Musk endeavours

[JamesMuenchen]

Interesting video showing the performance of motion eye & other systems compared to the uber system that was involved in the fatal crash (warning, shows the fatal crash):

https://youtu.be/QCCmqosHT-o

Regards,

In the footage of the fatal crash (around 9:24 in the vid), the victims dark jumper blends with the shadow on the road. I wonder if this could also be the cause of the problem for the sensors? To the eye at least, you just see a bicycle and legs. Obviously, I would say it should still stop rather drive into a bicycle, but just musing.

[odysseus2000]

TESLA motivational speech to employees:

Tesla Asks for Model 3 Factory Volunteers to Prove ‘Haters’ Wrong - Bloomberg
https://apple.news/ApqOjC_zRRiCAYedj6pZ-tg

Kind of interesting that management are using these statements to motivate model 3 production. CLEARLY it is no go see smoothly, but they do seem to be making more.

Regards,

[Itsallaguess]

Some details here regarding a fatal Tesla crash on 23rd March -

Electric carmaker Tesla says a vehicle involved in a fatal crash in California was in Autopilot mode, raising further questions about the safety of self-driving technology.

One of the company's Model X cars crashed into a roadside barrier and caught fire on 23 March.

Tesla says the 38-year-old driver, who died shortly afterwards, had activated Autopilot moments before the accident.

But they did not say whether the system had detected the concrete barrier.

http://www.bbc.co.uk/news/world-us-canada-43604440

Difficult to say, but judging from the photo in the above article, the concrete barrier looks to be of the 'lane-splitter' variety, and it seems to have been a real issue for the autopilot to judge correctly.

That seems to be corroborated in the following online article too, which discusses a very similar, earlier situation that a Tesla car had difficulty with -

Shaun Price, Director of Science for an environmental startup, drove us to the crash scene in his Tesla Model X.

"I mean, you have to think like a computer, right?," Price told Dan Noyes. "A computer doesn't know, it has no logic, so if it sees a line, it might think that's a lane."

And that raises so many issues. CalTrans has already started improving the infrastructure to accommodate automation, such as wider lane stripes in some places.

"We can imagine a day where all of our cars are in automated mode going down the road and they all fail at the same place on the freeway," Says Jim McPherson. "That raises the question at what point is CalTrans responsible to fix the road to accommodate automation."

http://abc7news.com/automotive/exclusiv ... e/3284757/

I shudder to think of the costs that might be involved to bring the majority of the major arterial routes in the UK up to a standard good enough for auto-pilot cars to travel large distances here.

Cheers,

Itsallaguess

[odysseus2000]

Yes, the whole autominous driving future is not holding up that well in practice.

Musk had said that a Tesla would drive itself from coast to coast across the US by the end of 2017, but it didn't happen and we have had the Uber crash where it is difficult to understand why the system didn't respond. Sure it might not have saved the pedestrian's life but unless I have misunderstood the timing there was a second or so before impact which to these systems is a lot of time.

Difficult also to understand how the system missed the concrete in latest Tesla fatality. The concrete wasn't moving and there seems no reason why it didn't see it but there is also the wonder as to why the driver put the auto pilot on just before the crash and whether there was some start up glitch that caused the accident.

There will be exhaustive research on all of these tragedies and this might throw up some understanding, but it begins to look like the systems currently deployed have flaws. It may be that such flaws can not be engineered out and one then has to make a decision based on the grim business of whether computers or people kill and injury the least number of people and develop the legal frame work for handling computer incidents. In my experience of using software to find signals in noise I always found that getting 100% was very hard. One could get close but there was often something that happened and even after many iterations there were still new features in the data that lead to erroneous false negative and false positives. The numbers I was working on were too many for a single person and one had to use computers and overall the algorithms were very good, but when its lives they have to be 100% reliable only wrong when circumstances are beyond any intervention.

The data suggests, for the moment, that we are some unacceptable way from that and that many of the statements on machine driving are hype. Would like to be proved wrong and might be, but for now I am leaning towards believing that the systems will need to be improved and that perhaps it may be impossible to prevent all machine mistakes putting the ball in the politicians court as to what they will allow.

Regards,

[BobbyD]

Basically the author is saying auto driving tech should be classed with nuclear, hospital tech, etc in that even one fatal mistake is one too many.

Presumably we apply the same test to human driven cars?

As a pedestrian I'm not that bothered about what flavour of car kills me, but if the situation could be avoided altogether I'd very much appreciate it, even if some other poor sod is still run over elsewhere and if the situations are reversed I'm sure he would think the same and I wouldn't begrudge him his survival.

Yes, the whole autominous driving future is not holding up that well in practice.

Musk had said that a Tesla would drive itself from coast to coast across the US by the end of 2017, but it didn't happen and we have had the Uber crash where it is difficult to understand why the system didn't respond.

Not for Tesla whose system is closer to an overdeveloped cruise control than an autonomous vehicle or UBER who have just settled a stolen IP claim from Waymo for $245 million... any reason other, arguably more reputable, players' systems should be considered suspect?

Coast to coast is so 2015, atleast that's when Delphi, now APTIV, did it with 99% autonomous control.

More conjecture on the UBER fatality:

- https://www.theregister.co.uk/2018/03/2 ... lind_spot/

They appear to have been trying to do it on the cheap, I thought the days of the single rotating LIDAR were restricted to the history books.

There will be exhaustive research on all of these tragedies and this might throw up some understanding, but it begins to look like the systems currently deployed have flaws. It may be that such flaws can not be engineered out and one then has to make a decision based on the grim business of whether computers or people kill and injury the least number of people and develop the legal frame work for handling computer incidents.

Grim how? 1.3 million people die on the roads every year. The decision to introduce technology which reduces that isn't grim, it would be a substantial achievement. And that's just accounting for those killed directly by automobiles. Autonomous cars don't have to be anywhere close to perfect to be a substantial improvement on human drivers.

[dspp]

A somewhat thoughtful piece on Tesla

https://www.nakedcapitalism.com/2018/04 ... italism%29

at this rate they'll be saying that someone with a red flag needs to walk in front

good luck to Tesla.

regards, dspp

[odysseus2000]

The promise of robot driving is that it could become a mass market technology available at relatively low cost that will substantially reduce road deaths. One goal is that the insurance policy for a robot car would be dramatically less than for a human driver and that public transport would move from the current use of many people devices (bus, train etc) only going to specified points to a one or a few person device going door to door.

To achieve these goals the systems have to be made at low cost and be reliable. Since humans rely predominantly on sight for driving there have been projects that predominantly use optical systems to replace the human driver. Such systems if reliable offer likely the lowest cost robot vehicles.

Other technologies using e.g. multiple radars are likely more expensive. If they work substantially better then they would have to be used but the engineering effort is always as Brunel noted, to do for six pence what any fool can do for a schilling.

Then there is the public/political aspect of driving. If the media start up campaigns about how unsafe robot drivers are and they are likely to do as “killer robots” make for many eye balls one can easily get a public backlash against robots.

Meanwhile the politicians have the calculus of death to weigh. If robots are much safer than humans then they would want to approve them, unless their is enough of a public hate that approving them may cost votes.

It is easy to post on a bulletin board about how much better robots are, but the grim realities of letting loose robots and the deaths caused by robot mistakes to be considered with the danger of the media rallying voters to rebel.

Sure one can argue that robots are much better than people and in the end that might be a convincing argument, but currently we have a public acceptance of road deaths and folk are notoriously reluctant to change and suspicious of system they do not understand. Compared to the number of deaths from human error the losses from robot drivers are very small, but these few are what make the headlines and which have huge impacts on Uber, Tesla and Nvidia and the other robot car players.

Regards,

[redsturgeon]

Has Elon Musk just had his "Ratner' moment by joking about Tesla going bust on Twitter?

Elon Musk
✔
@elonmusk
Tesla Goes Bankrupt
Palo Alto, California, April 1, 2018 -- Despite intense efforts to raise money, including a last-ditch mass sale of Easter Eggs, we are sad to report that Tesla has gone completely and totally bankrupt. So bankrupt, you can't believe it.

11:02 PM - Apr 1, 2018

Seem like his shareholders' were not amused with the price down 5%.

John

[odysseus2000]

@elonmusk
Tesla Goes Bankrupt
Palo Alto, California, April 1, 2018 -- Despite intense efforts to raise money, including a last-ditch mass sale of Easter Eggs, we are sad to report that Tesla has gone completely and totally bankrupt. So bankrupt, you can't believe it.

11:02 PM - Apr 1, 2018

Seem like his shareholders' were not amused with the price down 5%.

John

It will be interesting to see what the new deliveries are when they come out sometime this week.

Sentiment is now so bad that it will take really bad numbers to push the price down more, but could happen as could a rally if numbers are better than most forecasts.

But I haven't played this latest sell off well so best not to take notice of anything I type!

Regards,

[odysseus2000]

Tesla's view of the safety of automatous cars:

https://www.tesla.com/en_GB/blog/update-last-week’s-accident

Over a year ago, our first iteration of Autopilot was found by the U.S. government to reduce crash rates by as much as 40%. Internal data confirms that recent updates to Autopilot have improved system reliability.
In the US, there is one automotive fatality every 86 million miles across all vehicles from all manufacturers. For Tesla, there is one fatality, including known pedestrian fatalities, every 320 million miles in vehicles equipped with Autopilot hardware. If you are driving a Tesla equipped with Autopilot hardware, you are 3.7 times less likely to be involved in a fatal accident.
Tesla Autopilot does not prevent all accidents – such a standard would be impossible – but it makes them much less likely to occur. It unequivocally makes the world safer for the vehicle occupants, pedestrians and cyclists.
No one knows about the accidents that didn’t happen, only the ones that did. The consequences of the public not using Autopilot, because of an inaccurate belief that it is less safe, would be extremely severe. There are about 1.25 million automotive deaths worldwide. If the current safety level of a Tesla vehicle were to be applied, it would mean about 900,000 lives saved per year. We expect the safety level of autonomous cars to be 10 times safer than non-autonomous cars.

Tesla comment on the latest crash:

In the moments before the collision, which occurred at 9:27 a.m. on Friday, March 23rd, Autopilot was engaged with the adaptive cruise control follow-distance set to minimum. The driver had received several visual and one audible hands-on warning earlier in the drive and the driver’s hands were not detected on the wheel for six seconds prior to the collision. The driver had about five seconds and 150 meters of unobstructed view of the concrete divider with the crushed crash attenuator, but the vehicle logs show that no action was taken.

[Itsallaguess]

Tesla comment on the latest crash:

In the moments before the collision, which occurred at 9:27 a.m. on Friday, March 23rd, Autopilot was engaged with the adaptive cruise control follow-distance set to minimum.

The driver had received several visual and one audible hands-on warning earlier in the drive and the driver’s hands were not detected on the wheel for six seconds prior to the collision.

The driver had about five seconds and 150 meters of unobstructed view of the concrete divider with the crushed crash attenuator, but the vehicle logs show that no action was taken.

I still think that no matter how good an autopilot might get for domestic cars, it'll be our lack of trust of the people inside that'll actually stop the wide-scale roll-out of the technology onto our roads.......

Cheers,

Itsallaguess

[odysseus2000]

Itsallaguess
I still think that no matter how good an autopilot might get for domestic cars, it'll be our lack of trust of the people inside that'll actually stop the wide-scale roll-out of the technology onto our roads.......

This is possible and it is one the things that Tesla address in the excerpt from their web site I posted.

As I look at it there will be two important factors.

1 Can this technology be made cheap enough and reliable enough that most manufacturers will fit it. We currently have several fitting the optical system based on the Intel kit that they got when they bought the mobileye company which does not seem to have such a high cost that punters baulk, seemingly the reverse: They want it.

2 How does the government and insurance industry view the technology? If either or both of these decide that robot driven cars are worth having, they can force change by simply cranking up non-robot insurances to levels where folk will be reluctant to pay.

Dunno. My experience with computers makes me very nervous about having one drive me around, but looking at the accident statistics the case for robot driving is overwhelming. If we see many more deaths that can be blamed on poor robot performance as looks to be the case with the uber crash, then I expect the whole thing to be set back, but if not then it will likely happen very quickly as part of the transition to electric traction.

Regards,

[JamesMuenchen]

The driver had received several visual and one audible hands-on warning earlier in the drive and the driver’s hands were not detected on the wheel for six seconds prior to the collision.

Shouldn't the system pull over and stop in that case?

[BobbyD]

The wisdom of seeking to work up.through the automation gears starting with level.1 add one and evolving through to a level 5 system on road, whilst advertising future performance based on fitted hardware is questionable and risks poisoning the well for everyone in the sector. At the very least attention needs to be paid to the significant psychological effects which partial automation has, and ways to maximise the effectiveness of safety drivers during on road testing.

The point about more expensive systems using 4, 5, or 6 lidars is not just are they likely to kill less people, but they are not initially being targeted directly at consumers. Spending that much money on a car doesn't make sense for most individuals but for a car hire firm which can have it on the road for 18 hours a day it does. Once the technology is established and the volumes bring prices down then maybe it makes.the jump to the consumer market.

One of the reasons autonomous vehicles can be safer than human drivers is.because they are not limited to visual input of a small arc, but can be informed by all.sorts of sensors covering the.complete 360 degree surrounds of.the car. Human drivers aren't actually very good, and they make a strange model on which to Base non-human drivers.

[odysseus2000]

BobbyD

The wisdom of seeking to work up.through the automation gears starting with level.1 add one and evolving through to a level 5 system on road, whilst advertising future performance based on fitted hardware is questionable and risks poisoning the well for everyone in the sector. At the very least attention needs to be paid to the significant psychological effects which partial automation has, and ways to maximise the effectiveness of safety drivers during on road testing.

The point about more expensive systems using 4, 5, or 6 lidars is not just are they likely to kill less people, but they are not initially being targeted directly at consumers. Spending that much money on a car doesn't make sense for most individuals but for a car hire firm which can have it on the road for 18 hours a day it does. Once the technology is established and the volumes bring prices down then maybe it makes.the jump to the consumer market.

One of the reasons autonomous vehicles can be safer than human drivers is.because they are not limited to visual input of a small arc, but can be informed by all.sorts of sensors covering the.complete 360 degree surrounds of.the car. Human drivers aren't actually very good, and they make a strange model on which to Base non-human drivers.

Yes, this is exactly the thinking I was taught while doing my PhD and working in post-doctoral research.

It is a logical and plausible way of proceeding, but if you take this mind set into business it will get you killed.
.
In business one is always worried about cash and anything that can be done to bring in cash is done. Hence the modus operandi is always about getting the most simple and basic thing to punters and selling it at a good margin to fund the next version.

Many with university training fail in business because they are used to the academic environment which looks at all aspects of every thing and this provides work for lectures, graduate students and post docs who are so busy exploring all avenues that they lose complete sight of the concept that they are there not to make things better but to make money for the business.

There are endless conflicts between academic trained and entrepreneur mind sets because of these two divergent ways of thinking.

If you look at some of the big UK business disasters you will often see that the academic mind set had got too much power and ran the company into the rocks.

Regards,

[odysseus2000]

dspp
A somewhat thoughtful piece on Tesla

https://www.nakedcapitalism.com/2018/04 ... italism%29

at this rate they'll be saying that someone with a red flag needs to walk in front

This is a very useful article showing that the author is skilled in his craft, but at the same time misguided.

He argues that robots are not as good at putting fasteners in, welding etc than people, stating that this is evidenced in the build quality issues of Tesla cars. If the author had spent some time looking at human assembled crimps, welds, rivets and had driven a car pre-robot assembly he would have a different view. Humans are bad at all of these jobs which are boring, repetitive and have to be done very quickly on production lines. To argue that human stuff is so much better is indicative of someone who has never studied fabrication.

He then backs up his ideas by noting how all the big motor makers have tried automation and failed.

There is no guarantee that Tesla can make automation work, but that is not based on previous tries not working because the technology is now very different to what it was before, plus you have managers/engineers who have seen these failures and are attempting to make something better.

Ford in his tome (Life and times ?) railed against this kind of mindset, saying he would never keep the results of folk who have tried things and failed. Instead he would start with new people and let them go at it from scratch and this allowed him to make dramatic advances and lower production costs.

Sure the author claims to have Tesla internal folk saying this, but as with Henry Ford such folk are likely out of the door to be replaced by others told to get it done.

No guarantees that Tesla will make this work and failure probably kills the business, but if they get it right many good things follow.

Regards,

[dspp]

dspp
A somewhat thoughtful piece on Tesla

https://www.nakedcapitalism.com/2018/04 ... italism%29

at this rate they'll be saying that someone with a red flag needs to walk in front

This is a very useful article showing that the author is skilled in his craft, but at the same time misguided.

He argues that robots are not as good at putting fasteners in, welding etc than people, ........

No guarantees that Tesla will make this work and failure probably kills the business, but if they get it right many good things follow.

Regards,

02000,

The driving autopilot stuff is same as everyone else, that didn't interest me, yet it is what other posters picked out of the mix.

You are the first to pick out the assy automation which is what did interest me. What he is discussing is the same issue that the auto-industry is watching in fascination. If they succeed Tesla has a real chance as a game-changer. If they fail then, if nothing else, they have accelerated EVs by at least 5-years imho. But merely putting cars together is not quite as easy as it looks, and Musk himself has said as much. I wonder if that is the real reason for the uncluttered internal design styles, i.e. they have designed out all the fiddly bits.

regards, dspp

[BobbyD]

It is a logical and plausible way of proceeding, but if you take this mind set into business it will get you killed.

...where as cost cutting on unproven technology, and up selling non-existent technology thereby making your customers dangerously overconfident in your product is getting other people killed.

...and as to the jackpot they win if it all works out, ignoring the fact they've already had rights to trial withdrawn, halted their live trial scheme, been associated with a fatality and had the car manufacturer (volvo) and sensor system supplier (APTIV) distance themselves from Uber's working practices, how defendable do you think any autonomous gains Uber makes are likely to be? They are running their code on other people's equipment in other people's cars, and other people who have been developingbthe equipment for longerand are yet to kill anybody at that.

[Itsallaguess]

Nervous about computers driving cars?

How do you feel when you board an aircraft?

They are controlled by computers these days.

Of course they are - but when I step foot on an aircraft I don't think I'd feel like I was part of the live beta-test programme that currently seems to be under-way with large numbers of autonomous cars....

The fatal accident a few years back on a flight from Brazil to France was caused by the pilot not believing the computers.

This doesn't seem at all relevant to the current discussion then. Your example of pilot-error is in no way related to the non-driver-error issues that have recently killed people with regards to the Uber and Tesla incidents. In both cases, it seems that the autonomous controls of the cars were the prime causes of the deaths.

If it had been the aircraft autopilot that had caused the Brazil/France flight incident, it might have been relevant, but it wasn't; you've already said that the automatic systems were correct, and the pilot chose to ignore them.....

Cheers,

Itsallaguess

[odysseus2000]

BobbyD
...where as cost cutting on unproven technology, and up selling non-existent technology thereby making your customers dangerously overconfident in your product is getting other people killed.

...and as to the jackpot they win if it all works out, ignoring the fact they've already had rights to trial withdrawn, halted their live trial scheme, been associated with a fatality and had the car manufacturer (volvo) and sensor system supplier (APTIV) distance themselves from Uber's working practices, how defendable do you think any autonomous gains Uber makes are likely to be? They are running their code on other people's equipment in other people's cars, and other people who have been developingbthe equipment for longerand are yet to kill anybody at that.

Yes, I agree.

I am just saying how I view business.

Sure one feels there has to be a much better way, but in my experience business is about getting stuff to market and selling it.

If in this rush to market you make stuff that gives you a bad reputation by being of poor quality, or hurting folk your whole business can collapse or you can get shut down by the regulators.

So it boils down to an art of getting stuff to punters with enough quality that it doesn't back fire on you.

Folk who can do this repeatedly are few in number and they get very wealthy.

One can argue that Lord Sugar excelled at this in his early career and then got screwed by Western Digital and Seagate selling him hard drives that were faulty. He sued both of them, got a big chunk of money of Seagate and nothing out of Western Digital who had a very pro American company jury in Lord Sugar's view.

Regards,