Blog WSJ: Ambitious Autopilot Push Angered Tesla Engineers

Tesla engineers are reportedly jumping ship because they don’t believe Autopilot 2.0 hardware can meet the fully self-driving promise that Elon Musk proudly touts.

The Wall Street Journal says Sterling Anderson, previously the Autopilot director, decided to leave Tesla in December in part because he didn’t agree with the claims Musk was making about the vehicle’s potential for full autonomy.

According to the WSJ (paywall):

In a meeting after the announcement, someone asked Autopilot director Sterling Anderson how Tesla could brand the product “Full Self-Driving,” several employees recall. “This was Elon’s decision,” they said he responded. Two months later, Mr. Anderson resigned.

The Autopilot division has lost some 10 employees and four managers recently, according to the report. Satish Jeyachandran, the former director of hardware engineering for Tesla’s Autopilot team, and Berta Rodriguez-Hervas, a former machine learning manager also left the company in June. Anderson was succeeded by Chris Lattner, a former Apple developer, but he left in June after just six months on the job.

Tesla has declined to comment on the report.

 

[BigD0g]

The problem is the silence and more importantly the lack of evidence from Tesla is deafening. The only thing we have is some woefully outdated oversold demo from last November.

 

[Krugerrand]

Even though you use the word 'reportedly', it would be great if you would clearly call this out as speculation with only one source (former employee) stating this to be true.

It's great to keep the forum balanced, but this writeup feels like it's pushing the spin.

Not only that, but this is old, recycled information.

 

[Tam]

One might argue it should no longer be an "impossible challenge" when you are taking people's money for it.

Some might leave for ethical considerations, in such a scenario.

I wouldn't assign blame on Tesla or those who leave it.

I respect those who leave it due to various reasons including ethics.

But that does not mean Tesla is unethical in promising something impossible.

I think that's how Tesla was created: Promising things that sane people would think crazy, impossible or plain fraud.

It may take Tesla some time but it has proven that it has risen over obstacles and become successful along the way.

 

[Phrixotrichus]

Oh no! Not one more of the full self driving demo videos! BTW, Delphi did that in March 2015.
Delphi completes first coast-to-coast automated drive | KurzweilAI

Yep.

And those guys shortly after
http://gizmodo.com/this-self-driving-car-just-completed-a-1-500-mile-journ-1738266952


Delphi plans to "maybe" have autonomous driving ready by 2022...maybe. And their cars have quite a different sensor setup from Tesla btw.

 

[AnxietyRanger]

But that does not mean Tesla is unethical in promising something impossible.

Taking people's money for something impossible would definitely be unetchical.

We shall see what is possible, though.

 

[conman]

I wonder if it will be on AP2 or AP2.5. Has the goalpost moved or do they make a point of keeping the old ones?

Not that I don't think AP2 can't do it HW wise. I'm pretty sure it can - I worry more about its regulatory feasibility, but a coast to coast demo is not concerned with that...

AP2.5 can be retrofitted to AP2 if needed it's been said so AP2 vs AP2.5 won't be rate limiting per se. I understand Elon's point though, as others have said, that if it can't be done with the hardware included in AP2(+) then it can't be done at all. Relying on software recognition of everything with the many camera inputs, radar and u/s sensors seems logical... if a heck of a lot of work and smarts. I suspect that AND high resolution route mapping will be required for absolute level 5 FSD with our current technology limits (note I'm a software engineer.)

 

[AnxietyRanger]

AP2.5 can be retrofitted to AP2 if needed it's been said so AP2 vs AP2.5 won't be rate limiting per se

Not sure that is true. We now know HW 2.5 includes new radar, redundant wiring, redundant power and fuses... a CPU/GPU upgrade is one thing, a suite upgrade another.

 

[Tam]

...Taking people's money for something impossible would definitely be unetchical...

That's a true statement but Tesla does not think it's impossible.

Tesla has proven that it has never been unethical over and over again!

People said you can't convince any one to buy an EV because there's no market.

It's as plain as daylight.

But Tesla sold Roadster as an EV.

People said it's impossible to bring the price down for the next version but Tesla did with Model S and X.

Panasonic personally thought that it's impossible to sell that many cells so why even thinking about Gigafactory but now it admits it's wrong and yes it can.

People even said Model 3 was a fraud, but here now it is.

Just because some engineers who don't believe it's possible, and yes there have been many many problems and obstacles, but Tesla has proven again and again that it is possible.

 

[AnxietyRanger]

That's a true statement but Tesla does not think it's impossible.

Tesla has proven that it has never been unethical over and over again!

I guess some people with P90DL Vx's might disagree...

We shall see.

 

[Tam]

I guess some people with P90DL Vx's might disagree...

I am not familiar with that case. Did Tesla took their deposits and didn't deliver the product?

 

[AnxietyRanger]

I am not familiar with that case. Did Tesla took their deposits and didn't deliver the product?

They delivered product revision aftee another that was not capable of meeting its specs. Your only fix: buy a new one.

Some of it is in court now.

P85D another such debacle, of course.

There is precedence to worry about AP2.

 

[conman]

Not sure that is true. We now know HW 2.5 includes new radar, redundant wiring, redundant power and fuses... a CPU/GPU upgrade is one thing, a suite upgrade another.

I haven't read about the new radar but in terms of inputs to actions the others won't make a difference, but in terms of regulation requirements and safety, you're right - they're totally different goal posts.

 

[AB4EJ]

He seems to me like the kind of person who lives in the world of the possible, rather than the real world. This has advantages, such as when he pushes and pokes at reality and is able to do amazing things.. it also has disadvantages such as overpromising and underdelivering on everything from delivery dates to feature capabilities.

While I'm not going to cancel my M3 pre-order... As an engineer with limited understanding of the technology challenges I've never really believed that full self driving was going to arrive in this generation of vehicle and would have to be a kind of an extreme optimist to pay $3000 up front for the tantalizing promise of future tech that might never arrive.

This reminds me a lot of the "reality distortion zone" around Steve Jobs. Elon is a classic dreamer (think of Nikola Tesla himself). The AP2 hardware came out a month after I took delivery of new MS. I thought about this briefly, and then realized that by the time Tesla (and everyone else) gets this working, and state laws are updated to allow the actual use of it, my MS will be worn out and ready for replacement, and I will simply buy the new MS (or whatever it is by that time, in 10-15 years), and be ready to go! My new Tesla can take me, without my manual intervention, to my nursing home.

 

[Tam]

...specs...

I see why there is doubt of Tesla because of different specs expectations.

As you said, let the court decide.

 

[Tam]

...Oh no! Not one more of the full self driving demo...

It's easy to ridicule that Tesla is planning a self-driving Coast-to-Coast trip as others have done that already.

AP2 is like a baby so I would be glad that if it could stand and walk.

No one has ever seen AP2 can pass a traffic cone test yet so any thing to demonstrate its progression is welcome.

By the way, Delphi trip required 34 miles of manual human operation out of 3,400 mile trip.

That's is a very impressive 99% self-driving.

Let's see whether Tesla will be able to beat that if they can install Supercharger Cable Bot just in time so there will be no human intervention needed at anytime during the trip.

 

[AnxietyRanger]

I haven't read about the new radar but in terms of inputs to actions the others won't make a difference, but in terms of regulation requirements and safety, you're right - they're totally different goal posts.

Indeed the lack of redundancy in AP2 and the added redundancy in AP 2.5 might mean the difference between FSD regulatory approval and none.

We shall see. Tesla unfortunately does not have a very good history of remembering old promises when moving onto new generations of product.

Mind you, of course AP2 will still get major updates. I am however a bit less sure about the FSD capability.

 

[conman]

Indeed the lack of redundancy in AP2 and the added redundancy in AP 2.5 might mean the difference between FSD regulatory approval and none.

We shall see. Tesla unfortunately does not have a very good history of remembering old promises when moving onto new generations of product.

Mind you, of course AP2 will still get major updates. I am however a bit less sure about the FSD capability.

I only just got my Model S a bit over a month ago and FSD was an exciting extra feature they were offering to me, but not the reason I bought the car. Now that I'm more informed I've resigned myself to accepting that in all likelihood, for the effective life of my vehicle I am unlikely to get level 5 FSD. On the other hand, I am excited and anticipating various features of FSD as they become unlocked. Speed signs are currently relegated to AP1, but I don't think it'll be long before we see them on AP2. Then I hope to see things like traffic light, stop sign recognition, spontaneous lane changes and so on. I do think all of these will begin to happen sooner rather than later but the harder tasks... who knows At this stage I'd be happy to get autowipers even... auto high beams were delayed long enough and now that I have them I don't use them because inevitably, like all first releases, they're crap. But unlikely other manufacturers, at least there is always the potential for software upgrades to make them good.

My understanding and respect for Tesla as a technology provider has changed drastically since I received my car, not necessarily for the better...

 

[AnxietyRanger]

But unlikely other manufacturers, at least there is always the potential for software upgrades to make them good.

This is IMO where it all started to go wrong for Tesla. Starting with P85D in 2014 and its promised performance coming in a future update... that was the first distinct over-promise and underdeliver of this nature. Its specs were never met, same with P90DL V1, V2... AP1 traffic light detection, summon to curb, not even ramp-to-ramp is yet here though it may come...

A long list of forward-looking promises based on software updates unfulfilled and instead moved to the next hardware generation. And here we are again, AP2 replaced by redundant AP2.5 long before any EAP or FSD (let alone even AP1 parity) is even here for AP2...

Mind you, Tesla also has history of over-promising hardware retrofits. The promised lighted vanity mirrors for all Model Ss never came, promised 100 kWh upgrades to P90DLs bought late last year and supposed to come in early 2017 are not here...

A long list of broken and excessively delayed forward-looking promises means that IMO Tesla's word is so worthless on them that I would prefer they make none and concentrate on selling only what they are delivering today.

I know, doesn't make for quite as good demand levers, but it would be the right thing to do.

 

[hockeythug]

I really hope they stop accepting money for full self driving. Has anybody sued yet?

 

[AnxietyRanger]

Tesla could have introduced the exact same AP2 and later AP2.5 suites without making any EAP or FSD promises.

They could simply have sold software upgrades to them later once those became actually available.

There was no need to promise AP1 either originally, with curb summon and traffic lights and many things never materialized... they could have just added the hardware and started selling TACC, AP software etc. once each step was really done.

They could stick to making similar progress but marketing and selling only those features that are shipping today.