Elon Musk: human drivers are 'too dangerous'

Elon Musk: human drivers are 'too dangerous'(wired.co.uk)

45 points by riaface 11 years ago | 78 comments

I'm a little bit confused about why there's such a negative reaction to this on HN - it seems like a pretty reasonable statement to make, no? Sure, I don't think there's any firm idea of when 'robotic' drivers will overtake humans in terms of safety. But it does seem pretty clear that it will eventually happen.

Self-driving cars have a few massive potential advantages relative to humans, such as faster reaction times, no distractions, the ability to directly communicate car-to-car, and so on. It certainly seems likely that those advantages will far outweigh the adaptability of humans – after all, aren't most accidents caused by distraction, driving too fast for the conditions and the like?

drzaiusapelord 11 years ago | |

Because HN'ers are results based people in general and the results on autonomous cars are currently terrible. Still no solution to driving in the snow, heavy rain, and other poor visibility conditions. Still no solution to weirdo edge cases like a madman running up to your car with a bat (I used to live in a poor neighborood and had this happen once and accelerated away. My dumb smartcar would have stopped to avoid hitting him and let me be victimized).

Personally, I'd rather take my chances with the occasional drunk than worry about Google's bugs killing me because some H1B engineer with 4 hours of sleep was forced to write some really junky code that didn't understand some edge case. I look at what Google is able to do practically like Android and am not impressed. Even the best code quality for automation, for example NASA's work on rovers, is wrong occasionally and those things move at a snail's pace with almost nothing around them! The state of AI itself is in shambles. Its healthy to be skeptical of extraordinary claims like Musk's. We just aren't there yet and may never get there considering the fuzzy logic that driving requires.

Not to mention the anti-progressive thinking that better cars are the solution to our transportation woes and refusing to accept that we will reach a post-car age in urban centers sooner than later. This is like building a better horse drawn carriage. Sure its technically impressive, but its just not the real solution here.

matthewmacleod 11 years ago | | |

the results on autonomous cars are currently terrible

Sure, they're not there yet. But in 20 years they'll probably be much better. Might not happen, but it does seem pretty likely that technology will continue improving.

Still no solution to weirdo edge cases

So that might be one of things that are traded off against the benefits. How often does a car get attacked by a baseball-bat-weilding maniac? Is that risk worth ignoring the improvement in safety? Are there other solutions to the problem?

Not to mention the anti-progressive thinking

I can't agree with this. It's something common to hear from green-leaning movements around the world – that the car should be retired altogether in favour of public transport.

That's true to some extent, but public transport (and it's excellent in many cities around the world) can't completely replace individual vehicles. Regardless of how much it improves, it does not offer point-to-point connections, or the ability to transport goods, for example.

In any case, what would be the difference between a city centre populated by electric vehicles (let's say owned by Uber or Tesla or some other company) and one populated by public transport - simply the efficiency? It might be better to think of a city full of electronic cars as being the first implementation of something like personal rapid transport.

forgottenpass 11 years ago | |

I'm a little bit confused about why there's such a negative reaction to this on HN - it seems like a pretty reasonable statement to make, no?

"Too dangerous" is sophistry that is rightly derided when used for unquantified fear-mongering elsewhere and this is barely different.

I can't wait for the future where I don't have to drive, and safer is an attractive selling point too. But there is a level of risk associated with road travel, a level society has largely deemed acceptable (otherwise speed limits would be lower). Musk putting this in his cross-hairs is a disappointing choice of sales tactic.

matthewmacleod 11 years ago | | |

I don't think I agree with this.

Road traffic accidents are something like the 10th most common cause of death worldwide. It's clearly a rather dangerous activity.

The idea that we have 'deemed acceptable' the risk involved is only the case where the risk cannot be lowered without compromising other goals (e.g. fast travel).

In a hypothetical future where autonomous cars are widespread and much safer, the tradeoff changes – it suddenly becomes possible to have both fast and safe travel, with the proviso that humans can't be in change cars.

I'm not going to argue that this is a good thing, because I don't know. But it's not fair to say that it's fear-mongering – it's objectively true that cars kill loads of people, and quite possible that non-autonomous vehicles will be banned or at least heavily regulated in the future.

manicdee 11 years ago | | |

Lower speed limits will not reduce the number of deaths, unless all traffic is reduced to 40km/h or less.

Inattentive drivers, confusing road signage, insufficient separation of pedestrians and cars, drivers not following established rules and guidelines: these are the contributing factors which can be addressed to reduce road fatalities with greater reduction of fatalities than simply reducing speed limits.

Autonomous cars won't speed for thrills, they won't cut lanes or run red lights. They will be far safer, if more annoying, than most drivers: just like the more conservative human drivers we already have.

cheatsheet 11 years ago | |

The only thing I am concerned with is how secure / exploitable they are from a computer security perspective. I assume there will be some kind of link between the vehicular peripherals to vehicle control and configuration. Then I worry about who is actually controlling the car. I'm open to AI driven cars with redundant intelligent networks, but people are smart too, and some people are not nice people. How do you know that the entire system is functioning as expected? I'm pretty sure that's NP hard.

undersuit 11 years ago | | |

> people are smart too, and some people are not nice people.

What did you do to these people? Because there are two reasons these people aren't going to be nice to you: you offended them, or you're a random target of violence that we already experience, but now it takes the form of a homicidal self driving car instead of a boy in a movie theater with automatic weapons.

Are you advocating we stop pursuing self driving cars just because security is an open ended problem?

matthewmacleod 11 years ago | | |

Definitely a concern, but I'm not sure what can be done about that – except for "let's develop the technology so that can't happen", even though that seems unlikely.

closetnerd 11 years ago | |

I think the reason you're confused is because you're under the impression that people are arguing against the very obvious practical benefits of having only self driving cars: congestion, safety, etc. This is obvious to everyone.

However, there are other human concerns. I personally enjoy driving. I often go on drives where I'm simply exploring, not knowing which turn I might take next. It's my personal belief that driving is not just my privilege, but my right. If you want a self driving car, you're well within your rights to do as I am to drive my self.

Moreover, there's a huge scope of privacy issues and increased amount of control being put in the government. I am one of those individuals who is inherently distrusting of government policies that can restrict us under the guise of safety, etc.

matthewmacleod 11 years ago | | |

That's an interesting concern – I suppose we end up, like with most issues of public policy, trying to draw a balance. I imagine it's much like gun control: while most gun owners are going to be responsible, that damage that can be caused by those who aren't means that many jurisdictions prefer to restrict firearm ownership.

Sure, and there's good reason for that. The privacy issue is orthogonal and must be set to one side (though it's legitimate).

How would you feel about a move to regulate non-autonomous vehicle ownership more closely? Driver licensing could become much more rigorous, for example. Combined with the essential outlawing of non-autonomous vehicles in urban centres (almost inevitable in the long run), that would ideally assuage most safety concerns while still allowing people the freedom to drive.

panglott 11 years ago | | |

The right to travel is a very ancient right, from a time when people only walked, rode horses, or used draft animals to travel. The introduction of mechanical vehicles in the late 19th century changed things.

Especially, motor vehicles are so uniquely hazardous to their operators, passengers, and bystanders than operating them quickly became a privilege, not a right. http://www.bicyclelaw.com/road-rights/a.cfm/road-rights-cycl...

fricken4 11 years ago | | |

Mobility should be a right, driving shouldn't be a right. If someone is too old, too young, too poor, disabled, or otherwise unable to drive, and unfortunate enough to live in a car dependent area (as most inhabited areas are) then their rights are being infringed upon by your driving and others. For most of those who are victims of vehicle accidents- which number in the millions annually- what about their rights? Most of them did not cause the accident- they were passengers, other drivers, cyclists, pedestrians etc.

dagw 11 years ago | | |

I can't see a ban coming anytime soon, but I can see self-driving cars only roads. If you really want to/have to drive yourself feel free to take the 40-50 mph speed limit side roads. Or you can take engage the auto-drive and hop onto the speed limit free express way.

astazangasta 11 years ago |

The thing I keep emphasizing in this conversation is how terribly good human drivers are, on average. In the US, there are three fatalities for every 10 billion passenger miles traveled. That's tremendous. A machine that is currently incapable of driving in rain and unable to see a child in the street has a long way to go to hit this mark.

Musk is here to sell his car. He'll say all sorts of bullshit to make it look sexier.

ohsnap 11 years ago | |

What could be tremendous about that statistic? It seems like you are trying to contrast a small number with a big number to give a sense of some false magnitude. I could say people only murder other people once in every 100 billion seconds - thats tremendous!!

Retric 11 years ago | |

That's a load of FUD.

Modern Self driving cars track people on sidewalks not just the street. As to driving in the rain, they can do it, but like poeple there less safe at highway speeds in the rain. https://www.youtube.com/watch?v=YXylqtEQ0tk (6:45, 9:00) Also, 3:52 demos stopping for someone in the road.

They even read stop lights (8:00).

PS: If you actually look at that left turn on @9:00 the car has a much better idea what's around it than people do.

FranOntanaya 11 years ago | |

Miles per passenger depend on too many factors. I'd stick with 1 death per 6,200 driver's licenses (US, 2009)

jeffdavis 11 years ago | | |

Deaths per mile is more useful because it compares the cost (deaths) with the benefit (miles traveled per passenger). Additionally, we can compare directly against alternatives, such as buses, air travel, and self-driving cars.

Deaths per driver's license doesn't tell us the benefit we are getting, and doesn't give us a way to make comparisons with alternatives.

GraffitiTim 11 years ago | |

That statistic is orders of magnitude off. There are 30,000 driving fatalities per year. At the ratio you stated the number of miles driven per person per year would be absurd.

clarle 11 years ago |

FWIW, Elon Musk did say the following on Twitter:

> To be clear, Tesla is strongly in favor of people being allowed to drive their cars and always will be. Hopefully, that is obvious.

> However, when self-driving cars become safer than human-driven cars, the public may outlaw the latter. Hopefully not.

https://twitter.com/elonmusk/status/577946893646364673

So his goal isn't to shore up support for legislation against non-autonomous cars anytime soon.

closetnerd 11 years ago | |

He made that clarification because he was met with unexpectedly large audience which was critical of his statements.

He was quite obviously supporting future legislation against humans driving cars:

> In fact, in the distant future, I think it's probably going to be... people may outlaw driving cars, because it's too dangerous.

> You can't have a person driving a two-ton death machine.

gaius 11 years ago |

Heh, I have a friend who is all rah-rah for self-driving cars. He also likes to go camping. So I asked him, when you show up at the site and the farmer says, top field, past the sheep, turn left at the tractor, what's your self-driving car going to do? Hmm, he said, well it would need some sort of touch screen interface for that. Or I said, some sort of "steering wheel"...

elliottkember 11 years ago | |

I don't think "self-driving car" and "driveable car" are mutually exclusive. Self-driving mode for roads, highways, parking, commuting... manual override for the countryside.

dragonwriter 11 years ago | |

> Heh, I have a friend who is all rah-rah for self-driving cars. He also likes to go camping. So I asked him, when you show up at the site and the farmer says, top field, past the sheep, turn left at the tractor, what's your self-driving car going to do?

I don't see any problem with that, even with no direct controls. That's probably a situation that requires more detailed navigational input from a driver than providing a destination for road driving and letting the system handle the routing, but doesn't necessarily require actually manual driving. And, even if it did, even fully automated cars that are designed for personal ownership and (especially) for off-road use are likely to have full-control overrides available for some time.

Of course, camp sites and other destinations that people are expected to drive to will probably also have navigation information or guides available that self-driving cars can follow when self-driving cars are common, rendering the whole hypothetical largely moot.

gaius 11 years ago | | |

I can imagine the look on the average farmer's face when some city folk in their self driving car ask him to lead them up to the field himself. He'd probably just leave them there and drive off in his tractor, chuckling to himself.

lordnacho 11 years ago |

Humans are not going to stop needing sleep, get distracted, or react slowly. All we can really do (and have done) is mitigate. Seat belts, driver training, etc.

Whatever problems a machine has, we can improve on them gradually. We can spend loads of time looking at test cases, and improvements are ratcheting.

I can't wait for self-driving cars. It will be a massive change to be able to go to sleep in your own comfy bed and wake up in time for a meeting in another town.

I also imagine traffic will be smoother, as tolerances don't have to be so wide.

Methusalah 11 years ago |

Sensationalist bs. It'd take some pretty impressive programming to deal with issues like road construction, missing or faded road markings, unmarked dirt roads, and potentially outdated or incorrect maps.

sesutton 11 years ago | |

I don't think we'll go straight from total human control to total automation directly. I imagine that the first versions of automated cars will require some supervision and only work in certain conditions. This will likely require some sort of system to detect the driver is paying attention, like steering wheel sensors, to prevent people from just going to sleep.

lojack 11 years ago | | |

The first versions of automated cars already exist. Mercedes, BMW, and Audi have versions of their cars that maintain a safe distance from the cars in front of you. Mercedes also has a system which will automatically steer the car to keep it in a lane for you. Prius and Lexus have cars that'll automatically parallel park themselves.

jiballer 11 years ago | |

I don't think those are programmed into the driving software as much as it has learned them through machine learning. Most of these systems have gone through the equivalent of millions of years of humans driving so I'd trust it over a human driver any day.

0xdeadbeefbabe 11 years ago | | |

Well these systems weren't ever human, so I'm not sure a million human years would help them as much as you hope.

a8da6b0c91d 11 years ago | | |

I don't think any of the systems really use machine learning. It's more classical stats and control theory.

Maybe the object recognition stuff uses some trained classifiers, but the higher level systems doing the driving to my knowledge do not.

lukasm 11 years ago | |

They're gonna cover 98% of cases. The rest can be picked up be remote operators.

throwaway43 11 years ago |

I remain extremely skeptical that 95 % will be good enough and its the edge cases that really matter. It doesn't matter that they're better most of the time because the failure modes are different.

However since Elon is a really smart guy who has mastered multiple disciplines , I'll take his word. I'm waiting for my self driving car.

ecobiker 11 years ago |

The way I see it - it's hard to "transfer" all the learnings we get from our everyday driving experiences to other fellow drivers but a machine can do that. It's going to be a sum total of all of its learnings and its neighbours (all over the world). I like the idea of it.

rplst8 11 years ago |

He's right. Driving a car is something like 1000 times riskier than commercial aviation. We still have about 30,000 vehicle fatalities per year.

refurb 11 years ago | |

1000 times more risky as measure by what? People drive a car far more often than fly in a plane. I'm sure flying is still safer, but the comparison should take into account the differences.

rplst8 11 years ago | | |

The comparisons I've seen are usually on a person-mile basis. Commercial aviation can move more people a further distance with fewer fatalities.

drzaiusapelord 11 years ago |

Musk is pretty self-serving here. He writes about how advanced AI will "destroy humanity" to keep interest in manned space and to discredit robot and science missions. Now he's cheerleading automotive AI because Tesla wants to sell it.

I really am starting to see him as an empty suit. He just does PR for his companies' current projects. That's fine, but he's not a visionary. He's salesman. I can't wait for his celebrity to wind down. Its more than a bit over the top now. Tesla seems like a failure to launch an affordable electric or an electric with realistic range at a sane pricepoint. SpaceX is doing well, but LEO theatrics are boring considering the SLS is geared to take us way past that.

I wish he'd buy into causes that don't personally enrich him once in a while. I'd love to see someone of his caliber stump for a radio telescope on the far side of the moon or on replacing cars with more efficient tram systems powered by electric wires instead of the unsolved and perhaps unsolvable problem of batteries that can compete with the convienance of gasoline.

tokenadult 11 years ago |

Here are some facts and figures about driving versus flying for travel safety per passenger mile over a time period carefully measured by official statistics. "In absolute numbers, driving is more dangerous, with more than 5 million accidents compared to 20 accidents in flying. A more direct comparison per 100 million miles pits driving's 1.27 fatalities and 80 injuries against flying's lack of deaths and almost no injuries, which again shows air travel to be safer."[1] I thought everybody knew that airline travel is by far the safest form of travel.[2]

The basic issue about self-driving cars is that setting the standard of performance at "better than most human drivers most of the time" is a moderately low technical bar to clear, but would still result in a HUGE reduction of deaths, injuries, and damage to vehicles and roadside property. But I'm still not sure how soon self-driving cars will be good enough to drive on snowy roads (Google's self-driving cars have never been tested in snowy conditions, as of the last time I checked) or even in rain.[3] The story posted today is more nuanced than a story posted yesterday about Elon Musk's predictions, and reports, "He said he believed that we're still 20 years out from the roads being full of autonomous cars."

[1] http://traveltips.usatoday.com/air-travel-safer-car-travel-1...

[2] http://www.businessinsider.com/flying-is-still-the-safest-wa...

[3] http://spectrum.ieee.org/cars-that-think/transportation/adva...

lasermike026 11 years ago |

What about public transportation? We can make improvements to public transportation today.

mudil 11 years ago |

Actually, human pilots are too dangerous as well. Analysis of airplane disasters consistently shows that human errors are responsible for majority of crashes.

I would feel much more comfortable flying in a completely automated airplane.

mixmastamyk 11 years ago |

I'm glad they are working on this in general, however unfortunately the Model S is not able to slow or stop when encountering a car stopped in front of it.

I was quite disappointed when I found out they don't have this important technology, in a car packed to the hilt with it. http://my.teslamotors.com/forum/forums/collision-avoidance

panglott 11 years ago |

No question, if robot cars will obey speed limits and not blow through red lights. The biggest safety improvements will come from convincing motorists/passengers that it's OK to travel just a bit slower.