasn't? They passed on my left side?

My mistake. Yeah, people easing over from the left lane going slower than you or just in front of you is annoying. The only real way to combat this is to make them commit to going faster than you by speeding up, or going slower than you by keeping them boxed in. All in all, sounds possible like not great cruise conditions, which is pretty much all the time in most metro areas now.

> There were two main issues that had to do with following distance.

The trouble seems to be that they're being too conservative. Following distance is what it is because of human reaction times. You have to process the input with your brain, decide what to do, move your foot, then the car starts to slow down. In principle a computer can do it faster and so doesn't need as much space.

Moreover, everybody would like following distance to be dictated by safety, but in practice it's dictated by traffic volume. If you have traffic averaging three car lengths between each car and then twice as many cars merge into the highway, either you end up with one car length between each car, or you end up with stopped traffic. There is no third option.

The problem right now is that they're still using the human driver as a backup, so using the safe-for-computers following distance doesn't mesh with that, but using the safe-for-humans distance is a violation of standard behavior. So they're stuck choosing between the unsafe thing everybody does, or the "safe" thing that then becomes unsafe anyway because of how everyone else reacts to it.

Why does it let you speed?

going 75-80 on I-80 isn't speeding. that's the pace of traffic even in the slow lanes. people do slow to like 70-75 if there is a highway patrol car nearby but since speed limits aren't really enforced 80 is the de-facto limit. and since cars are no longer the greatest cause of death in the US, they gotta let it go to cull the herd a bit and get that death rate up. (/s)

We have a Honda with similar system. It can be annoying in heavy traffic, but wow is it amazing on long drives between cities. There's no more fighting the cruise control when you and the vehicle in ahead are a little out of sync speed wise - either due to 75mph being a little different in each or due to performance differences on hills.

On the whole, I find a 3+ hour drive a lot less stressful with the adaptive cruise control, and I arrive feeling much less fatigued.

It seems to me that self-driving cars can't be just as safe as the status quo, but have to be far far better. It's an unreasonable need, but human nature.

I'll bet that to really make it work well, you also need to redesign roads to suit the new cars.

In any case, I'll believe it when I see widespread/universal adoption of self-driving in constrained environments (mining vehicles, warehouses, storage yards). Step two will be things like garbage trucks (slow, expensive, otherwise automated, phone home when it gets in trouble).

Humans have a tolerance for human-style accidents, but not robotic-style accidents. A dipshit that pulls in front of you to t-bone is a human-style accident that is acceptable. A tesla that sees a river and thinks it's a road is a robotic-style accident. It's going to take centuries for us to become acclimated to how machines mess things up.

Is it unreasonable?

One can learn to drive defensively, and be able to spot the said dipshit in advance, decreasing the likelihood of collision.

Yes, not in all cases, but there's a bell curve. As humans, we can read other humans.

Robots aren't inherently bad either. I love my Roomba precisely because if it fails, it fails predictably. I can and do make my apartment Roomba-friendly.

But the self-driving deep learning-driven cars don't have this feature. They fail unpredictabily, as in your example. We don't have an understanding of what really drives the decision-making there. And we keep discovering failure modes as they manifest.

One day, the car might think the river is a road. Another, it will not recognize a road block as such. But only if it's a Tuesday. Or something. We really don't know.

So I don't think we'll get acclimated to that. Unpredictable failure is a reasonable thing to be afraid of.

I'd argue it is unreasonable given that (afaik) no one is marketing a self-driving car that doesn't also require a driver be attentive at all times.

You can learn to drive a self-driving car defensively, in the same way you learn to drive a normal car defensively. Pay attention to your surroundings, blind spots, and other cars.

Maybe self-driving is a poor term to use.

>You can learn to drive a self-driving car defensively,

Then it is not a self driving car, maybe let's named them human supervised self running cars

Exactly, that was my point.

There are no self-driving cars on the market, or in development by major car manufacturers.

They all explicitly require human attention.

It's especially bad because us humans supervising the cars are going to be worried about preventing human-style accidents. Maybe you're worried about it following too close or not noticing the red light - so you don't notice the warning signs in time when it veers into the concrete barrier.

> (afaik) no one is marketing a self-driving car that doesn't also require a driver be attentive at all times.

Tesla's marketing videos for Autopilot have a caption along the lines of "driver is only present for legal reasons".

Wow, that's downright irresponsible.

>You can learn to drive a self-driving car defensively

My point was that you can't learn to drive a normal car defensively if there are self-driving cars on the road (in their current form).

When a self-driving car decides to do something stupid, you won't see it coming because the decision tree is literally beyond human comprehension.

See: http://www.bbc.com/future/story/20181204-why-we-should-worry...

And Uber's and Tesla's accidents have shown us that you can't rely on the human in the car to catch it before it does something stupid. After all, teaching someone how to drive and watching them attentively is difficult work.

Humans fail in unpredictable ways all the time. Look at the cases of people using vehicles as a weapon of terrorism. How are we supposed to predict that?

Yeah, we can go back and investigate the person's life and say "see, they had a history of browsing extremist websites, making bigoted comments, suicidal thoughts, etc." The problem is that this description fits a very large group of people, compared to the subgroup which actually carry out the attacks.

So while it is true that we may have difficulty (or it may even be impossible) auditing a self driving car after the fact, it should not be any more comforting to know that human failure can be explained after the fact, given that the human failure was not predicted beforehand.

The US has 5 million car crashes a year. Vehicular terrorism is not even visible on the chart.

You would think then a cooperative system between human and robot would be therefore be the best option. The human prevents the robot from driving into a river and the robot prevents the human from t-boning someone else. Except most people on HN seem to be completely against this type of cooperative system and would only support self driving tech when it gets to level 4 or even level 5 autonomy.

I'm not sure how such a system could possibly work; human attention doesn't work that way. If the robot drive 2 hours and 35 minutes just fine and one minute later decides to drive off the road, how is a human going to react to that? It's not really possible.

If a human has to pay attention all the time, then they might as well just be driving.

And one of the advantages of self driving cars will be freedom for the elderly and disabled who already have difficulty driving cars. This is the group, in my opinion, that will benefit the most from self driving cars.

>I'm not sure how such a system could possibly work; human attention doesn't work that way. If the robot drive 2 hours and 35 minutes just fine and one minute later decides to drive off the road, how is a human going to react to that? It's not really possible.

Right now the self driving tech doesn't do anything to communicate with the driver besides just a general warning. However that doesn't have to be the case. Maybe some type of confidence indicator should be added. A simple green, yellow, red warning system could help a driver know when a self driving car might not be as certain in its surroundings.

>If a human has to pay attention all the time, then they might as well just be driving.

I just don't get this mindset. There are different levels of "paying attention" that require different levels of mental energy. The discussion also isn't made with nearly any other driver assist technology. No one suggests that automatic transmissions or cruise control are worthless because the driver still has to pay attention.

>And one of the advantages of self driving cars will be freedom for the elderly and disabled who already have difficulty driving cars. This is the group, in my opinion, that will benefit the most from self driving cars.

I agree with that, but it doesn't mean that other drivers won't benefit as well.

> Right now the self driving tech doesn't do anything to communicate with the driver besides just a general warning.

I think this probably fails to understand how quickly it's making decisions, how quickly the situation could change, or how it could even judge it's own confidence. It could quite confidently drive you into a river; it's not likely to do that unless it's pretty sure it's a road.

As self-driving technology gets better, the robot is going to be much better and the failure situations much more rare.

> I just don't get this mindset. There are different levels of "paying attention" that require different levels of mental energy.

If one doesn't have to provide direction or speed, there is literally nothing to keep a human attentive enough to make a difference. You expect someone to passively look at a green/red/yellow indicator for hours at a time while not otherwise engaged with driving? And then when that indicator turns red you expect them to immediately be able to make a useful judgement? I think humans are more than likely to make that situation worse!

> The discussion also isn't made with nearly any other driver assist technology. No one suggests that automatic transmissions or cruise control are worthless because the driver still has to pay attention.

There is a huge difference. If you have cruise control or automatic transmission, the vehicle still doesn't go where it's supposed to without you paying attention. The driver still has to pay attention because they have to pay attention or they don't anywhere. With self driving cars, the driver is now a passenger and doesn't have to pay attention for the vehicle to work.

I haven't seen any evidence of a self driving car accident in which the car had no potential to warn the driver. All the examples I have seen are when a car misidentifies on object in its path as a non-threat. The problem is that breaking and/or swerving are very drastic actions. There is therefore a deservedly high barrier before a system is willing to take those actions. At the worst you could have two systems running in parallel and one is tuned to be much more aggressive in making those decisions and that information is communicated to the driver.

No self driving car that doesn't have at least level 4 autonomy allows the driver to not pay attention like you are suggesting. The benefit of a cooperative tech is that it takes mental load off the driver. Like I said previously, there are different levels of mental energy that a person can spend on a task. It isn't like the choices are either 100% or 0% of your attention.

Such systems work fine. The human does have to be constantly driving, but the computer intervenes when it detects the human doing something dangerous. Look at automated GCAS in aircraft.

That's not self-driving though. There are plenty of driver-assist technologies (automatic breaking, for example) but that's not the subject at hand.

Full self driving in the general case will probably not exist in our lifetimes. It's like chasing an imaginary pot of gold at the end of a rainbow. So the industry should aim for goals that are actually technically feasible and will deliver tangible safety improvements.

I disagree. The progress of camera and laser technology, in addition to AI, is improving at an incredible rate. The market for self-driving vehicles is huge -- any company that figures it out will make billions (maybe trillions).

Computers can have a perfect 360 degree view and track and categorize every object around them. Humans will be both physically and intellectually outmatched eventually. And humans just aren't that good at driving on average.

You're welcome to disagree, but so far there's no hard evidence to support your viewpoint. Just a lot of hype.

Past rates of progress are not indicative of the future. The easy problems have already been solved. Now progress is already slowing down.

In particular the notion that computer vision can reliably track and categorize every object is just laughable. The state of computer vision research is nowhere close to that capability. Errors are frequent, especially under adverse conditions.

In November 2017, Waymo announced that it had begun testing driverless cars without a safety driver in the driver position. In October 2018, Waymo announced that its test vehicles had traveled in automated mode for over 10,000,000 miles (16,000,000 km), increasing by about 1,000,000 miles (1,600,000 kilometres) per month.

In Arizona, Waymo has fully autonomous taxi service you can use right now. https://waymo.com/apply/

Is that a lot of hype? Maybe.

You can get 99% of the way there with extremely detailed maps. Where the road is, where the lanes are, what the speed limit is, where to find stoplights and rail signals, where to be extra-cautious, how to navigate through a construction zone, all can be pre-programmed.

A Waymo is basically just following a virtual track on an extremely detailed LIDAR map of the area, obsessively watching for pedestrians and following the rules of the road the best it can. It will never think a river is a road.

Not to say that there aren't a million little things to be concerned about with this approach, or that there aren't major things to overcome (like heavy snow on the ground making your LIDAR map useless). But I think we'll get there, and in many places soon. There's nothing that says this tech has to exist everywhere out the gate.

Tesla's approach thinks white trailers are open sky. The river comparison is not unreasonable at all.

Almost all of the startups in the space are following the Tesla approach of just throwing machine learning at the problem which means almost all work in the space suffers from novel input producing unpredictable results. This is one of the things that has killed trained and evolved systems in the past and the fact that few of the companies in this space are even trying to manage it (either by building interpretable models or by using models and using ML to do parameter fitting) is a good indicator that the whole business is either a fraud or is built on the premise that with enough data or, for the un-cautious and unaware, enough simulation, the problem solves itself.

I think what we are actually seeing is that Waymo (where, in ten years, they might have a solution) and Tesla (which is mostly worst-in-class but as a company is the master of hype) drove hype around self-driving. Then Uber and Lyft latched on (because they needed a story to paper over their terrible economics) and pushed it even higher. The Otto thing and Cruise's acquisition made VCs pay attention.

So these companies are acquisition bait for the assumed-clueless big auto companies. They will not deliver self-driving cars. At best they will deliver next-generation enhancements to emergency braking, etc.

> Tesla's approach thinks white trailers are open sky. The river comparison is not unreasonable at all.

A Tesla is not a self-driving car. Autopilot is an assistive technology.

They can change the construction zones on any given day - adjust the lane changes, move the barriers a foot over, or add a new fence where there wasn't one yesterday.

> Humans have a tolerance for human-style accidents, but not robotic-style accidents.

I think it's more because there's someone to yell at - someone to blame.

Think of it like a taxi driver with a passenger; people ride in these all the time, or uber, or lyft - and have no problems doing so. Even though they aren't in control, and they have no idea about the driver's driving capability. They assume the driver is an average to better driver, when that may not be the case at all. So what happens when an accident occurs?

Well - if the passenger is still alive, and believes it to be their driver's fault, they can yell at their driver. If they believe it's the other driver's fault, they can yell at them...and so forth.

But you put a robot in place of any driver - then who can you yell at when an accident occurs? Nobody. You can yell all you want, you can even blame the corporation that made the car/robot - but you are still yelling at a machine. It won't change anything. It won't feel bad. It probably won't even make you feel better, because it won't restore your agency in the situation.

That's my guess anyhow - pulled straight from my nether regions, with nothing to back it up. I wonder if there have been any studies on this? Are there any data on what happens in other "self-driving" incidents; like perhaps completely automated trains with no humans other than passengers on-board? Do such things even exist?

Or do companies already know this - and put a human "in charge" to take the blame and heat for when an accident happens (maybe they also provide an e-stop button for the human to press - whether it's actually connected to anything or not - but it's logged as to whether it was pressed, just to help "assign blame" later)...?

While I have no proof or anything to back this assertion or theory up, it seems like it would match human nature. When things go wrong, humans want to blame somebody - some actual, sentient being. I'm sure there's a psychological term for this need and the action; I don't know what that term is, but I'm almost certain it's been studied.

Essentially one of two things would have to be done to overcome this: Either the machine would have to become perfect at driving and avoiding accidents (even to the point of avoiding things of natural random consequence, such as rocks falling on the road, or something of that nature) - or it will have to become more sentient, and ideally emotive; that is a GAI with human traits.

The former is likely impossible - randomness will always get ya; even trains have what should be "avoidable" accidents.

The latter - well, if it's achievable; that is, if we can make such a GAI, etc - it will likely also suffer the same problems humans do, most notably that we are failable and we make mistakes.

...but at least then we can blame them on something, right?

>When things go wrong, humans want to blame somebody - some actual, sentient being.

Exactly! When (not if) a fully autonomous self-driving car causes a terrible accident resulting in death, how are the loved ones going to feel when they've only got a company to blame? In such an emotional situation, the response needs to be better than Uber or whoever saying "we're very sorry our car did that, but hey, we're trying our best, and these things are going to happen". Are people really going to accept that? That's simply not going to fly.

We all accept risk in our lives, in most cases the qualitative characteristics of the risk are the more important one. I don't really have a good enough answer to your point, but I find it morally unsettling that many people would agree with this.

One example I can think of is Russian Roulette. I would be quite a different game if each player had to shoot the neighbour instead of himself. It would be literally the same expected outcome, but an entirely different game.

Each human I encounter everyday has a small chance of murdering me (let say one in a million) would I like a robot wandering around with one in a billion chance of murdering others?

We accept that we cannot endlessly manipulate and control other humans, that we cannot forcibly "fix" arseholes, we have no such limitations for robots.

We already have an example of strong automation-driven transports: aeroplanes. Honestly I see no reason to have lower standards for self-driving cars.

> self-driving cars can't be just as safe as the status quo, but have to be far far better. It's an unreasonable need, but human nature

Human nature doesn't end up meaning much as soon as insurers start seeing the statistical difference.

As the difference gets more pronounced with the lane assist and auto stopping features, rates on any new cars without these features will go up relatively.

Insurers don't care all that much if you rear end someone or get in some other "normal" accident. That happens enough that they have figured out how to cover the cost at scale. They don't take that much of a hit anyway because of your increased premium over time. They'd obviously prefer you do it less but so long as their slice of the population doesn't to it any more than their competitor's slice it's not a big deal because they come out even. That's why they offer $10 and $20 discounts for safety and security features. It's not a big deal to the insurer since they know exactly what the "value" of each of those features is at scale but of course they want the cheaper to ensure customers so they'll toss them a small perk, a fraction of the money that feature statistically saves them.

What insurers are really scared of is when your 16yo kid and four of his/her friends drunkenly go off a cliff on their way to prom. That's hard to assess the risk of and it costs them big bucks. These kinds of oddball high dollar edge case accidents are exactly the kind that self driving cars seem to be getting into. A Tesla killing someone by driving into a barrier or semi truck sends chills down the insurer's spine because it's only a little bit of bad luck (from the insurer's point of view) away from a not so dead customer that's maxes out their collision and medical coverage.

If I were an insurer I would not be offering any automation discounts until we have more data.

> until we have more data.

Well, yes, obviously. The commenter's point is that there will be a tipping point where insurance agencies look at those "crazy prom death" accident stats and see they happen 2x or 10x more often with human drivers. At that point, the cost of insuring a human driver is going to be 2x or 10x the cost of insuring an AI driver.

> It seems to me that self-driving cars can't be just as safe as the status quo, but have to be far far better. It's an unreasonable need, but human nature.

They need to be safer than me, not the average driver. I suspect the lower 50% of drivers cause > 50% of the collisions.

> They need to be safer than me, not the average driver

But I'd they are safer than the average driver, you gain from their widespread adoption even if they aren't safer than you. Because you don't have the choice of replacing other drivers with you...

Defensive driving allows me to counter most bad drivers.

If we're talking about removing dangerous drivers from the road, I'm all for it. It should be next to impossible to get a license for most people, if what we really care about is safety.

> It seems to me that self-driving cars can't be just as safe as the status quo, but have to be far far better.

They probably, for acceptance, have to be consistently as safe in like circumstances, they can't just be as safe on average. Which, if there are circumstances in which they are much safer, also means they end up needing to be much safer on balance, but it's not really the on-balance comparison that drives that.

> It's an unreasonable need, but human nature.

Not at all. Computers are supposed to be orders of magnitudes better than humans at certain tasks. If you're going to put computers in charge of our lives on the road at high speeds, why wouldn't we expect them to also be orders of magnitudes better at driving?

Also, if I'm a lot better than the "average driver" (where I assume such statistics also include people texting and driving or driving drunk or having not slept for 40h, etc), then I certainly wouldn't be satisfied with an automatic system that is only a little better than the average driver.

And I didn't even get into the whole thing about car companies, including (especially?) Tesla, which compare highway Autopilot accidents to all road conditions accidents, to show how their Autopilot is "better," or the thing about such systems making errors that humans would never make, just like language translation systems can make totally different translation errors compared to humans, even if they "score the same."

To keep things short, yes, I think it's completely reasonable that self-driving systems should be far FAR better than your "average human driver." I don't even know why this is controversial.

Also, I agree with the Ford CEO that all carmakers, even Tesla, overestimated self-driving capability over making an electric vehicle. The ability to make a great EV will matter far more for these companies' survivability in the next 10 years than making a good self-driving car. But it's almost like most of them focused more and invested more in the self-driving capability than switching to EVs. Big BIG mistake!

Even Tesla, the only major pure EV company, mistakenly almost "bet the farm" on self-driving to the detriment of making a high-value EV. How? Well, by putting very expensive "full self-driving hardware" (that turns out is not actually full self-driving) into every Model 3 car out there. Terrible decision. They would've done a lot better if every Model 3 was $5,000-$10,000 cheaper by default without all of that gimmick.

> Computers are supposed to be orders of magnitudes better than humans at certain tasks.

Who claimed that driving is among those "certain tasks"? No, in the near future computers will at best be marginally better than human drivers. Computers are better at decision latency. But that merely compensates for some of their current shortcomings.

And that will be sufficient to save thousands of lives every year. Do you not like preventing deaths?

Delaying "marginally better" for "orders of magnitude better" means delaying the deployment of life-saving measures.

>It seems to me that self-driving cars can't be just as safe as the status quo, but have to be far far better. It's an unreasonable need, but human nature.

What would be the point of self-driving cars which were no safer than the status quo? That's just removing freedom from drivers and adding more complexity to infrastructure for no added benefit.

It's also exactly the narrative that proponents of self-driving cars have been driving (pun intended), self-driving cars would eliminate all, if not nearly all, accidents and fatalities.

No added benefits? People don't have to own cars commuting drivers regain hours of their life.

Why are you assuming no one would own self-driving cars? It's likely they'll simply be sold the way any other kind of car is nowadays, and be far more expensive than typical cars.

Also, commuters will still have the same commutes whether they or their car is doing the driving. The deployment of self-driving cars is not going to magically transform a multi-hour commute into ten minutes.

Why would I want to own a "far more expensive" self driving car? I would love to get rid of my car and rely on fully autonomous Uber/Lyft. Even factoring in car rental for travel or moving or whatever I think it would be far cheaper that way. My car is a never ending money pit with insurance, gas, maintenance, registration, tires, it never stops.

>Why would I want to own a "far more expensive" self driving car?

It doesn't matter what you want, it matters what car companies do. They're not going to spend millions of dollars on R&D and marketing hype for self-driving cars only to make far less money on them than they do now on conventional cars.

>My car is a never ending money pit with insurance, gas, maintenance, registration, tires, it never stops.

Yes. That's the entire point, cars are micro-economies supporting multiple businesses and profit centers. And self driving cars will be the same, for the same reasons.

I'm not sure exactly what you're saying. Are you talking about some kind of monopoly or trust that prevents Uber/Lyft et al from getting access to self driving cars? That seems unlikely to me. As far as the various industries centered around car ownership I don't see why they'd go anywhere. Self driving cars will still need maintenance, oil, tires, etc. The only thing changing here is who owns the car. I don't want that to be me. As soon as it is economical for me to pay for transportation as a service and get rid of my money pit that is parked outside my apartment 24/7 I will do so.

I think you're reversing supply and demand...

You're going to pay for all those things even if it's indirectly through someone else who owns the vehicle. Most vehicle costs are related to mileage, not time--especially outside of areas where rust is a major factor.

I'm sure there will be a difference at the margins in cities where many people don't need a car on a daily basis. I assume that's already the case with Uber/Lyft today. And self-driving when it eventually arrives in those areas presumably will decrease costs some. But for someone who lives outside of dense city centers it seems likely they'll continue to want to own a car that is a model they like, equipped to their specifications, and storing all the various things they keep in a car.

This completely overlooks that in 90% and over of the cases a bus is all you would need

You're right I didn't mention it, but that fact is actually crucial to my point. I'm lucky enough to live in a place where walking, taking the bus, riding a bike (the one I own or a bike share), or taking a Lyft get me almost everything I want at a very reasonable cost. And as new transportation services emerge I need my car less and less. So I'm hoping that in the near future there is some service that can take over the few things I do need a car for. Something like Car2Go or ZipCar is actually almost there.

The main attraction is that automation will make car ownership non-sensible. Even right now, taking a cab every single time is economically at break-even or better for many urban dwellers. If you take out the driver, then cabs can be much cheaper.

Second, what you call "freedom", is called a "chore" by a good chunk of the population. I can't wait to not touch another steering wheel ever.

To each his own, but I don't want to have to get permission from Google, Tesla or anyone else before driving somewhere, then be escorted under surveillance like a prisoner in the back of a cop car with every possible metric being mined and correlated for the benefit of corporations, insurance companies and the state, leaving my safety to the whim of algorithms and systems likely built to the cheapest standard possible.

I'll gladly deal with the "chore" of being able to turn my own key and drive my own car as an alternative.

I wonder at what point they'll just solve the safety issue by removing the windshield and adding shit loads of impact absorption and a few steel plates that isn't possible on vehicles where the driver needs to be able to see outside.

The limiting factor right now is mostly that nobody wants to wear 4/5/6pt harnesses or have a properly fitting bucket seat (these things are a big pain in the butt for a daily driver) so we have to keep shoving explosively deployed cushions into places so humans don't bounce off of harder things.

The fact that we can't replace the windshield with something else will likely never be a practical limitation for the foreseeable future.

Yeah, as someone from a rural background, it always strikes me as far-fetched that people are going to give up ownership of vehicles because of driverless tech. People routinely spend 10x what they need to on a car, living well outside their means for essentially status (buying a new BMW instead of a used Honda for example). Moreover, there is more to owning your own car than privacy. It’s really damned convenient to have your car loaded up with your stuff. Your snow gear, your bike gear, your tools, your emergency electronic tools and first aid. I like to keep a folding chair in my car, it’s surprisingly convenient. I don’t have a ton of experience living in cold climates but I did spur-of-the-moment cross country road trip in February a few years ago. We left from the Bay Area headed to Indiana, and when we stopped outside of Chicago for gas and I realized how naive our plans were. I had to get back in the car while gas was pumping it was so cold. If we had car problems or an accident dying of exposure could have been a real thing. We didn’t have cold-weather gear with us, I was wearing jeans and a hoodie, and I really felt like an idiot for it. Such fears can be easily solved by having blankets and food and chemical heat sources in the car, which I’m pretty sure is normal for people in those climates. You can’t really have gear with you like that if the driverless car goes and gets another fare once you reach your destination.

What they're talking about I think is the usual statement that self driving cars only need to be slightly better than the status quo (in deaths per mile driven) to make sense because then anyone moving from driving to using a self driving car decreases the number of deaths. From a raw statistics POV this makes sense, if a self driving car is even just a little bit better than the average driver on average anyone switching to an autocar will decrease the number of injuries.

What they're saying (and I agree) is that human nature means people won't be willing to get into a car that's 'just a little better than average.' There's a couple reasons I think that will be true: 1) people generally think they're better than they are at driving 2) moving from people to autopilot moves the responsibility from generic 'bad drivers' causing accidents to a single system. Going from a diffuse responsibility to a more concentrated liability on the part of the manufacturer will probably mean they'll be blamed much more for the failures of their cars.

I think insurance companies can affect this behavior. If self driving cars are safer, then the insurance companies could lower rates for those cars, creating some economic incentive to override the irrational human brain.

> What would be the point of self-driving cars which were no safer than the status quo? That's just removing freedom from drivers and adding more complexity to infrastructure for no added benefit.

What?? Safety is a great side dish, not the main course. The point is to free up hundreds of millions of man-hours spent daily focusing on roads.

Then simply a better public transit system would be more efficient and effective.

They're a lot better already safety-wise (and likely in other ways too) than drunk drivers I'd assume.

It's in all likelihood already safer than that status quo by an order of magnitude. Not that they are that good but the status quo is really that bad.

But it's the devil we know, It's fine that we average 3,287 fatalities a day[0].

We also have more empty homes than homeless, more food than starving people, crippling homelessness and 1:4 malnourished children in the US alone. I expect nothing from humanity yet I am still disappointed.

[0]https://www.asirt.org/safe-travel/road-safety-facts/

Interesting.. I would actually think garbage trucks would be one of the last vehicles that are automated. They have to do some insane navigating in small spaces. It seems like the absolute hardest thing to automate that I can think of.

Do they? Most I see just cruise down pretty open residential streets grabbing convenient cans. Like most things it'll probably be a mix, most service would be covered by automated trucks with the more troublesome routes still driven by people until the kinks get worked out.

I would imagine actually picking up the cans is the hardest part. they often look different from each other, people place them haphazardly, and you might even have to distinguish trash from recycling.

"Survey: Self-Driving Cars Need to Reduce Traffic Deaths by at Least 75% to Stay on the Roads"

https://www.claimsjournal.com/news/international/2018/06/01/...

I am perfectly happy to accept autonomous cars to kill even a little more people than human drivers do in exchange for the utility that they offer. You can use your time in the car to do other things. That's actual life-hours saved.

And the promise is that they're only going to get more reliable with time.

Conversely, keeping back self-driving cars just because they upset your sensibilities who specifically should die, i.e. random group A instead of random group B even though A is larger, means you're effectively advocating that more people should die. That's grossly negligent.

> If you are going to geofence self-driving cars it's questionable whether they are better than trains and buses

They run on demand 24/7, they run point to point, you never have to stand, you don't have to share the space with people who behave antisocially, you have a range of choice of comfort options, there are unlimited competitors for the same route driving prices down and service standards up.

So, quite a few game changers for me at least.

If you live/work outside the geofence, this obviously doesn't apply, but if you do then you get the advantages 80% of the time, and for day to day purposes it absolutely is fundamentally better than trains and buses.

And they pollute much more and take more space.

I'm first in line to agree with you, I think cars in cities are awful because of these two problems, and honestly I support all sensible efforts to reduce the number of (current generation) cars on city streets.

I'm also a daily user of public transport in a major world city, however, and whilst buses and trains work they also have the fundamental problems highlighted above that are in my view insurmountable, and which (self driving) cars solve. So I'm in a dilemma over whether I think they'd be an improvement, but I ultimately believe they would be.

But yes, absolutely pollution and space saving would have to be addressed. I'm optimistic they could be:

Pollution: hopefully solvable with electric. Yes, of course, there are efficiency gains that trains/buses would perhaps always have (having to move many small things rather than one big thing carrying the same number of people) but I can see electric reducing the absolute level of pollution from cars to a point where it basically doesn't matter. Maybe overly optimistic, but I hope that'll happen.

Space: Driverless cars, particularly electric driverless cars, can be much smaller and can drive much closer together. Different cities have different challenges, but I don't see why we couldn't get to a situation where roads might effectively have capacity for 2x vehicles across and 2-3x vehicles along if the vehicles can be coordinated to drive closer together and are physically smaller (remember that on-demand unlocks certain design constraints - you don't have to provision for peak - if the average number of passengers is 1.5, then most cars on the road could be ultra small 2 seaters). Again lots of conjecture and optimism here, and I am probably assuming that 'manual' cars would eventually disappear or be banned at least from cities too, but nonetheless I think there is hope here.

Public transport will always be able to pack more people into the same space.

There is a minimum amount of space occupied by the vehicle's own matter. The ratio of that material to the occupant is a scaling law: the larger the vehicle, the higher the volume-to-surface, the denser the passenger-to-vehicle ratio can be.

A standard R160 subway car carries up to about 250 people in an area that would be about the same as 5 Corollas positioned nose to tail, a starting ratio of about 10:1. If cars are made twice as dense, it's still 5:1.

This is ordinary batching. The only thing that can carry more people more densely than a bus or train is the footpath.

Yes you're absolutely correct - I guess my main question is could the space savings, like the pollution savings, be 'good enough' for it not to matter in practice, assuming that driverless cars become the 80% usecase for road traffic and today's 'massive 5 seater car for an average of 2 people' model goes away.

It's also important to point out that of course, buses and trains shouldn't go away - they could indeed also have a bunch of problems solved by being self driving too, most notably the 24/7 running thing - they'd just be another option in the mix.

> Public transport will always be able to pack more people into the same space.

While that's true, that's also just an argument for something we already understand. Are there possible arguments to be made about the number of busses and routes because it's not taking people point to point. What about the lul-time of day, where on-demand model is more efficient than driving empty busses around. They reduce schedules obviously. But now you're punishing passengers that start a shift at noon. There are more things to solve than just "the least amount of pollution and space."

Totally unrelated but similar argument. The FAA introduced a new system and process for creating standard flight paths a few years ago. Flights around the nation started doing new things. TONS of people are now inundated by jet noise. They're now discussing just going back in certain areas to the old routes. They ONLY optimized for efficiency of route and didn't consider anything else.

> There are more things to solve than just "the least amount of pollution and space."

The argument I was responding to was that self-driving cars would be sufficiently more efficient users of space that it would obviate the need for public transport. This was demonstrably false.

> But now you're punishing passengers that start a shift at noon.

Sure, but (a) that's a financial question, not a capacity question and (b) rush hour is when both finance and capacity collide.

Public transport is a more efficient option per passenger-kilometre and per square metre of road or cubic metre of tunnel than self-driving cars. It is always going to be. Self-driving cars will have other virtues and will reshape transportation, but ascribing magical powers to them does nobody any good, especially if it leads to defunding of subways and buses.

>A standard R160 subway car carries up to about 250 people in an area that would be about the same as 5 Corollas positioned nose to tail, a starting ratio of about 10:1. If cars are made twice as dense, it's still 5:1.

You're not accounting for the space between trains. They will only take up 5-10% of the available rail space because you need gaps between the trains.

You need gaps between cars, too. Two seconds of following distance in good weather conditions, four seconds of following distance in bad weather conditions.

At highway speeds, 4 seconds of following distance is 120 meters.

> You're not accounting for the space between trains.

NYC's least efficient system is block train control, requiring 300M separation between trains.

In 300M you can fit about 66 Corollas, again packed bumper to bumper. That's 66 * 5 people = 330 people, or approximately 1.5 subway cars.

Subway trains are 6 or 8 subway cars long, depending on track and time of day.

With CTBC some platforms can run at 60 trains per hour (30 local, 30 express). That's 8 subway cars every 2 minutes, or approximately 1,000 people per minute.

To move 1,000 people at all with the Corollas you will need 250 * 4.55M of them, or approximately 1.1 kilometres of cars, bumper to bumper.

>With CTBC some platforms can run at 60 trains per hour (30 local, 30 express). That's 8 subway cars every 2 minutes, or approximately 1,000 people per minute.

Unless your trains are going at 10 kph you will have more of a gap than 300 meters if you see a train every 2 minutes.

Or the trains can move much faster and much closer because of improved control.

You should recognise this argument: it's the same logic given for why self-driving will solve everything.

But you're always limited by loading time. If you take X seconds at each stop then you will end up with X * speed distance between each car. Even at 30s loading and 50 kph you're looking at 400 meters between each train.

Why would driverless cars be smaller? They still need to be able to carry the same number of passengers; the only difference there is that one of those passengers is no longer also a driver.

With autonomous cars, robotaxis become feasible. If you're buying a car for yourself, you're going to want a certain amount of space for passengers or cargo in case you need to do something like drive your family to the airport. You'll also want long range in case you decide to go on a road trip. Even if 90% of the time you'll be commuting alone 20 minutes to work, you'll want this extra range and capacity for the times you need it.

If you're deploying a fleet of robotaxis, most of your fleet can be vehicles that only carry one or two passengers at a time on short trips. If a family wants a ride to the airport, then they'll specify that they need a larger vehicle for more passengers and luggage, etc. So you can buy a fleet of mostly smaller cars with a shorter range, and compliment them with a few minivans for long range trips. Overall your fleet will be quite efficient.

One thing that is often not accounted for is car seats. Do you have to bring your own?

Yep, and given that some big percentage of car seats are not properly installed, do you then have to take the requisite minutes to install and uninstall each seat properly? There are a lot of challenging edge cases.

Packing more self-driving cars onto roads is an appealing notion, but I don't think it holds up to scrutiny. Even if a computer detects hazards instantly and continuously negotiates with and anticipates the movement of other vehicles, the physics of moving cars remain the same. Safe stopping distances won't be dramatically different. Road surfaces, stray animals, mechanical failures, and any number of other hazards don't care who or what is driving a car.

Say you have 3 lanes of traffic. Each lane is occupied by 100% driverless cars.

Let's say, there is an object completely blocking the middle lane, and partially blocking the two outer lanes. What will the driverless cars decide to do?

My guess, is they wait there until the road is completely unblocked, there's no way for the cars to understand the proper course of action. It will be great fun to bring absolute gridlock with an empty cardboard box in downtown NYC once this all kicks off. Naturally, by then, card board will be illegal there.

> the vehicles can be coordinated to drive closer together

How do you make this work? There's no standard in self-driving systems, no standard in sensors, and no standard for vehicle to vehicle communication.

Prior to low cost ride sharing, a good bus system that people will use as an alternative to cars pollutes more because it has to run a lot mostly empty buses to satisfy long tail demand.

To the extend low cost ride sharing is a short term unicorn phenomena, that will again become true, although self-contained electric buses might change that a bit. Maybe someday self-driving cars will change this, but that's not in the predictable, foreseeable future.

Yeah this is the really interesting part - like where does the 'supply provisioned precisely to match demand' thing for self driving cars meet the 'buses/trains are more efficient assuming some % capacity filled', how does that relate to demand changing with hours of the day, where people live, etc.

It's just a fascinating optimisation problem. I think the answer is far from clear, and certainly couldn't be reliably approximated without lots of real world data.

Not necessarily, it's not that simple.

I've been on double-length public buses where I was the only passenger for most of the journey. And many times where there were only 1 or 2 other people. This is necessary e.g. at night where a bus only runs every 30 minutes, there has to be a bus, but there just aren't that many people.

Obviously in this case individual cars would pollute much less, and even take up less space.

Any realistic solution for a city is going to combine self-driving cars for the long tail of origins/destinations and times, with public transit for the busiest routes at the busiest times, including links between the two.

I'd love to be able to take a self-driving car from a house in Queens directly to the subway, the subway to Brooklyn, and then another self-driving car to my final destination that's 20 blocks away from the subway.

On average public transit produces 2% of pollution but transport about 28% of riders. (I don't remember where exactly I got the numbers but I've seen them in a few places).

The buses need to run at night so the people can rely on them during the day -- it's just the price of public transit, the public uses transit way more if buses run on 10-minute interval during off-peak rather than 30-minut interval. (another study I can't find right now).

I think as programmers we tend to try and optimize problem for least pollution, but the problem here lies in the domains of sociology and urban planning; we might not be the most qualified people to solve it.

For example if buses pollute 3% because of the extra buses on off-peak routes but now more people use them, this would have net effect of reducing pollution.

Was going to downvote but this probably merits a reply.

All your points are correct, but you're not responding to the point that SDCs can be superior to buses for those low-capacity times. If it were one or the other, then yes, buses could be less pollution on average, but that doesn't address the case the parent was talking about.

Yeah, this is something that really bugs me about public transport.

It's simply not acceptable to have to stand on a bus. Ever. It should be treated as some sort of breakdown/failure condition. If it happens on more than a few % of journeys, the public transport authority should be putting on more buses.

Otherwise people are just going to take their car because they actually have control over their own comfort.

With trains it's a little harder due to physical constraints on the network, but with buses it's inexcusable to not just have more capacity, particularly at peak time (London buses get silly in the morning peak, for example).

edit: Replies here are completely missing the point. A car has comfortable seats whenever you want them. It's honestly a comical joke to have to explain "why I don't want to stand up". The obvious answer is because I don't have to in my car.

Public transport can be, and should be as comfortable as that. It can't be as private, sure, but there's no need for it to feel like cattle class.

I'd gladly pay twice as much for it to be better. It easily could be; in London for example if everyone gave up their cars and put the money towards public transport we could double or triple the number of buses.

Why is it not acceptable to stand on a bus? It increases capacity hugely and if you're only going a few stops it makes perfect sense.

Most buses are rated for "X Standing, Y Seated" passengers.

Also, wishing for a blanket increase in total number of busses by some huge percentage to avoid anyone standing probably sounds like a real nice plan until you have to pay for it... Capacity management isn't as easy as throwing drivers and busses at the problem.

Because people will take a taxi or self driving car instead.

You're making a comparison here that's based on having no other choice.

There is obviously a difference between a 5 min journey down the road and a 30+ minute commute.

Will they? When I drive I have to worry about parking and, you know, driving. If I take a taxi or ride share I'm paying on the order of 6x what the bus costs.

I think you're overstating the annoyance of standing on a bus.

People don't mind in say, SF Chinatown. They pack those buses like sardine cans.

> It's simply not acceptable to have to stand on a bus.

you need bus fares to cost less than car travel or people'd just drive everywhere and car ownership/congestion comes with it's own set of issues.

it's annoying but having enough capacity to handle peak hours would also see lot more buses parked off hours, lot more technicians doing maintenance.

given the seated:standing capacity on current models, you'd end up with twice to thrice the buses to seat everyone.

now this wouldn't immediately mean to have triple the fares, but it'd be a significant factor and limit/reduce total bus usage, and there's no indication the consequences for society would actually be net positive compared to standing.

Really depends on the distance and the rides harshness, that's why we have buses and carriages filled with people who didn't take a car anyway.

Definitely not ideal but we can't optimize just for one thing. Putting more busses on demand has really hard scaling problems because of these busses don't appear and disappear like a virtual machine, it needs to be stored somewhere and delivered to the location where that bus is needed. Then you have load-unload problems. This is also why cars that are just like the cars of today but electrical and automatical won't scale too.

You can't really divorce the design of the inhabited areas and the transport systems. This is also probably the reason why the Americans seem so hostile towards public transport. In Europe and even in Asia public transport works fine and the most annoying thing about cars is not that you have to drive them but you have to put them somewhere when you reach your destination.

I almost never sit on a bus or a tram even if there are empty seats.

because poop and gum seats right?

Nope. In my location trams are new and clean. I just prefer to lean against the windows and people who need it more can use it. I don't travel more than 20mins with public transport, though.

> If it happens on more than a few % of journeys, the public transport authority should be putting on more buses.

That's unrealistic, because there are such huge peaks in traffic. You'd need a large number of extra bus drivers only from 7:30 to 8:30 and from 17:00 to 18:00 each day.

Who would want such a job?

Whether it is realistic or not is beside the point.

A bus cannot beat or even match a fleet of self driving cars if the suggestion is that a bus has to be a cramped standing area.

Given a service much cheaper than Uber is today, the only people using the buses would be the poor and environmentally focused.

I take the bus and I'm neither poor, nor a particularly good environmentalist. They're convenient and cheap and I sorta prefer it to putting my life in the hands of a sleep deprived wage slave. Uber rides feel _consistently unsafe_ to the point I don't even consider it an option anymore.

IME most of the resistance people have to taking the bus is classism. They don't want to share space with "those people".

Presumably then, in your area, buses are superior.

This is exactly what I am talking about. They are not everywhere, and they should be.

I prefer buses too when they actually work.

I tend to take public transport a lot more off peak. At peak time I vastly prefer cycling or driving depending on the weather because I find being crammed next to other people disgusting (a failure mode unique to public transport).

It’s not about class at all. It’s that people don’t want to share space every day for an hour each way with loud people, smelly people, crazy people, violent people, etc. And if they do, they sure as hell don’t want to do so standing and packed up against them. Not all cities and routes have these problems, but many do.

Nobody objects to having to share space with a broke working class single mother taking the bus to go clean hotels downtown.

Aside from that, the weather can be a pain in the ass (rain, snow, wind, scorching heat), transferring buses/trains is a hassle, you cannot travel on your own scheduling terms, you can’t swing by the grocery store or run some errand on the way home, travel time is generally much longer unless you happen to live/work right beside an express stop that does not require a transfer, and the list goes on.

I grew up on a council estate in a post industrial Northern town.

It's not classism.

I don't want to be rammed into a bus. I like buses that are at capacity, it's relaxing.

Or... cities see self-driving cars, draw the correct conclusion that they can put some smaller, cheaper and more efficient self-driving buses and blanket ban cars in places where there are too much of them. You either walk or take the bus.

I imagine Uber or Lyft drivers would be good candidates.

Would they be willing to get a bus driving license ?

"such a job" would not be made available as bus drivers are typically unionized employees and that is one reason part-time bus driver positions do not exist

It's solvable, especially since it's the government that is responsible for public transportation. For example, you can arrange another job for the non-needed hours. For example - a taxi driver.

Ok, but check this out: self-driving buses! It's easier to make a self-driving vehicle that follows a fixed route. Buy enough buses to cover the peak demand, then send most of them back to garage during non-peak times.

Buses are a half-million dollars or more, without fancy self-driving tech. Doubling your fleet (to double peak capacity) is a big investment.

Garage and maintenance costs also increase with the number of buses.

> Otherwise people are just going to take their car because they actually have control over their own comfort.

it depends on your definition of confort.

Ownership of time ressources for mental cognitive activity > sitting.

So in this case, inside a bus with a driver is a better option that consuming my mental cognitive activity for driving. From my POV, of course

This is a thread about taxis and self driving cars.

reply to the edit:

Well, when I was working in the City and living in south London I used to take the DLR (which is a self driving train btw) and walk for 10min.

I can confirm that there were drivers on the streets as well as cyclist, pedestrians, bus riders, tube riders and so on.

All of those at their limits at rush hour. Later the city banned private vehicles from certain locations at rush hour, so you can do such changes to favour one transportation method over the othet but it's far from solving it. It's a very hard problem, very unlikely to be solved without redesigning the city.

> I'd gladly pay twice as much for it to be better.

Some other comments mentioned classism, and anti-social passengers, but a less emotionally charged way of framing it is just in terms of price discrimination.

Busses and subways have no "first class" or "business class" section. The passengers who would pay double or triple are doing so in their car payments, insurance, and gasoline, and riding in relative comfort, and the people left riding the bus are mostly those who aren't willing to pay more.

It's going to be difficult to get people to put their car payments toward better bussing, while still keeping it accessible to those who want to pay less.

Yeah I think this is the main challenge - you can have private buses are that are comfortable and provide a net positive over driving yourself. These exist in the bay area (FB, Google, Apple all do this) but then they're not public transport.

> Busses and subways have no "first class" or "business class" section

There are. For example: the Shenzhen Metro Line 11 https://en.wikipedia.org/wiki/Line_11_(Shenzhen_Metro)

> Busses and subways have no "first class" or "business class" section.

It depends. New York City has "express buses" that are basically long-distance buses repurposed for commuting. You pay an additional fare and are more comfortable. Commuter trains are of course more comfortable than subway trains. (This is why I usually fly out of JFK. Taking LIRR to the airport is much nicer than taking the bus to LGA.)

Japan's rail systems have higher-tier options on some route. JR has "home liners" that are long-distance passenger trains basically repurposed for commuters (actually very similar to the default commuter train in the US, whereas the normal commuter trains are very similar to metros). Tobu has the "TJ Liner" which is similar. The demand does exist.

I think the reality is that public transportation is too popular to reduce capacity by having more seating. Everyone in NYC would pay for a more comfortable subway, but there wouldn't be enough trains to carry them all. The completely-packed-at-rush-hour is a byproduct of the fact that cars just aren't faster in NYC, so you might as well pay $2.50 to get home faster but in less comfort. (I've looked into it. There were many times when I was running late and thought, "maybe I'll just take an Uber to work today". But the delayed subway would take 20 minutes and the private car would take 30 because of traffic. It only made a difference in the middle of the night when the subway was running on 30 minute headways and there was no traffic. Even then, still close to 20 minutes to drive. The subway is just that efficient, despite everyone's complaints.)

Right, which is why public transport should be funded via taxation.

The way to achieve a reasonable service is actually to tax driving at a higher rate (to account for the externalities; pollution, congestion etc) and put the funds into infrastructure like buses/trains.

I often drive because it's more efficient and comfortable than using the bus. It's more efficient and comfortable than using the bus because the externalities are not priced in.

We can address that balance by taxing cars higher and putting on more buses / paying for more cleaners, etc.

> Otherwise people are just going to take their car because they actually have control over their own comfort.

The fact that many people don't indicates there are other factors at play.

I bet there's a diminishing return to this. The more buses you add at rush hour the more traffic you cause and thus slow things down reducing capacity. In heavy traffic buses tend to clump together as they get stuck behind one another due to the nature of smaller vehicles around them.

Also, what would be an acceptable cost increase to pay for this additional capacity? You'd have to balance the theoretical increase in ridership against the decrease from raising prices. Keep in mind that public transport is usually heavily subsidized, in the US its on average 50%.

I totally agree, and they can take care of the last mile problem for trains at both ends, making trains more practical.

And GP said, "If the solution is flawed even a little bit, people will die." Well, yes, and if humans drive, people will die.

If the flawed machines reach the point where they are provably statistically better than the flawed people who currently drive, lawsuits claiming that people who choose to drive and get in accidents are negligent will cause insurance companies to jack up the rates on "self-drivers".

That's where the state-change sigmoid curve will suddenly turn upward.

Also if you live for a part inside the geofence, You can give full AI autonomy inside the geofence and take over outside. This would be especially useful when you have to travel long distances on the highway.

The cost floor of driving even with automation is above the cost of public transportation in most cases. Because of traffic, the service standard will never approach public transportation in dense areas

It is seeming more likely by the year that Tesla is one of the biggest over estimators when it come self-driving cars. Musk has repeatedly promised "autonomous vehicles", and even promised to have a fully autonomous vehicle by the end of this year. He is notorious for his rather bold and exciting predictions, but fewer and fewer seem to come to fruition with each passing year. It seems especially unlikely considering Tesla doesn't even test any autonomous vehicles on public roads, unlike many other companies in the space: https://www.dmv.ca.gov/portal/wcm/connect/96c89ec9-aca6-4910...

It's disappointing to hear that Ford overestimated, however, it's reassuring that they publicly recognize the challenges and also put safety above being first out the door.

disclaimer, own a TM3 with full autopilot and purchased full self driving.

I have never believed they will deliver much more than level 3. I did buy the FSD package recently when it was discounted mostly for resale value; it was 2k; and not faith in the technology.

Too many companies want to label it a safety feature yet for the most part it doesn't operate like most safety features. almost all demos we see are fair weather events on simple roads or in some cases systems which can operate on specific roads. so I think they focus too much of the navigation part.

I am more interested in full self driving technology if it leads to more passive safety features like the forward emergency braking. as in an always on system which watches "over your back". that to me is the real value of all this research.

(personal usage stuff is below)

now don't get me wrong, Tesla's system is amazing for what it does do. It certainly isn't full self driving but it offers many of the features people could use to augment their driving. I used it for the majority of a six hundred mile drive both directions. It was useful in rush hour traffic, the stop and go variety, as well as the long haul.

two reasons I found it valuable. first being on long drives I engage it but still maintain the presence of mind to take over when I need to. In this mode I am using it to handle situations when I am distracted. The old issue of looking in your rear view mirror and seeing a bridge you went under but don't remember. I also used it on a typical mid Western two lane road at night in the pouring rain. It helped alleviate fears caused by the constant issue of people with poorly aimed lights or not dimming high beams. The car never lost track of the lanes whereas a few times I was having to hard focus on the white line to avoid being blinded by one of those hibeam drivers.

I agree 100% on why I purchased FSD at $2k. But I'll add that I almost never use Autopilot, as I've had too many negative experiences with it. I do frequently use (and generally love) the smart cruise control feature, though - steering the car in traffic just isn't that much work for me and I don't have to worry about the car lunging to the side when an exit ramp splits off.

Yup, have a Subaru and adaptive cruise control is great, but I doubt I'd want auto-steer. Steering just isn't much work but keeps you attentive without causing the fatigue that speed management does.

He is lucky if we have self driving cars by the end of the next decade. I am sorry but false positive and false negative situations are not ok when it comes to self driving cars and most if not all solution exist today are based on technologies that produce false positives and negatives.

https://arstechnica.com/cars/2019/03/feds-investigating-dead...

https://www.youtube.com/watch?v=-2ml6sjk_8c

> Tesla is one of the biggest over estimators

Tesla promised that the current hardware the cars come with is all it's needed to implement autonomous driving at some point in the future, which is currently an unknown unknown. if that doesn't scream of snake oil, I don't know what does.

Tesla's big bet on autonomous driving is very strange to me. They already have a unique selling point with fully electric cars with good performance and design, why are they overextending themselves to try to be first to market with such an advanced and difficult feature? They're a small company with little money, focus on the fundamentals of selling cars first.

It's like the falcon doors on the X, unnecessary, expensive, and caused delays when regular doors would have been just fine. But no, Tesla had to be different, for what? Elon's ego wanted cool doors?

It may not be enough to validate the choice, but the falcon doors make the X a minivan without the stigma of a minivan.

I'd say that the X is more of a crossover than a minivan. A true minivan would be the FCA Town & Country, Toyota Sienna, or Honda Odyssey.

Other SUVs accomplish that with regular doors.

The problem is that to get where they are they've needed huge investment and issued huge amounts of debt. They are only able to do that because of the huge valuation justified by self-driving technology. If they end up as a company churning out great electric vehicles that is not going to justify being valued at 50% more than Ford.

What costs Tesla money is building great big battery factories and car factories, meanwhile the the software engineers working on self-driving are a relative drop in the bucket but justify probably 80% of the market cap.

Their market cap was already over $30 billion in August 2014, before the first version of Autopilot was announced.

Reason Tesla is having trouble with autonomous vehicles: Elon Musk != Nikola Tesla.

People will die with self-driving cars. The only thing that matters is how many people are dying because of self-driving cars vs dying without. That's it. The moment it's an order of magnitude less, then self driving car will become default.

That is not the only thing that matters. There is some interesting psychology at play here.

One thing I see rarely discussed is the fact you are getting killed by a programming error, not a human. Of course, the humans made the said programming error, but they have no face, there is nobody (individual) to blame, no-one to apologise, no-one to mourn for their actions.

At least if someone is hit by another driver, they will be prosecuted in the case of negligence, or if it's due to a mistake, be able to apologise. It's human. It's different.

I don't know. People still fly despite the occasional mechanical error, programming error, or other non-individual error leading to fatalities.

I think if it is an order of magnitude safer that would override many of those emotional concerns. Insurance companies are decent at thinking rationally about problems. People might have emotional concerns, but I think a halved auto insurance bill would be enough to sway a critical majority of people.

But when you fly there is still a human face ‘in control’ (even if not to a sufficient degree) you can blame: the pilot.

That's a good point. Pilots are still an important psychological factor that makes it easier for people to fly.

You've articulated the idea underlying a nascent field of research: algorithmic accountability.

> Of course, the humans made the said programming error

Are you sure?

You do understand that self-driving vehicles inherently rely upon machine learning systems (and I am not talking simple regression here), which are trained and ultimately make decisions by means that we still don't fully understand?

What I'm trying to get it here is that, barring faulty sensors or mechanical systems, any decision that the control system of a self-driving vehicle makes is most likely arrived at via the inner workings of one or more machine learning systems (likely deep learning neural networks of some nature, but there are other systems being experimented with as well).

They are not a huge series of if-then-else statements or anything of that nature; such algorithmic approaches have been tried in the past and dismissed as unworkable due to the sheer complexity that is inherent in driving.

Honestly, what drove the field forward was mainly work done by CMU, mainly ALVINN; but there was interplay between CMU, Hans Moravec, Stanford (the Cart, among others), Thrun, and others. It has a very long, convoluted, but detailed and rich history (I encourage anyone who's interested in the fields of robotics, ai/ml, and/or self-driving vehicles to read up on it - it's very fascinating). ALVINN ultimately gave the hint toward using neural networks and deep-learning, but the tech couldn't make the leap forward until the convergence of a number of technologies.

I know this. You know this. But we are in a technology bubble. Machine learning also has no face to the regular public. Telling to the public that AI killed their relative is even worse than saying the team at Google did it.

> The only thing that matters is how many people are dying because of self-driving cars vs dying without.

This is not how human emotion or layperson risk assessment works.

> The moment it's an order of magnitude less, then self driving car will become default.

This is rational but I don't think it's true. ~37,000 people die from car crashes every year in the US, from about ~35,000 accidents. Each individual Tesla crash that results in any serious injury or death is still a national news story. Humans are bad at estimating probability; it's hard to imagine popular support for self-driving cars if they kill thousands of people each year. Even if the alternative is tens of thousands of deaths, we have already collectively priced that in as a risk of living in a developed, automobile-driven society, such that fatal crashes never rise beyond local news.

You're off by some number of orders of magnitude on the accidents, unless you mean, "There are ~35,000 accidents that involve one or more fatalities (and many more that do not)."

> ~37,000 people die from car crashes every year in the US, from about ~35,000 accident

OP reads correctly to me

A third of fatal accidents involve drunk drivers.

I find it interesting that the apparent objective is to make driving safer, but apparently the powers that be have decided that it is easier to make cars that drive themselves than to convince people not to drive while drunk.

Quibbling... 28% in 2016 (https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/...), but this is after decades of awareness campaigns and enforcement. Then there's the ~70% of fatal accidents that happen with sober drivers. There might not be more to be had by addressing human factors (though I personally believe graduated licenses with more rigorous testing and much higher investments in public transit would help) so the focus on self-driving vehicles may not be misguided.

> The moment it's an order of magnitude less, then self driving car will become default.

I have no surveys to back this, just intuition. But I don't think one order of magnitude is enough to counteract the fact that people don't feel as if they have control over their own mortality.

i.e. it'll have to be a lot better than just one order of magnitude before people can be swayed towards putting their lives in the hands of a robot.

That's not how people will see it.

Nuclear energy is the safest form of energy, it has much fewer casualties than gas, wind or solar. The difference is the risk of nuclear is indiscriminate, whereas people dying in a solar accident are the workers (rooftop fall).

If the driver is in control of their own car, they have themselves to blame (mostly). With self-driving, they are trusting their lives into a single entity they understand nothing about.

^This.

The census don’t, generally speaking, tend to believe statistics and facts. And the media will always spin a bad situation to boost their click through rates.

So, even though self driving cars may be 100% less prone to accidents (this is just an example), media headlines like “driver killed instantly when self driving car automatically steered into a wall without warning,” will cause the majority to push back and say they’re unsafe.

> “driver killed instantly when self driving car automatically steered into a wall without warning,”

Interesting example. A justifiable fear people have of self-driving cars is them crashing in scenarios where no (minimally competent, not ass-blasted drunk) human would reasonably crash, like randomly steering into a wall.

I think if self-driving cars were seen to crash in scenarios where a human could have made a similar mistake, their reputation would be better.

Not to say the media reporting of these incidents would be in any way accurate or fair.

This is why self-driving cars will feel more dangerous, even if statistically they are safer. I wouldn't much feel better if in the moments before my car drives into concrete barrier, after having determined impact is inevitable, it informs me that while it might have failed me and I will die, that statistically it has saved 10 other theoretical people.

If I had to guess, the decision will always be to protect the occupants of the car over everything else. No one is going to buy a car that will kill you to save others.

That's not what I was saying. That a car is statistically more safe is not comforting when it kills you in a way you would have easily avoided had you been driving.

there are plenty of sober people that randomly steer into walls/trees/you name it, every day

Nuclear energy is a good example of the situation that self-driving cars find themselves. A common saying in my industry is that with safety critical systems you do not get to decide whether the systems kill people or not, the best you can do is decide how many people get killed. Unfortunately for self-driving cars, like nuclear energy, even if the number is far less than existing technologies the public has little appetite for the indiscriminate death.

Nuclear power generation is also being used, you just don't hear about it much. You can see California's power supply by fuel here: http://www.caiso.com/TodaysOutlook/Pages/supply.aspx

Right now it's around 10%, but it's different for different regions.

I think the problem with nuclear energy is the overall number of people who die per incident. Because of this, the deaths are overreported.

We seem to be able to handle people dying one at a time, but once a whole town is in danger, even if this is something that happens once a generation (this is the hard part to understand), it crosses that threshold of "this isn't worth it".

That said, if self driving car deaths get reported as much as traditional car deaths (that is to say, not very often), perhaps people will think they are just as safe as driving themselves.

>I think the problem with nuclear energy is the overall number of people who die per incident.

IIRC those numbers are very, very, low. Your village should worry about the dam upstream, not the nuclear power station upwind.

Even if overall self-driving cars are safer, they won't feel safer to the individual if the situations in which it will kill them are situations they could avoid if they were driving themselves.

The self-driving car might be able to avoid that t-bone situation more often than the driver can, but if the death scenario for the self-driving car is it gets confused and drives into a river, that will feel unacceptable to a human because it seems so simple to avoid to them.

Most people aren't concerned how many other people survive, but are generally primarily interested in their own survival over population-level statistics.

I bet it'll be more about money than lives. As soon as car insurance is an order of magnitude more expensive when "driving yourself", then we'll have self-driving cars as default!

It's not like that's going to happen next year or ten years from now.

Insurers like boring, happens at a consistent rate, human error accidents so long as their slice of the population is not making them at an appreciably different rate than the competitors' slices.

Insurers really don't like hard to predict accidents that cause people to max out their collision and medical (which is exactly what would have happened in the Tesla cases had the customer not died) which is exactly the kinds of accidents self driving cars seem to be having.

It was very exciting when autonomous vehicles from a handful of companies first began demonstrating the ability to go a few miles in complex traffic and pull off a few fancy maneuvers along the way. I got really excited when I saw that, and for the engineers building these systems, it did go to their heads, and timelines to scaled deployment went from being 20 years away to being 5 years away in the span of 18 months.

When deep learning and simulation were first applied huge gains were made, and the limits of these tools were not fully understood. Now that these tools are better understood, we are back to old fashioned robotics development timelines. You gotta build features, and test them, over and over, down to a very granular level. Every time you roll out into a new geography, there is a new mountain of problems to be solved. To drive the error rate down another order of magnitude, the workload goes up exponentially relative to what was needed to achieve the previous order of magnitude. It's like inverse Moore's law.

Boston Dynamics had bipedal robots balancing dynamically and walking around in the early 1980s. 35 years later, now that we have a billion times more compute to throw at the problem, well, BD's robots are still mostly just walking around, and far from being good enough to walk around in the real world. That's the real rate of progress in robotics. The prospect of autonomous vehicles has drawn billions of dollars of investment towards solving these fundamental robotics problems, and certainly that has accelerated things, but how long does the money need to keep coming in for before it starts paying out? That's anybody's guess.

BDs controllers were/are all hand-tuned using good old convex optimization. Good luck getting fancy 'Deep-RL' to work on these without multiple man years tweaking every little knob in the cost function and the algorithm.

Trains involve massive capital outlays and 60% of bus operating costs are for the driver, resulting in buses that are larger than necessary and therefore increasing the non-driver costs.

If we had self-driving vans we could provide better and cheaper transit to everyone.

> Trains involve massive capital outlays

And self-driving cars don't? How many trains could be built with the billions invested so far in self-driving cars?

That's a one time cost. If we get the technology working building more self driving cars will be much cheaper than building new trains.

Self driving cars are simply a backwards-compatible transit solution. We have already invested so much money in car-based transportation to the detriment of trains. But, we will still need to continue paying billions and billions to repair and build new roads.

It costs NYC about 2.6 billion to construct a mile of subway.

In light of that, the money invested in self-driving cars is pretty small.

either that or NYC is just really inefficient in terms of cost.

Keep in mind you've cited probably the first or second most expensive mile of any subway ever built, and subways usually cost much more than other trains.

China regularly builds high speed rail for $30M per mile. Europe is usually closer to $50-60M. Tokyo has built subways for $400M per mile.

https://www.nytimes.com/2017/12/28/nyregion/new-york-subway-...

Beyond the geo-fenced solution, and system like Super Cruise, I doubt that we'll see truly self-driving cars for another 30 - 40 years.

What fascinates me is that company like Uber believe that they can use their current business model to finance the development of a self-driving car and the start to make profit when those cars a ready. That would mean investors pumping money into the company for at least 25 years more.

It makes more sense to slowly develop self-driving cars for technologies like Autopilot and Super Cruise. Gradually improving the solution until one day it can safely navigate a leaf covers dirt road.

Strategies like Autopilot are the worst of both worlds: They don't free up the driver and they are still susceptable to faults. Driving aids which engage to prevent an accident are a useful way to improve safety, driving aids which mean the driver is not in control and is just monitoring the system in case something goes wrong increases the likelihood of accident significantly.

> just monitoring the system in case something goes wrong increases the likelihood of accident significantly.

That depends on the rate of system failure. I used an early form of autopilot and you quickly started to pay attention to where the system would fail. It was like cruse control where you look for situations where you can turn it on but never stop paying attention.

At the other end with really safe systems they are going to default to safer than a human driver. So, there might be a significantly deadlier middle ground, but no company has deployed anything like that yet.

> I used an early form of autopilot and you quickly started to pay attention to where the system would fail. It was like cruse control where you look for situations where you can turn it on but never stop paying attention.

You may do this, but I bet the majority of people do not. I know that Google has data about it's self-driving cars that they know their "safety drivers" simply don't pay attention. In Uber's case - here in the Phoenix area last year - this inattention helped to cause an arguably avoidable fatal accident. These were people supposedly trained not to do this - yet they did.

As far as cruise control is concerned - there've been more than a few accidents of people thinking it did more than it does, and not pay attention to the road in front of them. Tesla's autopilot system, while better than simple cruise control, has - because of this better system (and poor marketing) - caused people to think it does more than it can, and has led them to sleeping in their vehicles while it "drove", to being completely inattentive, etc.

It is pretty well known at this point that those "middle levels" of self-driving technologies are things that shouldn't be implemented, because people either misunderstand them, or misuse them, leading to often fatal accidents.

When I mean early I am talking a system that stops working every 60 seconds or so. The constant failure instilled a fair amount of paranoia, and did not give enough time to cause boredom.

Google’s cars are in that middle ground, but Google has also not released their system. Presumably they are going to wait until it’s as safe as the average driver before releasing it.

I've been wondering: why don't the self-driving car companies build a system that requires a person to drive it but assures the human can't crash the car? In other words, if the self-driving models predict that the human inputs will result in a crash it forcefully overrides the throttle, brakes, or steering wheel.

The sensation would be like driving one of those tracked cars at an amusement park. Try and slam the wheel hard to the right to steer into a guard rail? Computer would physically block it from turning past the point that you would leave your own lane.

This system would:

  * force human drivers to drive the car 100% of the time
  * allow real world testing and enhancement of self driving capabilities (data gathering, model backtesting, etc.)
  * be strictly better than both fully automated and existing non-automated systems

We have the beginnings of this with the new auto-stop capabilities that are showing up in new cars. Why not build this hybrid system as a stopgap until full level 5 automation is working?

> requires a person to drive it but assures the human can't crash the car?

Because the 'not crashing' part is the hard part, the rest is just navigation.

Because now you have the worst of both worlds. The computer still has the final say in what the car does, but now you have a fallible human in the loop at all times as well. If the computer really can predict crashes better than the human, then have the computer drive the car. If it can't, then letting it override the human isn't a good idea.

Example: you slam the wheel to the right to steer into a guard rail. The computer overrides you and you remain in your lane. You flatten the toddler you were trying to avoid, that the computer didn't see.

Uber knows the routes ahead of time. They can decide if the route and time of day is safe enough for the limited self driving technology they employ or send a human driver if their SDCs can't handle it.

Unfortunately too many people subscribe the idea of startups bringing disruptive changes into car transportation because other startups could do the same with entirely different problems. Just because Elon Musk created Paypal it does not mean that he will be able to jump the hoops to get self driving cars on the roads by end of this year. Many people who are intimately familiar with self driving car technologies predict that we need at least 10 years and several innovations before we can seriously consider this as an option.

Of course you're correct that they might not pull it off, but it's not like people are basing any confidence on Musk's Paypal background. Tesla already brought disruptive change to car transportation by making electric cars popular. And Musk brought a third big change by commercializing reusable rockets.

I wonder what the model of survival is for Uber's regional competition like Didi and Ola. I can't speak for China, but driving in India is way way difficult than in the US.

I think self driving research has already benefited cars substantially via driver assistance, like you say. Recently I was in a base model Toyota rental (camry or corrolla I think) and was surprised when it automatically warned me that someone was cutting me off in traffic, before I even noticed them move a few car lengths ahead. Then I noticed there was automatic radar cruise control built in, which made the rest of the drive much nicer. I hadn't even looked too closely before since I assumed that tech wasn't pervasive enough yet to trickle down to a basic rental car.

I think that the technology is probably already saving lives - Volvo is trying for zero fatalities right now, and I think driver assistance will reduce the lethality of cars substantially before we're able to adopt self driving vehicles, which would likewise make it less of a disaster if we aren't able to adopt self driving vehicles in the short timelines originally predicted.

I think you're discounting exponential growth. We (humans) overestimate the short term, and underestimate the longterm.

On the contrary, humans are keen on extrapolating "linearly", ignoring that progress is discrete, and that the next discrete step isn't on any guaranteed timeline. See: humanity's idea of the next twenty years of space exploration in 1970.

Humanity's ideas of space travel 1970-1990 were right on, assuming US government investment had remained the same (4.4% in 1966, <= 1.01% since 1975) https://www.theguardian.com/news/datablog/2010/feb/01/nasa-b...

Technological growth is an S curve. Exponential at first but hitting a point of diminishing returns after. The 747 is roughly equidistant from the Wright Flyer and the 787. Aviation hit the wall before we got, for example, routine supersonic flights. In computers, the 1.5 GHz Pentium IV is roughly equidistant between the 8 MHz 286 and the i9-9900k (which clocks up to 5 GHz, but that’s not even a fair comparison because the 286 dissipated under 15 watts, the Pentium 4 was a 60W chip, and the i9 is nominally a 95W chip but in reality can dissipate much more than that during boost.)

> If the solution is flawed even a little bit, people will die.

Of course they will, but as long as it's more convenient and the death rate is lower than human drivers, they will take off.

Will regulators allow machines to kill people? Human drivers killing themselves and others are a part of human nature, but will politicians and the public accept that people get killed through no fault of their own by mistakes machines make?

> Human drivers killing themselves and others are a part of human nature, but will politicians and the public accept that people get killed through no fault of their own by mistakes machines make?

This has already happened with manufacturer defects. It's just more of the same. As long as a clear liability boundary can be established in principle, there's no problem.

If autonomous driving can be delivered with accident-causing failure rates similar to those of general manufacturer defects (i.e., much lower odds than accidents caused by human error), then I think that would be sufficient. But if say we go from ~35k human-caused road deaths/year to ~10k autonomous-driving caused deaths/year, that may be seen as unacceptably high.

> But if say we go from ~35k human-caused road deaths/year to ~10k autonomous-driving caused deaths/year, that may be seen as unacceptably high.

Sure, but there wouldn't be an outright ban. They'd instead establish higher standards that need to be met.

Even ignoring the answer to that question is yes for other industries. They are allowing SDVs even after the Uber/Phoenix incident, so we already know the answer to this question.

The sobering thought is if you work in machine learning, big data, or talk to someone who does.

What do you consider a good accuracy rate for your machine learned / hand-engineered heuristic? 95%? 99%? 90%?

We've built some real marvels of machine learning, but you can get away with the occasional not-very-rare horrible failures in speech recognition, photo labeling, or machine translation, and it just results in some giggles on Reddit.

> If the solution is flawed even a little bit, people will die. That will give self-driving cars a bad reputation and people will be unwilling to ride in them.

There is lots of debate on every autonomous car story, including this one, about "is as good as humans good enough", but that's a technical threshold. The real blocker (from my view) currently is and will be liability for "negligence."

If I'm driving and fiddling with the radio and as a result blow a light and kill someone, at most you can turn out my shallow pockets in civil court. The automaker has no liability.

But if the _car_ is driving and blows a light and kills someone, then the automaker can and will be sued because they've got deep pockets. I can see the ads now running during the Price is Right. "Hit by a runaway robot car? Jon from Orlando was awarded $3.2 million! Call Richard Cranium Attorney at Law and get yours!"

Both GM and Ford have been hiring a lot of software engineers in Detroit in the past few years. The problem is that they need NASA-style software engineers (crazy-anal nit-picky consider-every-possibility style engineers)--not engineers from the regular software industry. If your car crashes, you can't just reboot it.

It seems to me that there's less of that than there used to be.

In days of yore, I shipped quite a few ROM-based products. Bugs were frowned upon.

That, and they pay shit relative to the regular software industry.

IIRC, it was Elon Musk that said something along the lines of it's not too difficult to get self-driving cars to 99.9% safety - but that's still 1 accident in every 1000, which is terrible all around. You need to shoot for many, many 9's before it can be considered a practical alternative.

People die today. Self driving cars just need to improve on that but a significant amount. I predict that despite not being perfect (which might not even be a reachable bar) they will become mandatory for all cars in the near future because they are that much better than humans.

Who's going to be liable for those deaths? The manufacturer? I don't think they'll be so hot on that. So really, they need to be orders of magnitude safer than humans, to the point that they are close to flawless.

I don't think that is the blocker. You can pass a law that says "manufacturers that meet XYZ self-driving certification are exempt from liability".

The question of liability is always tough to resolve. Is Boeing responsible for the 737 MAX deaths because they wrote the faulty software? Is the airline responsible for poor training or buying the version without the "disagree" light? Or is the FAA liable for their poor oversight of the whole program? Arguments can be made for all three.

Selfdriving.. oh wait, this is in the air, where there's no traffic and all these different situations. And wait, it was not even an autopilot. Even this was too hard to implement.

Lane-assist on the highway? Sure.. autopilot? Not any time soon. Probably not within 20 years

Exactly, this is also a legal one. I'm curious what happens when we get human driven cars crashing into auto piloted. Are we going to have an assumption of guilt on the human or the auto or will it actually be balanced.

I agree, but the optics are awful. It's not a technical problem, it's a psychological one.

I don't have any sources to back this up, but I'd wager that people, on average, would prefer to take a risk an order of magnitude greater if they believed they had some degree of control over the outcome.

I.e. let's say 10% of conventional cars experienced a fatal accident, and 1% of self-driving cars experienced a fatal accident. I think most people would choose to be in the 10% fatality group because they feel like they would be in control.

Anecdotally, I have a co-worker who hates flying because she feels like she isn't in control, even though aviation deaths are extremely rare.

Humans are deeply emotional creatures and it's most evident when you observe group-level decision making.

Individually, you and I can both agree that the bar is quite low for self driving cars. It just needs to do better than current drivers to be safer. We should adopt it immediately, if we can just prove that it crashes less.

Society doesn't see it that way. 1 crash is horrific. 1 accident, is proof the technology is flawed. 1 accident is 1 too many.

It's stupid, but that's how it is.

The metric shouldn't be #crashes but #innocent lives lost. I've never seen a study that accounts for all the influences to prove that autopilot and co. are better in that regard.

> If you are going to geofence self-driving cars it is questionable whether they are better than trains and buses.

If they could manage to geo fence them to highways and get full functionality there it would be a big win for commuters and road trippers.

However the companies can't then make billions off of automated city taxi services.

Exactly. It's relatively easy to build a self-driving car that works well on good, well-marked highways in good weather in the daytime when every other driver on that road is acting responsibly. But that's just the base case, and the edge cases are wickedly hard.

> If the solution is flawed even a little bit, people will die. That will give self-driving cars a bad reputation and people will be unwilling to ride in them.

Some people have died while using elevators, but I don't see the rest of us not using because of that.

> Some people have died while using elevators

For some reason I forgot about this in an earlier comment I made about people wanting someone to blame.

But here we have what could be argued as a "simplest version of self-driving vehicle, with no human operator at the controls, and no one to immediately blame when somebody is hurt or killed by the system"...

So the question is - has there been a study done on how people respond to such incidents, when they have no one to blame? At least, in the moment? They can't "yell at the elevator". Eventually, they may yell at the owner of the property or the manufacturer of the elevator, or whatever company last did maintenance (ie - yell -> lawsuit).

But does that stop them from using elevators ever again (I imagine that a percentage do become fearful; I know I read something of that nature regarding people who've been "trapped" for a long period of time in an elevator, especially those who are alone when it happens)?

Also - why do others continue to use elevators, and not take the stairs (this is more a rhetorical question - stairs are not an option in a skyscraper over more than a few stories for most people)?

I'm just interested in why a self-driving car is so different to people, beyond the relative "novelty" of it? Will we completely reject it, because it can't be made 100% perfect? Or will we get used to them in time, even though statistically someone somewhere will die occasionally from an accident in one?

I'm not looking for answers here; it's just something that I ponder when these kinds of discussions come up, as I've had a bit of minor experience in learning about (and how to create) these kinds of systems.

The parallelisms don't end there: elevators used to be much more complicated that they are now. Instead of having one button per floor they had motor controls. They literally had to be "driven" by human operators.

I don't remember if any studies were cited, but the "elevators vs self-driving cars" was studied in the Planet Money podcast: https://www.npr.org/2015/07/31/427990392/remembering-when-dr...

Agreed. Advanced driver assistance is where the market opportunity is for a considerable time. From there it can evolve closer to full self driving. However full self driving will never be “done”

So we want a car that we don't want to drive. It's calld a taxi.

I used to think this too -- but Tesla has killed half-a-dozen or so of their customers with no apparent brand repercussions, let alone an impact on the broader AV space!

Self driving trucks on highways driving from hub to hub with manual drivers for the first/last miles seems doable today. That’s most of the way to a working solution and still a major potential cost/labor/crash savings.

Reply on news.ycombinator.com