Elon: FSD Beta tweets

[diplomat33]

The other metric is the cost of what Waymo's doing in creating their geofence. Add the cost of roving support vehicles (staffed with humans) and that's a helluva sunk and ongoing cost proposition to sell a service at or below taxi/Lyft/Uber per passenger mile.

The cost of remote assistance and roving support vehicles will come down over time. That's a temporary evil because it's the only way to help an autonomous car when there is no safety driver. As the Waymo Driver gets better with edge cases, Waymo will need fewer and fewer roving support vehicles or remote assistance. Eventually, Waymo will not need any remote assistance at all.

 

[FloridaJohn]

So when can we bring a Waymo down to Irvine and have it go head to head with Beta 9 FSD?

Serious question: when?

When will Tesla not require someone in the driver's seat? Seems like that would be important for a fair head to head challenge.

 

[qdeathstar]

Maps isn't the issue. Waymo can generate the 3d centimeter accurate maps fairly quickly. The issue is risk taking. Waymo won't take risk of harming people. Tesla has and will continue to do so. Although Tesla tech is inferior it will win in the end against everyone else because of ability to take risk.

Why don’t they make the maps and release level 5 beta everywhere? Or at least more than one city. There is zero evidence what waymo is doing is scalable....

 

[Daniel in SD]

Why don’t they make the maps and release level 5 beta everywhere? Or at least more than one city. There is zero evidence what waymo is doing is scalable....

The lose money on every ride but make it up in volume strategy?
This argument is a bit silly. Is there any evidence that they won't be able to continue to reduce costs to the point that it is scalable? That's sort of the way new technology usually works. The original Roadster only sold a couple thousand units...
I'm not saying it's guaranteed that they'll be able to get the costs low enough but it certainly seems possible.
There's also the whole problem of achieving greater than human safety, they may not have even done that yet...

 

[mikes_fsd]

This argument is a bit silly. Is there any evidence that they won't be able to continue to reduce costs to the point that it is scalable? That's sort of the way new technology usually works.

Except the last generation of the Waymo driver hardware was the one that was supposed to be "ready for the job and scalable".

 

[Daniel in SD]

Except the last generation of the Waymo driver hardware was the one that was supposed to be "ready for the job and scalable".

Haha. Yeah, maybe if you leave out all the other costs of running the service. Don't they still have support vehicles following them around? Extremely frequent use of remote assistance? The hardware is probably fine, it's the software that still needs a lot of work.

 

[rxlawdude]

When will Tesla not require someone in the driver's seat? Seems like that would be important for a fair head to head challenge.

No, because you will note my challenge says TODAY. Waymo, outside its pitifully limited geofence, vs Tesla V9 FSD.

THAT would be a fair challenge in late July 2021, not some future date. I'll buy the beer after the head-to-head. So, diplomat, use your diplomatic skills to get a REAL, TODAY test* so we can stop this absurd debate.

Oh, wait. You and Waymo can't. Pity.

*Subject to a willing Beta 9 FSD owner in the LA/OC area who would volunteer. (I'm guessing several would be interested, but get back to me after securing that Waymo car.) FSD 9 is running on real streets, with no geofence. Is it "driverless?" Nope. But can Waymo actually work in a novel city? Nope.

 

[DanCar]

Why don’t they make the maps and release level 5 beta everywhere? Or at least more than one city.

They are money losing operation. The more they scale the more money they lose.
They don't even cover all of one city yet.
Waymo has trouble merging on freeways. Afraid they will get a law suite if they touch another vehicle.
Waymo is switching to new gen hardware, are they waiting on that before expanding?
Waymo wants to be very careful where they expand and make sure there is very close to zero chance of an accident. Entails running lots of tests and fixing everything that could be a potential issue.
Waymo has been running in many cities with safety drivers, so they already have HD maps for many places.

There is zero evidence what waymo is doing is scalable....

I remember Chris Urmson from 5+ years ago saying HD Maps is no big deal. So maybe Waymo isn't going to scale, but HD Maps isn't the issue. Google maps has mapped the world and I suppose Waymo is using similar tech.

 

[DanCar]

...
There's also the whole problem of achieving greater than human safety, they may not have even done that yet...

They are way past that from my inside sources and external disengagement data. Being better than average human driver is woefully inadequate for Waymo. When the Waymo driver had a scrape with a bus, the entire team was depressed for a week.

 

[Daniel in SD]

They are way past that from my inside sources and external disengagement data. Being better than average human driver is woefully inadequate for Waymo. When the Waymo driver had a scrape with a bus, the entire team was depressed for a week.

How can they possibly know that? Their collision rate looks ok, about 1 per 100k miles in their safety paper, but they only drove 6.1 million miles. That doesn't seem like enough when the fatality rate is 1 per 100 million miles.
It shouldn't be that hard to dodge a bus, I would be upset too.

 

[DanCar]

How can they possibly know that? Their collision rate looks ok, about 1 per 100k miles in their safety paper, but they only drove 6.1 million miles. That doesn't seem like enough when the fatality rate is 1 per 100 million miles.
It shouldn't be that hard to dodge a bus, I would be upset too.

The company said that no one was seriously injured and “nearly all” of the collisions were the fault of the other driver. In other words Waymo shouldn't get penalized for being rear ended by a careless driver.

Waymo pulls back the curtain on 6.1 million miles of self-driving car data in Phoenix

Waymo was involved in 18 collisions and 29 near-misses during the last two years in Phoenix.

www.theverge.com

 

[Daniel in SD]

The company said that no one was seriously injured and “nearly all” of the collisions were the fault of the other driver. In other words Waymo shouldn't get penalized for being rear ended by a careless driver.

Waymo pulls back the curtain on 6.1 million miles of self-driving car data in Phoenix

Waymo was involved in 18 collisions and 29 near-misses during the last two years in Phoenix.

www.theverge.com

Of course they should. An autonomous vehicle that drives erratically or not defensively enough such that it gets into more accidents than an average human driver isn't really acceptable imho.
Really though my concern is edge cases that haven't shown up in 6 million miles of testing. For all we know it runs over a group of kids every 50 million miles. They need more testing.

 

[diplomat33]

The company said that no one was seriously injured and “nearly all” of the collisions were the fault of the other driver. In other words Waymo shouldn't get penalized for being rear ended by a careless driver.

Waymo pulls back the curtain on 6.1 million miles of self-driving car data in Phoenix

Waymo was involved in 18 collisions and 29 near-misses during the last two years in Phoenix.

www.theverge.com

While the safety data that Waymo released is very encouraging that Waymo is safe, 6M miles is not enough data to statistically prove that Waymo is safer than humans. Waymo needs more data to actually prove safety with high confidence. I believe it is one reason why Waymo is investing so much in Simulation City. If you try to brute force it with only real world miles, you need about 2B miles to prove safety. So if Waymo wants to get out of the geofence, they are going to need a lot more than 6M miles to prove safety. If they can get Simulation City to be a reliable measure of safety, then Waymo can use it to supplement their real world safety data and prove safety that way, a lot faster than trying to just brute force it.

 

[qdeathstar]

They are money losing operation. The more they scale the more money they lose.

yea. I know. That’s my point. Lol. Waymo: a waste of time and resources.

 

[DanCar]

Of course they should. An autonomous vehicle that drives erratically or not defensively enough such that it gets into more accidents than an average human driver isn't really acceptable imho.

The way accidents per mile is calculated is fraught with issues. If you think Waymo is in more accidents than average driver then examine your sources. For example are statistics based on serious accidents, accidents reported to the police, accidents where driver is at fault? May find it hard to correlate with the Waymo data.

Really though my concern is edge cases that haven't shown up in 6 million miles of testing. For all we know it runs over a group of kids every 50 million miles. They need more testing.

You might be surprised how sophisticated the 30 billion miles of simulated driving is.

 

[momo3605]

Let's try this challenge: TODAY, let's take a Waymo vehicle with all its sensors on non-mapped streets, vs an FSD Tesla that hasn't been on that same roaute. Let's compare how the two drive over a couple of miles.

Oh, wait. This can't happen, because Waymo can't do it. Am I exaggerating? Waymo guys?

Is this a trick question? Because the correct answer is Neither cars can safely navigate a route in Irvine autonomously!

One car makes mistakes and requires a supervising driver (Tesla). The other can operate without a driver but is limited where it’s allowed to operate (Waymo).

 

[DanCar]

And I have spent hundreds of hours of my life and terabytes of disk space to make this a reality. Nobody had be better saying there's a lack of Waymo videos from independent third parties

I haven't seen many videos except yours. How many videos are there from others? How to find videos from others? Does someone have a feeling how many non JJRicks video exist? Last time I search some months ago, I didn't find any recent non JJRicks videos.

 

[qdeathstar]

The fact that there needs to be someone in the drivers seat is irrelevant. If the Tesla makes it there without intervention it has made it there autonomously.

 

[diplomat33]

I haven't seen many videos except yours. How many videos are there from others? How to find videos from others? Does someone have a feeling how many non JJRicks video exist? Last time I search some months ago, I didn't find any recent non JJRicks videos.

If you do a youtube search, you can find other videos. But they are not as good as @JJRicks videos. The other videos tend to be shorter, not the entire trip. JJ really did a great job. Not just his videos show the full trip unedited with a nice camera, but he did a great job organizing them as well.

 

[momo3605]

The fact that there needs to be someone in the drivers seat is irrelevant. If the Tesla makes it there without intervention it has made it there autonomously.

Sorry…did I miss the line in the Tesla beta agreement that says, “You, the tester, are irrelevant to this beta FSD functionality. Feel free to lie back and watch Netflix while your car whirs you away from Irvine to LAX”