Slashdot Mirror
Tesla and Autopilot Supplier Mobileye Split Up After Fatal Crash (usatoday.com)
An anonymous reader quotes a report from USA Today: Tesla and Mobileye, one of the top suppliers to its Autopilot partial self-driving system, are parting ways in the wake of the May accident that killed an owner of one of its electric Model S sedans. Mobileye is considered a leader in developing the equipment that will be needed for fully self-driving cars. The Israeli tech company will continue to support and maintain current Tesla products, including upgrades that should help the Autopilot system with crash avoidance and to better allow the car to steer itself, said Chairman Amnon Shashua in releasing the company's second-quarter earnings Tuesday. Shashua said moving cars to higher levels of self-driving capability "is a paradigm shift both in terms of function complexity and the need to ensure an extremely high level of safety." He added there is "much at stake" in terms of Mobileye's reputation, and that it is best to end the relationship with Tesla by the end of the year. Tesla CEO Elon Musk, meeting with reporters at the company's new battery Gigafactory outside Reno, indicated that Tesla can go forward without Mobileye. "Us parting ways was somewhat inevitable. There's nothing unexpected here from our standpoint," Musk said. "We're committed to autonomy. They'll go their way, and we'll go ours."
Shocked! I say I'm shocked that a boutique car made by a startup company directed by an eccentric tycoon could possibly have had a supplier or design change. Unfathomable!
First world problems mate.
Impressed but not surprised Mobileye would ditch Tesla. If one of my customers was using my shit recklessly on public beta experiments that got people killed I would ditch them too.
I bought a car recently.
It has 'frontal collision avoidance' and 'speed limit reader' built in. It does that with a bit of lidar and a bit of camera work. Actually pretty cool CS wise.
The car companies have a real issue on their hands. Sun and temperature blindness. Twice already my car has freaked out and went into an error mode. The temp on the dashboard was easily over 120f (~50c) (the camera is behind the driver mirror). The camera freaked out and just stopped working. All of the other cameras that the car would let me see out of were smeary and 'grey'. You could tell the temp had basically made the sensor over sensitive. So the second I got a couple of reflections off a truck in front of me the system stopped working. Now some of you may be thinking 'oh just take it in'. Already done. 'common problem in the summer'. Bit or research on the internet? Same thing. Parked the thing in the garage and everything was fine again. Now these things are basically 'toys' in my car. But if they were important for real day to day usage I would be a bit more mad.
I am excited for the upcoming self driving car revolution. But it will be awhile. I would not buy one and rely on it too much today. They will need to put some very high end camera parts in these things. I am not seeing that today with the ones I have seen. Most are in the 720 and lower range with poor color contrast, and poor refresh. Little better than a cell phone camera from 10 years ago.
This sounds familiar. Oh yeah... Ford and Firestone parting ways because they were blaming each other for Explorer rollovers. It's funny how the rollovers continued after they put everything else but Firestones on them.
Two musky stories in a row? Some PR firm must be getting some big dollars.
Your hair look like poop, Bob! - Wanker.
I'm not at all convinced that anything will change. Car companies have to be forced by the government to spend $5 on seat belts, they're always going to be looking for the absolute minimum cost to produce economy/midrange vehicles and full automation will never fit into that.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
"is best to end the relationship with Tesla by the end of the year"
it's extremely rare that a parts supplier stops being a parts supplier by their decision. only reasons could be the customer(tesla) not paying it's bills or using them out of spec in a manner that might bite them later, and it basically says just that on the inv. call. so tesla and the sensor manufacturer were probably at odds over them shipping the autopilot sw. the sensor manufacturer knows that it can't be good enough with the sensors they sold tesla.
the mobileye system is mainly for collision avoidance, it's just an extra safety feature kind of a thing. thats how other car manufacturers are selling the same thing.
but I can imagine Musk looking at the spec sheet for their product and just going "AUTOPILOT!!!! WHEEE!!" after seeing that mobileyes off the shelf system has lane departure, headway monitoring/warning and forward collision warning. because hey, all you need for autopilot(tm)(r) is to hook up lane departure software output to the wheel and you're DONE! right??
autopilot, a name chosen to make the illusion that it can drive itself. they can't even trademark it, it's only purpose as a name is to deceive.
Most people buy a Tesla because it's electric and fast, NOT because it has a bot.
Table-ized A.I.
Two humans and one computer failed to avoid the collision. One human was at fault for failing to yield to oncoming traffic. Fatal accidents involving left hand turns are common.
About 100 people die every day on US roads. Beyond regretting that yet another person failed to yield while turning left in front of traffic and it resulted in a death what, exactly are we supposed to care about regarding that particular accident?
I don't know how many lives would be saved with autonomous vehicles, I only know that about 30,000 deaths a year on the roads are caused by human errors. Far more accidents involving serious injuries and billions of dollars in damage also occur each year due to human errors.
To quote Elon Musk:
“This was expected and will not have any material effect on our plans. MobilEye’s ability to evolve its technology is unfortunately negatively affected by having to support hundreds of models from legacy auto companies, resulting in a very high engineering drag coefficient. Tesla is laser-focused on achieving full self-driving capability on one integrated platform with an order of magnitude greater safety than the average manually driven car.”
This sounds quite reasonable to me. Tesla wants to go faster than anyone else in autopilot. Mobileye starts selling its chips to many car-makers. Mobileye is unwilling to make a special chip only for Tesla. Tesla then decides to come up with their own solution, using their in house chip expertise as well as possibly other companies' products (Nvidia perhaps?). This post is a subtle troll on Tesla.
This and no other is the root from which a tyrant springs; when first he appears as a protector - Plato (423 to 327 BC)
LIDAR is likely not the best tool for self-driving cars. It has problems in snow and rain. Radar and video are likely better, especially radar that makes a sparse point cloud, as Tesla's is going to do in the next update.
This and no other is the root from which a tyrant springs; when first he appears as a protector - Plato (423 to 327 BC)
That's another thing. The speeding. Autopilot breaks the law and LOGS IT.
Laws are rules for the court, but merely a bottom bar to hit for life. Think beyond laws in your actions always.
The autopilot did not break any law. The driver broke a law. If we can't get simple things like that correct, there is no way to have a logical conversation or debate.
Despite initially feeling the same as you about the Autopilot name, I've been cooling rapidly on it for one simple reason: in practice an aviation autopilot can handle pretty much anything that it is likely to encounter well enough for a distracted pilot to take many seconds assessing any crisis situation before having to take control. And that's key, because if the pilot's input is not needed, then human nature dictates that they're likely to be distracted when a situation that *does* need their input arises.
Tesla's Autopilot is not yet anywhere near that competent, not because it's technical competency is lacking in comparison, quite the opposite in fact, but because its expected operating environment is far more crowded and chaotic, and most crises will unfold far more rapidly, having already reached a conclusion before an inevitably distracted driver can hope to assess the situation. As such, Autopilot will need to be FAR more competent than it currently is just to be able to offer the same level of real-world functionality and safety as its relatively crude aviation counterpart.
I'd say it's currently got 70-80% of the needed functionality worked out, which means, as any programmer can attest, that only 90% or so of the work remains to be done.
--- Most topics have many sides worth arguing, allow me to take one opposite you.
The driver sets the maximum speed when they activate the autopilot, in much the same way as you set the speed when you use cruise control on any other car. Or are you saying speeding isn't the responsibility of any driver if they're using cruise control to break the limit?
Autopilot will slow down if there is traffic ahead, otherwise it travels at the speed set by the driver.
-- Pete.
Monochrome - Probably the UK's largest internet BBS
I think the name "autopilot" starts to suffer the same fate as the name "hacker": they both have a precise meaning, but in the general use by the public, the meaning has shifted
Hacker used to be someone who is good at McGuyvering, at finding creative uses, etc.
But the press ended up using it for Cracker, someone who just breaks into things, not necessarily showing any creativity.
Same happenned with Autopilot: in aviation, it is a very precise thing - an apparatus which can take care automatically of the small minute details of flying the plane. The human need to provide it an order (a destination) and then only watch over it and control that everything is going well, but not actually hold the commands themselves.
Nobody has ever deigned this for the whole crew to take a nap while it is on.
Same in a boat: the autopilot will keep a destination, so you don't need to hold the wheel. That doesn't mean that you should be napping, you still need to whatch out for dangers, obstacles, etc.
But suddenly, the general public has taken a different meaning: as you say, now the think of it as Chauffeur: the Chauffeur (not necessarily electronic, it can be a human) takes care of everything, while you can safely take a nap or whatch some harry potter.
Elon should have called it "Ship's Commander mode" (as the one which gives orders instead of holding the wheel) sound both mor awesome and a little bit less passive role for the driver.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
Designing a chip like the EyeQ v3 chip is a very large multi-year project. Tesla has hired a couple of designers, but they can't possibly be designing their own chip to replace the EyeQ. It would be akin to Dell hiring a couple of s/w engineers and saying they are writing a replacement for WIndows. I expect that in reality, Tesla are trying to integrate some of their existing s/w with other chips and new board design. From what I understand, they have disagreed about futures, and on realistic vs unrealistic expectations from existing system. For public info and speculation see: https://www.technologyreview.c...
Lasers, obviously!
Remember George Hotz?
http://www.theverge.com/2016/6...
He developed some self-driving technology and Elon offered him a job with a bonus if they developed technology independent of MobileEye. Elon has wanted to part ways with them for a while.
The accident gives him the excuse he needs.
Elon likes to do as much as possible in-house. You see that in both Tesla and SpaceX.
Its OK. Don't worry about me. I'll just give it to my kid and get the new one when it comes out.
Great, maybe he'll kill himself abusing autopilot.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Gee really nice, asshole. You want my kid to die?
No, I want you to die before making him, but causality prohibits it.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Really, many cars today "know" the speed limit. A friend's nissan versa tells her when the car is exceeding the limit. So shouldn't auto-pilot obey the limit automatically, which even econoboxes can somehow figure out.
Ok, let's talk about the purpose and practical capability of a plane's autopilot:
Autopilot is there to reduce pilot workload, it is not there to automate the whole process. It does not take a plane gate-to-gate. Safety organizations worry that autopilot makes pilots loose skill and reaction time. Autopilot in the Tesla behaved much like several other crashes attributed to pilots mis-using or over-relying on the autopilot.
The thing that is different with Tesla's system is that they do have ambitions of it taking you door to door eventually.
Ownership won't become obsolete until automated car companies can have a car in your driveway five minutes before you open your front door. If people have to wait for a car they'd rather use their own.
I'd "rather" not have to drive a car at all, and call for a chauffeur every morning to drive me to work. The reason millions of people don't have their own personal chauffeurs is the same exact reason car ownership will become obsolete and fairly quickly. You won't be able to afford it.
When autonomy is proven to be orders of magnitude safer than having a meatsack behind the wheel, insurance rates will adjust accordingly. Sure, you can choose to spend $2,000/month on human-powered car insurance for the privilege of still controlling your own vehicle, but who would want to do that when the next-gen autonomous ride service is ten times cheaper? Cost is king today. That is why you see Uber/Lyft fans wait for a car while the cab sitting right next to them is ignored.
And yet, despite all that, in normal clear-sky conditions a distracted pilot that simply sits in the chair and scans the environment every few minutes will be able to utilize it safely. Not because the autopilot is particularly sophisticated, but because the operating environment is extremely forgiving. Away from high-traffic airports airspace, any obstacles will be clearly visible for at many minutes before action is required.
Yes, Tesla's "Autopilot" is leaps and bounds more sophisticated, but it is unable to deliver the same level of reliability simply because the difference in operating environment complexity and urgency is far greater than the increase in sophistication.
--- Most topics have many sides worth arguing, allow me to take one opposite you.