Speculate: what the heck happened to Chris Lattner?

[gowthamn]

Apple does not mind releasing a feature late, but they do it well. That is not how tesla works as they release unfinished software. May be it was a clash of culture.

 

[croman]

Bottom line is it's obvious EAP isn't where it needs to be and someone was responsible and it wasn't going to be Elon's come to Jesus moment.

 

[whitex]

Is it though? People have been saying this for years yet sales keep increasing as does the stock - those are real metrics whereas judgment about what people tolerate is anecdotal at best. As long as its capabilities exceed everything else available for purchase, my speculation is people will continue to tolerate less than perfect software.

As mentioned @supratachophobia , us long time customers no longer hail the product as loudly, and no longer buy into the hype. I honestly tell people that Tesla is a really great car (we are a Tesla only household today), but expect some quality issues which will be taken care of but it's still a hassle, don't buy "P" as it is bleeding edge and potentially unreliable (see 691hp or countergate controversies, etc), and don't buy unfinished hyped options like EAP or FSD. I bought 3 Model S'es so far, but no longer willing to spend the money on P or EAP or FSD. No worries, once the AP2 car can drop me off at work, drive itself home, then come pick me up, Tesla will get my $10K for FSD. According to Elon, I'll be writing that check "sooner than people think".

 

[whitex]

Apple does not mind releasing a feature late, but they do it well. That is not how tesla works as they release unfinished software. May be it was a clash of culture.

"release unfinished software"? Lol. Try "release software ideas", as was EAP when initially released and FSD still is today. Gotta hand it to Elon though, he's been able to sell complete vapourware for thousands of dollar a piece.

 

[boofagle]

Apple seems to only announce things once they are imminently ready for the public.

Apple just gave specs and announced a new "iMac Pro" last month. It's not supposed to go on sale until December.

 

[BigD0g]

From Chris's resume:

[paste:font size="4"]first early release (which had few features and was limited to 45mph on highways) to effectively parity with HW1, and surpassing it in some ways (e.g. silky smooth control).

• This required building and shipping numerous features for HW2, including: support for local roads, Parallel Autopark, High Speed Autosteer, Summon, Lane Departure Warning, Automatic Lane Change, Low Speed AEB, Full Speed Autosteer, Pedal Misapplication Mitigation, Auto High Beams, Side Collision Avoidance, Full Speed AEB, Perpendicular Parking, and 'silky smooth' performance.

• This was done by shipping a total of 7 major feature releases, as well as numerous minor releases to support factory, service, and other narrow markets.

• One of Tesla's huge advantages in the autonomous driving space is that it has tens of thousands of cars already on the road. We built infrastructure to take advantage of this, allowing the collection of image and video data from this fleet, as well as building big data infrastructure in the cloud to process and use it.

• I defined and drove the feature roadmap, drove the technical architecture for future features, and managed the implementation for the next exciting features to come.

• I advocated for and drove a major rewrite of the deep net architecture in the vision stack, leading to significantly better precision, recall, and inference performance.

• I effectively stopped regretted attrition from the team, and ended up growing it by over 50%. I personally interviewed most of the accepted candidates.

• I made massive improvements to internal infrastructure and processes that I cannot go into detail about.

• I was closely involved with others in the broader Autopilot program, including future hardware support, legal, homologation, regulatory, marketing, etc.

In the end, Elon and I agreed that he and I did not work well together and that I should leave, so I did.

Overall I learned a lot, worked my butt off, met a lot of great people, and had a lot of fun. I'm still a firm believer in Tesla, its mission, and the exceptional Autopilot team: I wish them well.

 

[McRat]

Probably unrelated.

In aerospace, when there was a problem with quality, you needed to perform Corrective Action.

Rather than actually finding the root cause, they would declare it was a failure of the quality system management.

The Corrective Action was firing the QC Manager. Some companies would go through more than 1 per year until they figured out you should not heat your home by burning it down. It is better to build a robust heating system that will reliably heat the home and leave it standing, even if a box of matches and a gallon of gas is cheaper.

 

[Shateam]

I think we can stop speculating...


Chris Lattner's Resume

Tesla’s Autopilot team grew by ‘over 50%’ in the past 6 months, says former VP of software

 

[suraj1194]

My speculation is that Tesla and Chris both realized they weren't dealing with a software engineering problem but a deep learning / computer vision problem .

Chris is justifiably famous for the LLVM infrastructure, but he is ultimately a great software engineer, not a machine learning expert .

Andrej Karpathy on the other hand IS one of the most well known persons in the CNN/computer vision field . He's among those who have most pushed forward the state of the art from Krizhevsky, Sutskever and Hinton's groundbreaking 2012 AlexNet work .

Considering Tesla didn't just hire some computer software engineering guy but a theory expert from academia , in my opinion they want to accelerate image artifact discrimination in future AP, and found they don't have necessary inhouse expertise .

 

[S4WRXTTCS]

I never really understood his hire to lead an Autopilot team. It never did seem like a good fit for that part of it, but Tesla desperately needed someone who could manage software teams.

I fail to see why he couldn't lead the teams responsible for the other aspects of the SW. Like getting the media player up to par.

 

[suraj1194]

The creator of LLVM and Swift would be hopelessly overqualified to handle media player / misc updates...

 

[BluestarE3]

In the end, Elon and I agreed that he and I did not work well together and that I should leave, so I did.

In the end, it could be as simple as this: two strong personalities not meshing. Elon certainly isn't leaving, so it had to be Chris.

 

[suraj1194]

Well, Chris' responsibilities have been split between the Autopilot hardware guy and Karpathy, who's neither a software engineer not a hardware person, but a theoretical computer scientist among the most well known in the young field of ConvNets for Computer Vision . Karpathy and Lattner don't have related backgrounds at all .

 

[S4WRXTTCS]

The creator of LLVM and Swift would be hopelessly overqualified to handle media player / misc updates...

It's not just that, but the overall software outside of AP.

In fact one reason why I'm strongly suspect of self-driving is the current nav is nowhere near the point needed for a self driving car.

It's a monumental task, and it's not just deep learning. So I fail to see why there isn't room for both of them. So it's more like he's culturally not a good fit. He certainly brings awesome assets to the table.

 

[suraj1194]

Even the overall software outside of AP doesn't need someone like Lattner to manage . The guys already done far more . Running general software would be a downgrade for him; he created LLVM and Swift .

 

[S4WRXTTCS]

Even the overall software outside of AP doesn't need someone like Lattner to manage . The guys already done far more . Running general software would be a downgrade for him; he created LLVM and Swift .

But, there is a massive amount of things that need to be done for edge computing.

Edge computing is the next big thing and it's happening at Tesla, Apple, Google, etc. Whether it's a phone, car, robot, etc.

So no I don't agree that there wasn't room. He likely simply wasn't a good fit culturally. I don't think it had anything to do with his knowledge or his capacity to contribute on a very high level.

I'm a bit bummed because I felt like he could have gotten the SW under control. Where we would see more polished releases. Instead of the "oops, we broke something again" releases we get.

 

[GasKilla]

Chris should go back to what he knows best, basketball!

Christian Laettner - Wikipedia

 

[whitex]

Hmmm, so given their software release history to-date, Tesla decided to put a scientist in charge of production software?

 

[suraj1194]

But, there is a massive amount of things that need to be done for edge computing.

Edge computing is the next big thing and it's happening at Tesla, Apple, Google, etc. Whether it's a phone, car, robot, etc.

Lattner is not a cloud/edge computing guy either. He developed LLVM and Swift, which are a compiler framework and computer language respectively. He's justifiably famous in his own domain of expertise, but it makes him suitable for managing a general software development effort, not a domain specific one. If they wanted a cloud expert, there are plenty to poach from Amazon AWS, MS Azure or elsewhere.

In the case of computer vision, just throwing people at it is not enough. It's a new, rapidly evolving field where Karpathy is among the pioneers. He's an expert on dense image classification, something Tesla would find valuable in Autopilot. On the other hand, Karpathy is probably not suited to general software engineering.

 

[john_lam]

I was fully expecting Chris to have been leading a team to implement a high level language for writing GPU and FPGA code. That's in his wheelhouse, and something that could greatly improve the productivity of their engineers. Given Elon's history of donating patents to the world, I would also have expected this new language to have been Open Sourced as well.

Clearly that's not going to happen now. But it's still needed ...