Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Speculation: Model X will contain "full autopilot" IE self driving.

This site may earn commission on affiliate links.
Bef 2015: 40 signs in 25 countries
Beg of 2015: 250 signs in 50 countries
End of 2015: 1000 signs in 100 countries

Stoplight recognition in Model X? Some comments in the video below that indicate that possibility.

Certainly contains EyeQ3 like is in the Model S as of (~Nov'14) [EyeQ4 coming out in Oct'15)

Source: CTO of Mobileye in which Amnon Shashua talks about how Mobileye is incorporating deep learning into its self driving car research.
The Future of Computer Vision and Automated Driving by Prof. Amnon Shashua - YouTube

So here's what I don't understand: Tesla uses Mobileye for the hardware and presumably a lot of the software. Yet Tesla has been on a tear hiring autonomous driving engineers for at least a couple of years. Who is doing what? Is Tesla using the hardware but doing their own custom software? Only some of the hardware is Mobileye?
 
Apologies for posting this in two threads, but I didn't see this one until now.

From Car Safety IPO Mobileye Ties Future To Tesla, Others:

A U.S. carmaker will soon launch a semiautonomous, nearly hands-free-driving car for highway use, using Mobileye's camera technology, Aviram says.

The driver need only push a button and leave much of the highway driving to the car itself, he says. For speeds of up to [sic] around 40 miles per hour on the highway, it'll stay within a lane, keep the right distance from the vehicle in front and brake or adjust speed as it reads signs and traffic signals.

Aviram wouldn't disclose the name of the automaker behind the vehicle, but Shanker of Morgan Stanley said, "We believe it is going to be Tesla Motors (TSLA)."

He expects Tesla's semiautonomous car to come out toward the end of the year, by way of the Model X.

Probably meant to say 40+ mph.
 
I came across an interesting small piece today in the Swedish publication "Ny teknik" (New Technology). The title is misleading and a bit of click-bait though, since I don't really see any opposing views between Tesla and Volvo necessarily. Anyway, this article is mostly about Volvo's view but since they are one of the leaders in the field I found it interesting. I took a stab at translating it, since Google translate did a poor job with this one:

Tesla och Volvo oeniga om självkörande bilars moral - NyTeknik

My translation:

Tesla and Volvo disagree on the morals of self driving cars

Tesla's opinion is that the main obstacle with self driving cars are the rules governing the control systems, the ones that make decisions that can mean the difference between life and death. Volvo disagrees.

Peter Carlsson, chief logistics officer at Tesla, was asked about the company's work on self driving cars during the Sthlm Tech Fest conference last week.

- Of coursew we have a team working on it. All new cars would be able to drive without a driver thanks to all the sensors and cameras. But for them to actually do this, of course many other things are required, said Carlsson.

Erik Coelingh, technical specialist at Volvo, disagrees that decisions like these [this is poor writing by the journalist, since they reference something that hasn't been mentioned, except in the title I guess] are the hardest ones.

- Our strategy is for the car to never find itself in a situation where it needs to make these kinds of moral decisions. Ever.

- Because of this the self driving cars will always drive very defensively. They will always keep a following distance to the car in front of them longer than the expected braking distance.

According to Volvo the self driving cars will never go above 70 km/h.

- If, for example, a trailer has stopped along the side of the road making it impossible for the self driving car to see if there is a potential danger ahead, it should change lanes to get a better overview or slow down. It should never take a risk and try to pass in the same lane it's already in, says Coelingh.

But there must be situations that the seld driving car can't anticipate, and there could be technical problems with the car?

- Yes, but they are few. If for example a skydiver suddenly lands in front of the car. That would be an extremely uncommon event.

- In this case the cars should follow the traffic rules, stop in its lane and brake as hard as possible. It should not attempt to make moral decisions. The same goes in a situation with technical problems that couldn't have been anticipated.

- There will never be a table to decide whether the seld driving car should choose to hit a Fiat or another Volvo.

Neither if the car should choose to hit a child or a tractor?

- No the car should do anything in its powers to avoid those kinds of situation. If something extremely uncommon occurs it should still follow the traffic rules, stop in its lane and brake as hard as it can.

Eric Coelingh says that the important moral issue is whether society should keep accepting the 30 000 yearly traffic deaths in Europe, approximately the same number as in the US, when there is technology - seld driving cars - that can radically reduce those numbers.

He compares the issue to vaccinations, used to provent a great deal of human suffering dispite causing a few cases of illness.

- The same moral issue is present with self driving cars. A few new types of deaths could occur as a result of this technology, at the same time that it prevents a much larger number of deadly accidents.

How many of today's 30 000 yearly traffic deaths that could be prevented by seld driving cars is not something Coelingh wants to try to estimate.

- We just don't know. But we do know that almost all accidents are caused by human behaviour. So on paper the potential to reduce the number of accidents is big, but we need to confirm this in reality.

Volvo's triles with "Drive Me" self driving cars in Gothenburg start in 2017. Some time during 2018 or 2019 Coelingh believes he will have good real-life data. He has high expectations:

- If we were only able to reduce the number of traffic deaths from 30 000 to 3000 every year by introducing self driving cars, I'd be dissappointed.
 
I suspect that Tesla Model X has all the technology for the full autopilot but will release it slowly by firmware updates. This is possibly due to legalities in having the full self driving car. As many have noted, it is better to have full autopilot, only car driving experts can do better than the autopilot. Especially with the drivers being obsessed with their phones, the autopilot is less error prone. The auto insurance companies will speed up the autopilot by having lesser prices for cars with autopilot due to reduction in accidents.
 
According to Volvo the self driving cars will never go above 70 km/h.

Assuming that wasn't a transcription error where you actually meant 70 mph, I'd say don't look for Volvo to be around twenty years from now, since "never" is a long time, and 70 kph is really pretty slow. All the more so when you consider that the Volvo spokesman was evidently advocating for autonomous eventually being the sole permitted mode of operation (reading between the lines a little). If you put those two things together, the necessary conclusion is that all road traffic would be restricted to 70 kph. "Not gonna happen."

I don't follow autonomous vehicle research closely but I'm pretty sure others do not have such timid goals.
 
My thoughts are that humans are able to drive using only vision. Smart software should be able to get by with vision only too. With that being said, you'd need hardware capable of processing 4K+ resolution video at as high of a frame rate as possible. I don't think that will be possible for 5-10 years.

Seems to me your drive with human vision paradigm is an unnecessary constraint.
Eg when evaluating a stop sign, you could take all video inputs of all data sources -mobile and stationary- in all possible spectrums and sensor types (IR, LIDAR, weight, RFID etc), and all registered events of all history, for that and every relevant comparable location on the planet, simultaneously.

Machine learning is not at all like humans'.
 
I came across an interesting small piece today in the Swedish publication "Ny teknik" (New Technology). The title is misleading and a bit of click-bait though, since I don't really see any opposing views between Tesla and Volvo necessarily. Anyway, this article is mostly about Volvo's view but since they are one of the leaders in the field I found it interesting. I took a stab at translating it, since Google translate did a poor job with this one:

Tesla och Volvo oeniga om självkörande bilars moral - NyTeknik

My translation:

I agree completely with Volvo. I've been saying this for years. There should never be a need for a car to make a "moral decision." Thank you for translating.
 
Assuming that wasn't a transcription error where you actually meant 70 mph, I'd say don't look for Volvo to be around twenty years from now, since "never" is a long time, and 70 kph is really pretty slow. All the more so when you consider that the Volvo spokesman was evidently advocating for autonomous eventually being the sole permitted mode of operation (reading between the lines a little). If you put those two things together, the necessary conclusion is that all road traffic would be restricted to 70 kph. "Not gonna happen."

I don't follow autonomous vehicle research closely but I'm pretty sure others do not have such timid goals.

No, it did say 70 km/h. Which I agree is very defensive. And the spokesperson did say that a self driving Volvo would "never go above 70 km/h" however you should be careful to say never when speaking about the future IMO.

I agree completely with Volvo. I've been saying this for years. There should never be a need for a car to make a "moral decision." Thank you for translating.

I agree largely, but they are forgetting the more philosophical question which is that if the car doesn't "make a moral decision" i.e. identifies a situation where the risk of hitting a child is very large, and the hardware+software suit "understands" that it could swerve in to another lane and hit another car, but since this wouldn't be 1) legal and 2) not the default action in a crisis (which according to Volvo is "to stay in its lane and brake as hard as possible) the car "refrains" from making a decision. The problem is that this "refraining from making a decision" is actually also a decision, that needs to judged according to some type of moral philosophy. See where I'm getting at? It's the age-old question of consequantialistic ethics v.s. virtue/rule based ethics in a real-word application.
 
Google, Mercedes and I believe Tesla, have also accepted responsibility. I think they realize the only way they'll get legislation changes (in any near-term decision) to allow autonomous driving is to step up and say they'll accept blame.

Yeah, this is a smarter move than I expected of them, though. And it really says a lot about the confidence they (and their insurers) have about the upcoming system.

My biggest fear is that when the first accidents happen (and of course, they will), people will say, "See? These things aren't safe." It's fine to continually evaluate, but my guess is that they'll wind up being significantly safer than human drivers and will still be judged harshly for any "mistakes".
 
Has anyone heard any more rumours of the 8 camera package from Mobileye coming to Tesla?
<a data-cke-saved-href="https://www.youtube.com/watch?v=3yZjw1KgJ68" href="https://www.youtube.com/watch?v=3yZjw1KgJ68">


Like in this video posted earlier somewhere on the TMC forum it sounds like the 8 camera system will come already this year. I am so curious about what car manufacturer will get it. The Model X will be very expensive for us, especially now that our Norwegian currency is much weaker today than when we reserved the car soon two years ago. I am very much hoping for this new autopilot hardware to arrive with our car.
 
Last edited by a moderator:
Like in this video posted earlier somewhere on the TMC forum it sounds like the 8 camera system will come already this year. I am so curious about what car manufacturer will get it. The Model X will be very expensive for us, especially now that our Norwegian currency is much weaker today than when we reserved the car soon two years ago. I am very much hoping for this new autopilot hardware to arrive with our car.

I believe the 8-camera system was the Nissan discussed in the Q3 earnings call in November: "We confirm that our technology is supporting (the) camera processing showcase by Nissan’s Intelligent Driving Prototype Car at the recent Tokyo Motor Show. It is managed by an eight camera system as shown in pictures released by the media and the processing is powered by multiple EyeQ3 chips."

Note he says "first car," not "cars." We already know Tesla has confirmed the X and S will be "lockstep" in autonomy. This makes sense: If the X and S don't get new Autopilot hardware on the same day, it would halt sales of the other model while potential owners wait for the upgrade. That interview (and others) were in September, and the Nissan with 8-cameras and multiple EyeQ3 chips was revealed in November.

It's December 5th, they haven't even delivered a single Signature X. The Production Design Studio that was just released shows identical Autopilot features to the Model S. I just don't see them complicating things in the next three weeks.