With Tesla using Intel, does this mean Mobileye may come back?

[activejustin]

It seems like Tesla is struggling a little more than expected with the autonomous features since they switched to Nvidia.

To me it seems strange to have half your system from Nvidia for the autonomous driving and the other half from Intel for the infotainment.

Do you think since Intel owns Mobileye, Tesla might reengage with Mobileye technology once again and eventually switch to a double Intel package from the current Nvidia and Intel mix?

 

[mblakele]

To me it seems strange to have half your system from Nvidia for the autonomous driving and the other half from Intel for the infotainment.

That doesn't seem strange to me. Many general-purpose computers in the world use multiple vendors for different components. For example it's common to use one vendor for the CPU and another for the GPU. It's also not much different than a car maker sourcing the stereo from one vendor and the ABS from another.

Do you think since Intel owns Mobileye, Tesla might reengage with Mobileye technology once again and eventually switch to a double Intel package from the current Nvidia and Intel mix?

Possibly. I wouldn't bet on it just yet. I think Musk continues to believe in vertical integration, and sees autonomous driving as a key technology for Tesla. I also think Musk has more tolerance for risk — of driver fatalities, not to put too fine a point on it — than Mobileye. Intel is a fairly cautious company, so I don't see why their influence would change that.

 

[eye.surgeon]

The Tesla-Mobileye split was acrimonious. There will be no reunion.

 

[activejustin]

Possibly. I wouldn't bet on it just yet. I think Musk continues to believe in vertical integration, and sees autonomous driving as a key technology for Tesla. I also think Musk has more tolerance for risk — of driver fatalities, not to put too fine a point on it — than Mobileye. Intel is a fairly cautious company, so I don't see why their influence would change that.

I agree Musk is huge on vertical integration, but seems like they are blowing by all the goals he set.

Provide Time for a Better Long Term Solution
Switching back to Mobileye may solve some of the problems they are having in the short term and take the pressure off. Tesla could work over a longer time period to develop a better in house solution without a gun to the head. Going back to the joint Mobileye and Tesla until one day they could just switch everything over to Tesla.

Musks Original Preferred Strategy
I think that was Musks original strategy to have Mobileye and Tesla tech side by side for a better safety margin during the development process. Now they are partnering with Intel already, not sure why they might not go back to it.

Intel Has Experienced Lawyers
Mobileye was a relatively small company that would have had a difficult time with any litigation, pretty sure that Intel would not be so scared off by that, they have great lawyers and deep pockets. They are a conservative company, but they would be somewhat shielded by Tesla on the liability front and also give them a good opportunity to continue vetting out their technology.

Everybody At Risk
At some point in the self driving continuum, everyone who wants to be a player is going to have to take a big liability risk when they first get started. If I was Intel I would rather be partnered with an experienced company like Tesla, that has semi proven technology. Rather than partnering with someone totally untested like a GM etc.

Autonomous System without Over the Air Updates
I do think it is funny when an autonomous system rolls out without over the air updates. If there is a serious problem, the automaker will be required to do an expensive recall to get everyone in the dealership for your updates vs the over the air process Tesla is using. I think I recall reading that Cadillacs system cannot be updated over the air, I can't remember if Mercedes is in the same boat.

 

[activejustin]

The Tesla-Mobileye split was acrimonious. There will be no reunion.

I would say that was a small spat in the scheme of things and they are now under new ownership. If the relationship was that bad, Tesla would not be working with Intel now.

There was this guy who had a really acrimonious split with the technology company he worked for and then one day he came back. They treated it like the second coming, Steve Jobs.

At some point every true technology company will seek an optimal solution when faced with a very difficult challenge. You may not realize how great the technology was, until you can't use it.

Even if it means eating a little crow and asking out your ex... Do you not remember when Elon went back with his other ex for a time...

 

[Cosmacelf]

It seems like Tesla is struggling a little more than expected with the autonomous features since they switched to Nvidia.

To me it seems strange to have half your system from Nvidia for the autonomous driving and the other half from Intel for the infotainment.

Do you think since Intel owns Mobileye, Tesla might reengage with Mobileye technology once again and eventually switch to a double Intel package from the current Nvidia and Intel mix?

Yeah, I made this point in another thread, but no one seemed to understand what I was getting at. Since we now know that Tesla is using Intel for infotainment in Model 3, I just can't believe that Tesla would have not also negotiated to re-use Mobileye at the same time, now that Intel owns Mobileye. That's step 1. We also know that the original V2 AP processor board actually had a spot on the board for a Mobileye chip. So it wouldn't take a huge energineering effort to start using it again. And we know that Tesla has been struggling trying to even replicate Mobileye, let alone surpass it.

Someone will have to do a tear down of the Model 3 AP hardware for us to know, but I think it is a possibility that they go back to using Mobileye as a backup. Going forward, it is clear that Tesla wants to develop its own chip technology for AP, so not likely to use Intel for that, other than whatever Mobileye has as a backup or even primary system depending on how well their chip project goes.

By the way, that's another agreement that Tesla and Intel probably signed. Obviously Tesla isn't going to actually make their proprietary AP chip, so they may have inked a deal for Intel to fab it.

 

[stopcrazypp]

It seems like Tesla is struggling a little more than expected with the autonomous features since they switched to Nvidia.

To me it seems strange to have half your system from Nvidia for the autonomous driving and the other half from Intel for the infotainment.

Do you think since Intel owns Mobileye, Tesla might reengage with Mobileye technology once again and eventually switch to a double Intel package from the current Nvidia and Intel mix?

I don't see how that is unusual at all. In AP1, nVidia ARM chips handled all the multimedia and the Mobileye MIPS chip handled semi-autonomous driving.

So why would it be unusual for Tesla to use an Intel x86 chip for infotainment, while they used Nvidia ARM and CUDA based architecture for semi-autonomous driving?

Note the Intel infotainment chip (Gordon Peak, Apollo Lake) has a completely different architecture with the Mobileye chips (which will continue to be MIPS based).

Going with a hybrid approach made sense in the empty period when AP2 had nothing, but right now it doesn't, since AP2 is much further along.

 

[anticitizen13.7]

I doubt Tesla will return to Mobileye. The hiring of Jim Keller, allegedly to create custom chips, indicates that Tesla is likely to go with proprietary hardware to run its Autopilot system in the future.

 

[grashelm]

I doubt Tesla will return to Mobileye. The hiring of Jim Keller, allegedly to create custom chips, indicates that Tesla is likely to go with proprietary hardware to run its Autopilot system in the future.

^^^ Agreed...don't see a return to mobileye/intel for autopilot.

 

[calisnow]

...it seems strange to have half your system from Nvidia for the autonomous driving and the other half from Intel for the infotainment.

Why?

 

[calisnow]

I agree Musk is huge on vertical integration, but seems like they are blowing by all the goals he set.

If you look at the actual timeline of AP2 it is progressing very rapidly - look at dates, releases. AP2 is not stagnating and is 90-95% of where it needs to be to have parity with AP1. The releases are now coming every 2 weeks, massive video uploads are going to Tesla, refining the neural nets for exceptions and corner cases. The progress is just fine. The real "hard part" of FSD is going to be the logic of driving with other cars in complex traffic scenarios - and that has nothing to do with Mobileye. That is a whole different software problem and of course Tesla has been working on that in the background. Mobileye is just object recognition and path planning - which Tesla has now spent a year building as rapidly as possible from scratch and is making rapid progress on. Mobileye may have a big head start in object recognition but it doesn't have any advantage over Tesla in driving logic. In fact, Tesla still holds a possible ace in the 100K+ AP2 cars on the road allowing the FSD software to "shadow drive" and test its decisions in traffic. No other automaker has that capability and Waymo doesn't have nearly the fleet. During the next 12 months the existing AP2 fleet will drive 1.3 billion miles and the cars sold from now til then will drive possibly a billion more (depending on the speed of the model 3 ramp). Each car is uploading tens of gigabytes of video data to Tesla on a regular basis now. This project started only in May and there's no other fleet in the world that can remotely compete. Hold your horses and see where we are in 12 months.

Provide Time for a Better Long Term Solution
Switching back to Mobileye may solve some of the problems they are having in the short term and take the pressure off...Tesla could work over a longer time period to develop a better in house solution without a gun to the head. Going back to the joint Mobileye and Tesla until one day they could just switch everything over to Tesla.

Not sure if you remember but Tesla has already explicitly stated that they were doing exactly what you advise - but that Mobileye demanded Tesla stop working on its own solution, demanded Tesla share what it learned with Mobileye, demanded to retroactively change the terms of their deal and more. Then Tesla says Mobileye refused to license any more chips to Tesla. Mobileye denies all this and says it was about "safety concerns" - but Amnon is a hard nose, shrewd businessman. Be your own judge - if Tesla is telling the truth the split wasn't its choice.

... Now they are partnering with Intel already, not sure why they might not go back to it.

If you were Mobileye - why on God's green earth would you work with Tesla at this point? Musk is well known for building his own tech - he isn't a partner, he's a user and a temporary one. What a stupid biz move for Mobileye if Mobileye agreed to sell chips to Tesla again - knowing that Tesla was simply using them to buy time while they build their own tech. Musk's entire life history is of pulling this move over and over again with all kinds of purchased tech.

Intel Has Experienced Lawyers
Mobileye was a relatively small company that would have had a difficult time with any litigation, pretty sure that Intel would not be so scared off by that, they have great lawyers and deep pockets.

This is irrelevant if Tesla's claim is true - that Mobileye's safety concerns were a smokescreen. In any case that's all in the past now and it certainly is not in Mobileye's interest to provide Tesla with vision technology.

They are a conservative company, but they would be somewhat shielded by Tesla on the liability front

and also give them a good opportunity to continue vetting out their technology.

How so? Mobileye has plenty of other suitors - higher volume suitors - to run its vision on public roads. It doesn't need Tesla and Tesla doesn't need it.

...If I was Intel I would rather be partnered with an experienced company like Tesla, that has semi proven technology. Rather than partnering with someone totally untested like a GM etc.

Wha??? What are you talking about? Mobileye doesn't need Tesla - any car is fine.