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

Waymo

This site may earn commission on affiliate links.
Tesla fans have been saying that big data is why Tesla will win for years. Yet Tesla is still at L2. You'd think with such a huge data lead that Tesla supposedly has, they would have more than L2 by now.

Tesla has roughly as many L4 cars as Waymo. They just haven't used them for a massively cash-burning PR spectacle in Arizona.
 
What cars? Show me the part in the disclosure that says this. Tesla said that FSD Beta is L2. Tesla has no L4 software. Does Tesla have a secret fleet of cars with additional sensors that would be L4?

We don't have any evidence Tesla is testing anything higher than L2 internally; but we also don't know what software the test drivers Tesla hired in Austin, Texas back in June, 2020 are doing.

It would be a bit odd for Tesla to hire dedicated autonomy test drivers if they were just testing the exact same software given to FSD beta drivers (why pay folks to do the job so many are willing to do for free?).
 
We don't have any evidence Tesla is testing anything higher than L2 internally; but we also don't know what software the test drivers Tesla hired in Austin, Texas back in June, 2020 are doing.

It would be a bit odd for Tesla to hire dedicated autonomy test drivers if they were just testing the exact same software given to FSD beta drivers (why pay folks to do the job so many are willing to do for free?).

If Tesla were testing L4, we would have seen some evidence of it by now. Heck, Elon would have mentioned it. He tweets about FSD all the time. But like you said, we have no evidence that Tesla is testing anything higher than L2. It makes no sense to assume that Tesla is somehow secretly testing L4 but keeping it so secret that there is no evidence at all. And with no evidence, for shrineofchance to say so unequivocally that Tesla has as many L4 cars as Waymo is wrong IMO.
 
We don't have any evidence Tesla is testing anything higher than L2 internally; but we also don't know what software the test drivers Tesla hired in Austin, Texas back in June, 2020 are doing.

It would be a bit odd for Tesla to hire dedicated autonomy test drivers if they were just testing the exact same software given to FSD beta drivers (why pay folks to do the job so many are willing to do for free?).
Fairly sure Tesla has only the software they have sent out and what Musk tweets about. Its a tight ship - I'm sure all hands on deck approach to FSD.

Tesla has drivers to test the software before its put out to beta testers. Most likely they are asked to test particular conditions / scenarios etc. No different from any other company that spends thousands of hours testing software internally before releasing beta versions.
 
A letter from Tesla to the California DMV:

As described above, Tesla analyzes data from billions of miles of driving received from our customer fleet via over-the-air (“OTA”) transmissions. We supplement this with data collected from testing of our engineering fleet in non-autonomous mode, and from autonomous testing that is done in other settings, including on public roads in various other locations around the world.


(download
here)

A Tesla job posting:

“We are looking for a highly motivated individual to accelerate our vehicle-level testing for autonomous vehicles. ... This full-time role requires the flexibility of traveling (up to 50%) to various domestic and international destinations as necessary, to test vehicles on public roads, test tracks and Proving Grounds.”


original | archive.org | perma.cc

Another Tesla job posting:

“As a Software QA Engineer, you will be responsible for validating all current and future Autopilot features on the path to full self-driving. You will work very closely with the development team to drive the validation of new and existing features, and you will be responsible for reporting and investigating issues that are detected in the field. ... Help transition Tesla's Autopilot to full self-driving by testing features including Autosteer, Navigate on Autopilot, Smart Summon, and more. ... Flexibility of traveling to various domestic and international destinations when necessary, to test vehicles on public roads, test tracks and proving grounds.”


original | archive.org | perma.cc
 
Last edited:
A letter from Tesla to the California DMV:

As described above, Tesla analyzes data from billions of miles of driving received from our customer fleet via over-the-air (“OTA”) transmissions. We supplement this with data collected from testing of our engineering fleet in non-autonomous mode, and from autonomous testing that is done in other settings, including on public roads in various other locations around the world.


A Tesla job posting:

“We are looking for a highly motivated individual to accelerate our vehicle-level testing for autonomous vehicles. ... This full-time role requires the flexibility of traveling (up to 50%) to various domestic and international destinations as necessary, to test vehicles on public roads, test tracks and Proving Grounds.”


original | archive.org | perma.cc

Another Tesla job posting:

“As a Software QA Engineer, you will be responsible for validating all current and future Autopilot features on the path to full self-driving. You will work very closely with the development team to drive the validation of new and existing features, and you will be responsible for reporting and investigating issues that are detected in the field. ... Help transition Tesla's Autopilot to full self-driving by testing features including Autosteer, Navigate on Autopilot, Smart Summon, and more. ... Flexibility of traveling to various domestic and international destinations when necessary, to test vehicles on public roads, test tracks and proving grounds.”


original | archive.org | perma.cc

This is not L4. I think you are misreading the job descriptions. This is testing the current AP/FSD software on the path to becoming autonomous some day. It says it in the job description:

“As a Software QA Engineer, you will be responsible for validating all current and future Autopilot features on the path to full self-driving. You will work very closely with the development team to drive the validation of new and existing features, and you will be responsible for reporting and investigating issues that are detected in the field. ... Help transition Tesla's Autopilot to full self-driving by testing features including Autosteer, Navigate on Autopilot, Smart Summon, and more. ... Flexibility of traveling to various domestic and international destinations when necessary, to test vehicles on public roads, test tracks and proving grounds.”
 
If Tesla were testing L4, we would have seen some evidence of it by now. Heck, Elon would have mentioned it. He tweets about FSD all the time.

I'm still not so sure. If Tesla had test drivers testing L4 in Austin, what evidence would we possibly have? The vehicle wouldn't look any different from the outside, and you would have to carefully watch the person behind the wheel to see if they ever had to interact with it.

And we know Tesla is capable of L3 in a limited capacity. They had to report the miles drive on Autonomy day as such. If that's possible, why haven't they produced a single video of anything higher than L2 since Autonomy day? Publicity wise, we already know Tesla isn't showing all of their cards.
 
Anyway, I don't want to sidetrack this thread any more than it already has been. The topic is Waymo, and while it's useful to compare to Tesla, it's not always the most productive way to talk about Waymo's progress.

I'm curious about that paper you're proposing @shrineofchance . Is that being done as part of an academic institution, financial organisation, or your own personal knowledge?
 
  • Like
Reactions: diplomat33
Yes that is the large dataset that I was referring to. It was released for research purposes, not for commercial use.

570 hours of recorded driving is not a lot in the context of what companies like Waymo, Cruise, and Tesla must have internally. Say you have 500 cars driving every day for 2 years. Say the average number of times a car encounters something worth snapshotting per day is 1. And assume 20-second snapshots, like the Waymo dataset. You'd end up with a dataset of over 2,000 hours of recorded driving.
 
  • Like
Reactions: mikes_fsd
I'm curious about that paper you're proposing @shrineofchance . Is that being done as part of an academic institution, financial organisation, or your own personal knowledge?

I'm commissioning the paper out of my own personal burning curiosity. As an investor (an individual retail investor), it's worth paying for original research that sheds needed light on this subject.