[jtchang]

Being able to test your code out in the real world and dark launch a feature gives Tesla a ridiculous advantage. No other company has the infrastructure to be able to do that. Plus they can even tell if the decision the neural net made was wrong because I assume even Tesla knows you got into an accident. And this over thousands of cars. The speed at which they are collecting data is unprecedented.

[BoorishBears]

I'm so tired of this unapologetic Tesla hand holding.

> No other company has the infrastructure to be able to do that.

Because they don't want it. They don't want AP regressions we've already seen happen. They don't want engineers testing code out on real people's lives.

You really think other companies just don't know how to send OTA updates to an ECU or other car components?

I could literally put that together with scraps in my garage, my car's ECU can be flashed over OBD, it'd be trivial to write firmware for a microcontroller to repeat a set of OBD commands downloaded from a webserver.

Obviously an auto manufacturer would do more than that, but there is literally nothing but choice stopping other manufacturers from doing this.

Why is everything Tesla does always suddenly some sort of competitive advantage. Next people will be saying the fact they have cars that spell out "S3XY" is a competitive advantage.

[icanhackit]

Running the latest neural net build in a shadow mode pitted against the real world driver and relaying back telemetry where the shadow NN took an opposing action is a little more sophisticated than pushing out an ECU update.

I think you're being a little uncharitable or simply don't understand what Tesla are doing.

[BoorishBears]

I say you're intentionally confusing two points.

>Being able to test your code out in the real world and dark launch a feature gives Tesla a ridiculous advantage.

Is the comment I'm responding to.

You're intentionally dragging the rest of it into the conversation to make my point seem absurd.

Par for the course for Tesla defenders.

[icanhackit]

>> Being able to test your code out in the real world and dark launch a feature gives Tesla a ridiculous advantage.

> Is the comment I'm responding to.

And that dark launching includes sending the latest neural net builds to the fleet to safely shadow test the changes, collecting millions of hours of real-world telemetry and rapidly iterating without end users even needing to pick their nose. You either intentionally or through ignorance reduced OP's statement to being an OTA ECU update that you "could literally put that together with scraps in my garage".

But feel free to keep shifting the goal posts.

[BoorishBears]

I hate it when people say "keep shifting goal posts" in the most random contexts to sound pithy.

I said no OEM wants their engineers to have a capability and you're fixating on the uses of the capability.

My comment is saying it doesn't matter if they wanted to use it for something as trivial as ECU updates, They. Do. Not. Want. It.

If you want a case study in why, AP regressions are a fine one.

They don't want the liability when an updates goes wrong and kills people. They do not see people's lives as prod vs dev. They do not want to hotfix safety critical systems. They want to get it right the first time, and that's why they're conservative and by god, they get it right a lot more than they get it wrong.

Because no matter how much spin TSLA fanboys put on these things, they are liabilities as much as, if not more, than they are strengths, especially combined with the flippant approach to valuing human life the company that called Adaptive Cruise Control + Lane Centering "Autopilot"

[icanhackit]

> My comment is saying it doesn't matter if they wanted to use it for something as trivial as ECU updates

What I'm saying is that you've completely missed the OP's point: The advantage Tesla has in the autopilot space is that they are able to have hundreds of thousands of real-world test units for their autopilot systems by running the neural net in parallel with the driver inputs (i.e. shadow mode) versus companies that at best have a few hundred test units thus limited material to improve/iterate on their deep learning models.

A metaphor for the situation is Apple playing catchup in the AI and voice recognition space to Google because Google had years of material and huge sums of it to build their deep learning models while Apple had comparatively little. The underlying technologies used by either company might be very similar, but you can't throw hardware at the problem to play catch up -- you need that real-world input/learning material for the deep-learning model to be robust.

[snuze]

As someone in the industry (with domain knowledge), I can tell you that you are oversimplifying the problem and ignoring Tesla’s advantage and leadership in this space. Other manufacturers have to account for legacy systems/designs, it is not as trivial as you make it sound.

[BoorishBears]

I even state that I'm oversimplifying the problem.

Other manufacturers do have to account for legacy systems, you don't need to be in the industry to understand the relationships manufacturers have with companies like Bosch and know they're not in full control of their destiny.

But their advantage is not a "let's ignore dumpster fire financials quarter of quarter to earn them a multi billion dollar market cap" big.

Tesla is not worth 42B dollars because manufacturers can't figure out OTA updates, and the fact other manufacturers don't have OTA updates should be a rounding error when you list advantages that warrant that market cap. It doesn't take any insider knowledge to know that, just common sense, and for far too long people have been ignoring that in preference for TSLA hype.

I got out last year during the 420 nonsense so honestly I don't know why I'm even bothering with all this, the fact anyone needs more than that to see TSLA is not worth your time is interesting. Anything from that point onward was just icing on the bear cake for me.

[mavhc]

Everyone else not doing updates is the symptom of the larger issue, everyone else is a slow moving hardware company that outsources their software. Tesla is a software company first. And software beats hardware whenever you're doing something complex.

Plus lack of legacy self competition, no one else wants to cannibalise their other sales, and no one else has really invested in batteries. Most other car companies can't make enough EV cars to fulfill demand. They're all out of stock until next year because they make so few of them.

[BoorishBears]

Treating safety critical software like a web server is not being fast moving, it's being reckless.

We've seen regressions in AP behavior.

Situations where a route that was safe yesterday will send your Tesla into a concrete barrier if you don't catch it

Simply. 100%. Unacceptable.

Honestly, the more I think about it, the more I do not want to be on the road with these people. I did not sign away my life to be someone's SDC test environment after all.

[mavhc]

Are you referring to https://www.reddit.com/r/teslamotors/comments/b36x27/its_bac... ?

Not good, and they fixed it again.

They should probably put a warning like: Warning: Autosteer is intended for use only on highways and limited-access roads with a fully attentive driver. When using Autosteer, hold the steering wheel and be mindful of road conditions and surrounding traffic. Do not use Autosteer on city streets, in construction zones, or in areas where bicyclists or pedestrians may be present. Never depend on Autosteer to determine an appropriate driving path. Always be prepared to take immediate action. Failure to follow these instructions could cause damage, serious injury or death.

[ReptileMan]

This is a glorified bumper car. Why does it needs updates at all? It should be KISS... instead they complicate for the sake of complication.

[mavhc]

Mostly because people ask for features. And to support new charging stuff. Everyone loves Sentry mode, automatic video recordings of people damaging your car while it's parked.

Then there's "entertain people while charging" stuff.

And then there's autopilot stuff. NoA, Stop light detection, emergency lane departure avoidance, automatic lane changing, conditional speed limits (eg slower in winter)

[gamblor956]

Tesla has to account for legacy systems and designs, since they refresh models all the time without even bothering with distinct ("set in stone") model identifiers (like year). For example, there is no such thing as a "2018 Model 3" as there were multiple variants released over the year with changes over the year, not including the range-based option packages.

In fact, Tesla's legacy systems are actually a disadvantage perspective, since they have to account for a much larger range of hardware configurations than do legacy automakers.

[daryas]

> They don't want engineers testing code out on real people's lives.

When Tesla pushes and runs code in the shadow mode, it is run in parallel to the "production" version of the software. The actual behavior of a car, even with the test code in, still relies on decisions made by the "production" version of the software[1]. The purpose of such testing is to compare decisions made by the updated code vs those made by the old version, and use this info for fine-tuning the next gen algorithms.

[1] https://youtu.be/Ucp0TTmvqOE?t=6679

[BoorishBears]

So? Tesla is still changing the production mode code on their cars with forced updates and no oversight in addition to running code in "shadow mode"

They've used this capability to introduce regressions in the safety of autopilot before.

When a stinky old "traditional automaker" wants to update a safety critical component, they spend months developing and testing before slowly rolling out to dealers. They want to change as little as humanly possible to limit their liability if something goes wrong.

Now some of you think this attitude is a bad thing, and that safety critical systems should be something you can update on a whim like you update your web server and keep iterating on.

But knowing developers like I know developers, being one myself, slow-as-molasses processes for safety critical code is a feature, not a bug for me.

[jayd16]

>Why is everything Tesla does always suddenly some sort of competitive advantage

Because the difference between could do and did is massive.

[Fins]

And difference between "did" and "should or should not do" is even more massive, although completely beyond Elon's (admittedly rather limited) powers of understanding.

[leesec]

You are giving big OEM's way too much credit, imo, haha.

What you suggested would take at least a year for them to setup at all, let alone work the bugs and making it useful.

[mandeepj]

Rest of the car companies are like Boeing. I hope you are keeping up with that. If not here's the latest - https://www.cnn.com/2019/07/24/business/boeing-loss/index.ht...

[693471]

Have you considered that OTA updates undercut the dealership network and could be seen as servicing vehicles which in many states is illegal for the manufacturer to do so they could suffer consequences, maybe getting banned from selling in that state?

[tracer4201]

Seems like those states should change their laws. There’s no law guaranteeing me a job. Why are dealers so special?

[tssva]

Car dealers produce significant local employment and generate significant state and local tax dollars. What is the advantage to the state of favoring an out of state company over them?

[693471]

I agree they should change their laws. This is where Tesla has an advantage. People are importing the cars in states that don't allow it.

[skybrian]

Well, maybe. Since most of this data will be very repetitive, I don't think we can know for sure how much it helps?

[imglorp]

You (and neural nets) learn the most when a boring repetition is violated. The error function was activated and that's worth feeding back into learning.

Ask any kid, they probably couldn't say what happened at school all week. But if they do remember something, it would have been something outside of their routine.

[imglorp]

You (and neural nets) learn the most in the smaller number of cases when a boring repetition is violated. The error function was activated and that's worth feeding back into learning.

Ask any kid, they probably couldn't say what happened at school all week. But if they do remember something, it would have been something outside of their routine.

[skybrian]

Yes, exactly! We don't know the amount or quality of interesting data from the total amount of data.