The huge problem with the old auto industry is that they profit from making cars that are easy to crass. Every crash means expensive repairs and possible a new car so they profit big time from crashes. They are interesting in making their cars better at protecting their occupants when the car crash (because that is a selling point) but they do not want to prevent the crashes from happening in the first instance as that is bad for business. This is a strong reason why the old auto industry now seams to be covertly campaigning against what Tesla is doing with their autopilot and that significantly reduces the number of accidents per mile driven as measured by airbag deployments because not yet enough death events to make any scientifically based conclusions based on that definition of accident.

Tesla’s auto pilot is on a fast track to eliminate all types of crashes to trivial levels when it is fully developed. It can be improved in terms of the percentage of miles it can handle autonomously for any car and in terms of how much it can reduce the probability of accidents on those roads compared to an average human driver. The goal is to cover 100% of all roads and reduce the probability of accidents by at least 90%. The first goal could be achieved in 2 to 3 years whereas the latter goal may take a decade or two.

Autopilot should be mandatory on all new cars and it can’t go fast enough. It will not come automatically because of the economic interest in keeping a high rate of expensive crashes so we need legislation that requires it. I also think it is in the best interest of the public that all automakers are required to publish their detailed accident statistics so that the public can investigate independently these data for to reasons. 1) So that the industry can use these data to make better auto pilots that will prevent similar accidents and 2) to make sure that those autopilots that are used are in fact reducing the number of accidents per mile and are not increasing them.

@mahonj thank you for that idea blogged elsewhere about fully disclosing accident data. Retrospectively it is obvious it should be required.

Not true Bernard. Tesla's data proves that their auto pilot reduces the number of crashes already in a lot of places where it can be deployed. As they improve it and many more miles are documented thing will only become even more convincing for making auto pilots compulsory for all new cars sold ASAP.

I think the real problem here is driver education/training. Doesn't matter how "safe" we make vehicles, someone will always find a way to crash them. Think "make something idiot proof, and they'll build a better idiot".

While new technology makes newer cars harder to crash and safer when they do, as Tesla videos prove it also makes drives more complacent. They are coming up with newer ways to crash and sometimes cause much worse crashes.

Haven driven in several countries and various states in the US I can say that road courtesy and etiquette is worse in US. One example - down south you have be on the look out for people just randomly stopping in the middle of the road because "they are being nice to let someone out" or "stopping for pedestrian strolling across 7 lanes of busy parkway". I don't advocate hitting pedestrian, but stopping randomly is going to cause an accident.

I think the biggest improvement in highway safety, in this country, will be an overhaul of the driver education and licensing procedure. And I guaranty everyone will have to be dragged, kicking and screaming, into the new procedure.

TRD you are stating completely untrue things about what Tesla has said. The info I have about the issue you can read at Tesla’s website link given below. You say Tesla does not compare accidents when autopilot is on with accidents when autopilot is off. However, this is exactly what Tesla has done let me quote from Tesla’s webpage “…the data is unequivocal that Autopilot reduces driver workload and results in a statistically significant improvement in safety when compared to purely manual driving.” So all your rambling about Tesla drives can’t be compared to non-Tesla drivers is irrelevant because the comparison is between Tesla drivers using autopilot and the same Tesla drivers not using autopilot.

Then you say that “even if Tesla compared Tesla drivers driving with or without autopilot then it would still not be a correct comparison because the conditions could be different when people drive on autopilot and when they do not.” Sure enough that could be true but you are guessing as you obviously do not have a clue about what data Tesla is comparing. Also Tesla could have controlled for different driving conditions when concluding that their autopilot is safer. This is pretty basic knowledge for a statistical expert and off cause Tesla has employed competent statistical analysts to mine their data for useful information and extract that information in a scientifically correct manner.

Tesla already has enough data to prove that using autopilot is safer than not using it and that data is getting better and more certain for every mile they log in their database. Tesla said they already logged 130 million miles on autopilot and I believe that is more miles than any other automaker that develops auto pilot systems. I don’t know for sure as the old auto industry is very secretive about their data on autopilot. Only Google is open on this issue as Tesla is and Google has logged less than 2 million miles on autopilot.

https://www.teslamotors.com/blog/tragic-loss

ECI I have a lot of respect for you being one of the brainy bloggers at this forum and also because you drive a Tesla not out of vanity I believe but because green cars sincerely interest you and you actually care about not helping to destroy our common planet. And I don’t mind your critique. On the contrary we all need critique to sharpen our thoughts and simply to learn faster.

If you could give some links to the info you talk about then I could check it out. I expect the data gathered by the AP programs of the old automakers to be utterly uninteresting as they only have a few cars logging a few hundred thousands miles. Google at least has logged more than a million miles and Tesla has logged over 130 million. To make a scientifically meaningful death statistics on AP we need minimum 5 billion miles or about 50 deaths. Tesla is the only auto maker that actually has enough data to start making good conclusions about airbag deployments.

In the current environment where a lot of people who have shorted Tesla stock are digging aggressively for dirt about Tesla whether relevant or not or true or not I would advise Tesla not to publish any more than they have to according to the law as it will just be used unfairly against them. Also Tesla cannot defend themselves by comparing with the same data for their competitors as they are not yet having this kind of data in any meaningful quantity.

Make some legislation that requires all new vehicles to have autopilot by 2019 and require the car makers to log all accident data and make it publicly available with monthly updates. That would quickly save thousands or hundred of thousands of lives globally and it would speed up the time it takes to develop really safe and fully autonomous cars.

ECI I do not understand your animosity against AP tech. I suspect the Model S you drive is one without AP. However, AP is the most interesting thing that is happening in the global auto-industry since the invention of the combustion engine and fully autonomous AP systems are critical for making BEVs mass market affordable so I really hope you will get a Tesla with AP soon so that you can get first hand experience from this tech.

It is hard to take impaired drivers away from the driving wheel when there are no viable transportation options. Without viable public transportation (most of the US), how does an individual get to work, home, or the grocery store that is 3 miles away? Europe has an advantage in that respect.

Henrik, thank you, I enjoy the productive dialogue.

> ECI I do not understand your animosity against AP tech.

I assure you I have no animosity against AP whatsoever. Just a very healthy skepticism about an unwarranted belief that it is as robust as many people seem to imagine.

I talk with AP engineers routinely at conferences. I read the white papers (I am a software developer and former automotive technician with a great interest in the tech).

I've driven cars with assist technology from Volvo, GM, Tesla and others. Much good progress, real benefits available.

I now routinely talk to Tesla AP drivers about their use of and attitudes about the systems. There is a huge knowledge gap.

I agree that it is a very exciting development in the automotive field. One of the most significant in 110+ year history.

I am simply stongly cautioning against unjustified optimism about the rate of progress and over-reliance on systems that have no capability to handle situations you are likely to eventually encounter on the road.

ECI I guess we are more on the same page with AP than I thought we were. We consider it equally important and interesting but differences are that I believe it will come sooner than you think it will.
I think the first commercial driverless taxi services will start no later than 2020 and all cars sold to fleet or private in the developed world will be fully autonomous no later than 2030. At that time I also expect 50% of all known car companies of today to have bankrupted because they did not change fast enough towards driverless cars.
We may also differ in our attitudes regarding how fast we should push AP tech into real world cars. I think it can’t go fast enough. Let the driver be told in crystal clear language that they are responsible at all time and should be ready to take control instantly when the AP is active and drive into a situation it is unable to handle. When the fully autonomous AP is done and it can be statistically proven that it can drive more safely than the average human driver it will be ready for commercial and fully driverless taxis with no steering wheel or speeder or brakes. I think Tesla will make and deploy such taxies by 2020 at the latest.

The Tesla autopilot incidents are under discussion in another engineering forum that I am in. It seems that the system is only capable of operating properly under a fairly narrow set of conditions, and anything outside of those can cause it to fail. It does not do cross traffic, it does not do stop signs and traffic signals, etc. It does not do something that all good human drivers do ... scan their environment for potential threats, predict those threats, and take preventive action against the threat materializing. (It can be as simple as noting that there is a vehicle in an adjacent lane which is ending, predicting that this vehicle will do a lane change, and positioning oneself such that the other vehicle can do this action without conflict. Good drivers do this ... many drivers do not ... Autopilot does not)

"Driver aids" - systems that bail drivers out when they screw up - are already here and will develop further.

But Sweden and elsewhere don't have any more "self-driving" than the USA does. If there's a difference in safety statistics, this isn't the underlying factor. It cannot be.

I can say what some of the differences are ... Better training of drivers, periodic re-testing of drivers, periodic safety inspections of vehicles (and these are inspections that actually have some teeth), better public transit, enforcement of traffic rules other than speeding (plenty of intersections have cameras), better lane discipline, drivers who know how to yield, roadway designs with roundabouts in place of traffic signals or stop signs (much harder to have a head-on or left-turn-into-oncoming-traffic), mandatory seat belts for car drivers, mandatory helmets for motorcyclists, and in some cases, roadways designed with consideration for what happens when a vehicle leaves the roadway.

Henrik > Tesla will be ready to deploy such a fleet of driverless taxies by 2020

This speculation is based on what evidence?

Based on billion Km driven, Sweden and UK have the lowest crash fatality rate and USA has the highest.

The crash fatality rate by billion Km driven in USA (17.1) is about 2.75 times the Canadian rate.

The Canadian rate varies from a low 4.5 in Ontario to a high 48.3 in NVT. The average is about 6.2.

Impaired drivers (drinks and drugs or both), texting and high speed are the main reasons.

I would beg to differ on what the prime factors of the US traffic fatality rates are.

Drunk/impaired driving, yes, I agree.
Texting or otherwise distracted, that as well.
High speed, not so much. Most of Europe has a motorway speed limit of 130 km/h give or take. Ontario (Canada) has the lowest such speed limit (100 km/h), but it's widely ignored. The actual travel speed on motorways is little different. (120 - 130 km/h, or 75 - 80 mph if you wish) And in any case, motorways are the safest roads, city streets the most dangerous. One factor Toronto shares with european large cities is that about half the traffic fatalities are pedestrians (many of whom have their noses buried in their phones nowadays).

The USA has a problem with people not wearing seat belts in cars and not wearing helmets on motorcycles. Canada and all of Europe require these. This is a rather large factor.

ECI you keep bitching about Tesla and their cars. I wonder why you say you drive one. I am beginning to doubt it and on whether you have ever driven a Tesla with AP activated. Tesla can already prove that using their auto pilot result in safer driving than not using it. All of your arguments are emotional and about one event. Give us a break. If you and others succeed in slowing the development of AP systems it will mean thousands of more dead people from traffic accidents that these AP systems would otherwise have saved. So my conclusion is that what you agitate is basically immoral because it will result in more death not less and it is also based on ignorance and emotions about spectacular events that are inevitable. Accidents happen and they will keep coming no matter how careful you are. Only large scale statistics can determine whether we are making progress or not.