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

Frustrated with FSD timeline

This site may earn commission on affiliate links.
I've seen the demo, yes. I wonder how much it was just nVidia's SDK talking, though.

Be that as it may, I guess my point is: I don't believe that demo was the full monty with just neural net training and a slight coat of polish missing. If it were, I might agree with you.

Frankly, in late 2016, I believe the results show that they had tons of actual programming missing. The spat with the MobilEye probably caught them a bit off guard, software-plan-wise. Also the constant personnel changes within the division are not very reassuring.

Probably today they are far more advanced now, but for me a lot of this is going through Tesla's statements in October 2016 through January 2017, and how accurate those were compared to what has unfolded and what we now believe the real status to have been then.

Not very, IMO.
I never thought Chris Lattner was a good fit. He never had the background for that role. I'm sure he can lead a team, but his resume was a little light on the machine learning experience (at least from his Linkedin page).
 
I never thought Chris Lattner was a good fit. He never had the background for that role. I'm sure he can lead a team, but his resume was a little light on the machine learning experience (at least from his Linkedin page).

If that was the only bit of turn-around it might be different, but I think a lot of stuff points to inner turmoil.

Just pondering on this one. We're delving into areas where details are scarce, so better not make too many proclamations. :)

We shall see...
 
  • Like
Reactions: JeffK
For them to do this as a driver's aid they'd need constant nagging to the point where it doesn't work at all without a human there (otherwise it'd be classified as an autonomous vehicle). I think we may see a phase in of features instead, which still allows periods of nag free driving, but incapable in all driving situations until regulatory approval. This is the way to most easily get around the law.

You can't simply say it's a lvl2 driver's aid though unless you omitted features otherwise it's an autonomous vehicle (according the the state of CA at least).

People seem contradictory. On one hand they want Tesla to release potentially unsafe software before it's complete and on the other hand these same people criticize the software for being unsafe. It boggles the mind.

This is completely incorrect. GM Supercruise for example allows you to drive hours on hours with no nag. aslong as you are paying attention. that's the entire point.
 
Done would imply that validations are complete.... so it's pending validations. I hope that makes sense.

Personally, I try to test all my software before release too and if I find issues I fix them.

Its not pending validation. its pending development. do you even know what validation is? or do you simply accept Tesla BS pr?
 
I imagine my company selling a customer a software application that has a lifecycle of 3-5 years. What sold the customer however was a software extension to the product that we said we are building. The customer loved the demo but we didn't disclose the issues we ran into making the demo, the fact we just dumped our the industry leading SDK to try and do it ourselves, or that we lost key people on the project. We just took the money while making public statements about where we would be by the end of 2017.

Now imagine that we start giving our customers inaccurate project updates via twitter and then just stop giving updates at all. When the customers call to complain, we avoid the calls and tell our reps to cite the fine print about QA and be vague.

I imagine I would have unhappy customers....
 
Not an autonomous car by the California code definition because it still requires driver watching the road or the car pulls over and disables autopilot. That's the software piece that keeps it from being autonomous so they can avoid the self-driving requirements of the California code.
And it stops at stop lights, makes left hand turns, etc? If so this must be a new and improved supercruise.
 
The customer loved the demo but we didn't disclose the issues we ran into making the demo,
They disclosed the disengagements to the DMV at the end of the year.

... but yeah, for now, it's an interest free loan to Tesla.

Now imagine that we start giving our customers inaccurate project updates via twitter and then just stop giving updates at all. When the customers call to complain, we avoid the calls and tell our reps to cite the fine print about QA and be vague.
You mean inaccurate estimates... they are just that, estimates. Elon is pretty optimistic. Even on the financial calls you can hear the team rein him in sometimes.
 
I do this for a living actually. So there's that.

Obviously you don't because what you regurgitate from TESLA narrative is completely not true.

You are clearly not a software engineer who works in a software company.
Unlike you, I'm a senior software engineer and work for one of the biggest healthcare insurance service company.

Validation (Unit, QA, SIT, UAT) in the SDLC only happens once development is complete. Not when its in its infancy.

Tesla incorrectly masquerades early development stage as validation to give the impression that development is done when 0 codes have been written.
 
  • Love
Reactions: oktane and NerdUno
They disclosed the disengagements to the DMV at the end of the year.

... but yeah, for now, it's an interest free loan to Tesla.
I didn't see that disclosure on the Tesla website next to the demo or product description. I only see the fine print about regulatory approval.

Not that it matters given the above, how does disclosing at the end of the year help people who purchased beforehand? They obviously knew well before they disclosed to DMV.
 
I didn't see that disclosure on the Tesla website next to the demo or product description. I only see the fine print about regulatory approval.

Not that it matters given the above, how does disclosing at the end of the year help people who purchased beforehand? They obviously knew well before they disclosed to DMV.
They [customers] knew an LA-NY demo wasn't even scheduled until the very end of 2017, so they weren't/shouldn't have been expecting anything sooner. This was definitely disclosed publically.

Validation (Unit, QA, SIT, UAT) in the SDLC only happens once development is complete. Not when its in its infancy.
Except there's this thing called agile software development... maybe you've heard of it. ;) It's based on the same principles Tesla has worked toward from the very beginning, even with their manufacturing process.

Obviously if Elon truly thought FSD features would diverge from EAP by now, then whatever they were planning to release didn't pass validation right away.
 
  • Like
  • Disagree
Reactions: J1mbo and NerdUno
JeffK - what is your relationship with Tesla? Please disclose:

1) Tesla owner
2) Tesla investor
3) Tesla employee
4) Tesla contractor (including employees of companies paid to post positive comments on message boards)
5) None of the above

Would be good to know to put your comments in proper context. Let us know. I'm sure it will come out anyway when this thread gets subpoenaed in upcoming litigation.
 
Just saw this article from end of April.

Elon Musk clarifies Tesla’s plan for level 5 fully autonomous driving: 2 years away from sleeping in the car

It helps to clarify that demo is a demo. Fsd is about 2 years away (being developed by Tesla, not stuck at dev complete and waiting for approvals). some fsd features will be enabled along the way before fsd is deliver 2+ years from April.

Article conclusion which I would agree with:
"Musk believes Tesla will solve computer vision by the end of the year, hence the timeline for the demo drive across the country, but it may take another year to prove the reliability of the system and bring it to a level 5 autonomous capability in all driving modes."
 
JeffK - what is your relationship with Tesla? Please disclose:

1) Tesla owner
2) Tesla investor
3) Tesla employee
4) Tesla contractor (including employees of companies paid to post positive comments on message boards)
5) None of the above

Would be good to know to put your comments in proper context. Let us know. I'm sure it will come out anyway when this thread gets subpoenaed in upcoming litigation.

JeffK - I see you rated the post as funny, but a disclosure would be helpful to the discussion. I am happy to disclose that I am (1) a Tesla owner.
 
  • Like
Reactions: NerdUno
Article conclusion which I would agree with:
"Musk believes Tesla will solve computer vision by the end of the year, hence the timeline for the demo drive across the country, but it may take another year to prove the reliability of the system and bring it to a level 5 autonomous capability in all driving modes."
They've already indicated the coast-to-coast drive isn't happening in 2017 so it appears the timeframes in that article are now off too.
 
  • Disagree
Reactions: MP3Mike and JeffK
So GM supercruise is an autonomous car now?

Cadillac has not classified their SuperCruise to any SAE autonomy level. They are referring to it as a geomapped hands-free high/low-speed cruise control system augmented with other safety systems such as low and high speed AEB, FCW, blindspot, pedestrian, and long range thermal imaging. The night vision defeats blinding headlights, and has far more range than high beam headlights can offer.

It is a component from a family of digital driver's aids, not the entire program. SuperCruise claims to have OTA update ability, which will actually be a requirement as the geomapped database changes.

There are at least 3 basic AV technologies in concurrent development at GM. V2V released in the 2017 Cadillac CTS (1000 signals per second, 300m range), the LIDAR system on the CTF Bolt EVs which are now assembled on a production line, and the geomapping used on Cadillac CT6.

When, how, and even if, they merge these technologies is not being discussed publicly. But all three technologies are on public roads today, not on concept car brochures. Mostly Captured Test Fleet cars. These are production cars that are instrumented and operated by engineers. All GM cars and trucks log millions of miles before retail release.

I'm not sure how systems without thermal imaging will fare in the real world of deer strikes, and the new brighter headlights of oncoming cars in hazy conditions.
 
Last edited:
  • Informative
Reactions: NerdUno
They've already indicated the coast-to-coast drive isn't happening in 2017 so it appears the timeframes in that article are now off too.

I agree that timelines are off, but the situation is clearly stated. Tesla doesn't have fsd worked out. When they do, it won't be reliable enough for a while. During all this time, we will see new eap/fsd features incrementally added. Plus, it's Elon timelines so I thought that was obvious by now that milestones are unlikely to be met.

Finally, when fsd is reliable enough in 2+ years time, it may or may not be gated by government approval.

I do wish all of this was stated on Tesla. Com when adding fsd feature. I was really hoping to see the roadmap or something similar for fsd features. However, I'm not going to make a fuss about buying fsd because in the end, all should be aware that fsd isn't ready, isn't approved by govt, isn't available in any fashion, and will not be for a while.
 
I'm sure many customers did their due diligence by reading the specifications on the website, reading what the big car magazines/websites said, and talking to people at the store/dealership. With most all other car manufacturers, this is a pretty reliable process. What most people didn't realize is that Tesla plays by different rules, requiring an understanding of things like "Elon-time" and Tesla-speak.

The primary issue is that Tesla should not be selling features that aren't expected to be available as described for at least a reasonable portion of the car's lifecycle.