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

Waymo Says Its Self-Driving Tech Is More Advanced Than Tesla FSD, Others

This site may earn commission on affiliate links.
All the Tesla fans getting upset that their Lord Musk might be beaten to it by Waymo.

To be fair I supposed Tesla probably will be pretty badly screwed when someone does get there first. Musk has bet too much on FSD being soon and being first.

I actually don't think Tesla fans should be too upset. Tesla produces great cars. I think most Tesla fans really love AP. And even if FSD is not fully driverless and requires driver attention, it could still provide a lot of benefits to Tesla owners in terms of added convenience and added safety. Imagine a system that does require the driver to watch the road but that can basically drive you anywhere, anytime, is safer than human driving, is cheap to add to cars and is available on all Tesla cars. That's essentially what Tesla seems to be heading towards with FSD Beta. That would still be a huge win for Tesla IMO. It would crush anything that other automakers offer on their vehicles right now.

As I see it, at least in the short term, it looks like there won't be any 1 company that wins FSD. Rather, we will have different companies offering different types of FSD. Waymo and Cruise will offer driverless ride-hailing in limited areas. Tesla will offer a driving system that requires driver attention but can drive anywhere, anytime. Other automakers might offer L3 that offers hands-free driving but only on highways under specific conditions. I think those different business models will coexist.

Also, even if Waymo did "win" by offering driverless robotaxis in all US cities, there would still be a place for Tesla because Tesla offers very compelling EVs. I think there will still be folks who will want to own a Tesla especially with the FSD package.
 
Imagine a system that does require the driver to watch the road but that can basically drive you anywhere, anytime, is safer than human driving, is cheap to add to cars and is available on all Tesla cars. That's essentially what Tesla seems to be heading towards with FSD Beta. That would still be a huge win for Tesla IMO.

I dislike your hypothetical. You'd have an autonomous system that entirely replaces the human driver, when: 1) it is known that humans fail to remain as attentive when unprompted; 2) The human would not be driving the car, FSD would.

To legally require "the driver to watch the road" would not just be an evidently absurd cop out to avoid liability, but also make me question the utility of the feature: now I'd need to pay mind to both the road and the car's the decisions. If not allowed to legally relax the mind in a situation that should mostly leave me unprompted (I wouldn't be driving) while adding a new variable to monitor (FSD's decisions), why not drive the car myself?

How much trust do you think people would have in claims that FSD is safer than human driving without liability assumed by Tesla? And would that trust suffice to offset the discomfort of the effort required to remain attentive without driving? I imagine it'd be a product for either hardcore enthusiasts or the ill-informed. I disagree that there's a place for something that can "drive you anywhere, anytime" but requires "the driver to watch the road", in that the latter means Tesla wouldn't take responsibility for its algorithm's faults, when it is effectively driving the car.
 
I dislike your hypothetical. You'd have an autonomous system that entirely replaces the human driver, when: 1) it is known that humans fail to remain as attentive when unprompted; 2) The human would not be driving the car, FSD would.

To legally require "the driver to watch the road" would not just be an evidently absurd cop out to avoid liability, but also make me question the utility of the feature: now I'd need to pay mind to both the road and the car's the decisions. If not allowed to legally relax the mind in a situation that should mostly leave me unprompted (I wouldn't be driving) while adding a new variable to monitor (FSD's decisions), why not drive the car myself?

How much trust do you think people would have in claims that FSD is safer than human driving without liability assumed by Tesla? And would that trust suffice to offset the discomfort of the effort required to remain attentive without driving? I imagine it'd be a product for either hardcore enthusiasts or the ill-informed. I disagree that there's a place for something that can "drive you anywhere, anytime" but requires "the driver to watch the road", in that the latter means Tesla wouldn't take responsibility for its algorithm's faults, when it is effectively driving the car.

What I mean is a system that has the capability of driving you anywhere, anytime but does not have the reliability yet to remove the driver. That seems to be what FSD Beta is right now. It can handle traffic lights, stop signs, making turns at intersections, navigating around parked cars, avoiding obstacles etc... You can put an address in the nav and turn on FSD Beta and it will drive you there. But Tesla says in the release notes that the driver must have hands on wheel at all times because quote, "it might do the wrong thing at the worst time".

I agree that a true autonomous driver system should not require driver attention. The car is driving, not the driver. But Tesla has chosen to release FSD features when they are still "beta" and do still require driver attention because it is not reliable yet.

And Tesla has used the cop out for awhile now. Just look at the FSD order when you buy a new car. On one hand, Tesla sells a bunch of features as part of the FSD package but on the other hand, still requires driver attention. Every FSD feature released has had a disclaimer in the release notes about how it is beta and the driver needs to keep their hands on wheel at all times.
 
  • Like
Reactions: Microterf
There is another hypothetical and that is the driver doesn't have to pay attention to immediately intervene, for example to avoid an accident.. But the driver has to be able to take over within a minute to deal with extreme edge cases. For example, a road closure or an officer who has to divert traffic because of an accident. Situations where FSD would be smart enough to alert the driver.
 
There is another hypothetical and that is the driver doesn't have to pay attention to immediately intervene, for example to avoid an accident.. But the driver has to be able to take over within a minute to deal with extreme edge cases. For example, a road closure or an officer who has to divert traffic because of an accident. Situations where FSD would be smart enough to alert the driver.

That sounds like level 3 autonomy.
 
What I mean is a system that has the capability of driving you anywhere, anytime but does not have the reliability yet to remove the driver.

You had previously proposed a scenario with a system that is "safer than human driving" while requiring "the driver to watch the road", and that's that with which I disagreed ;) I do agree that even if Waymo succeeds first, Tesla fans shouldn't be upset if Tesla does eventually succeed afterward; it's an entirely different service, and will be a massive boon to Tesla vehicles over alternatives.

And Tesla has used the cop out for awhile now. Just look at the FSD order when you buy a new car. On one hand, Tesla sells a bunch of features as part of the FSD package but on the other hand, still requires driver attention.

I wouldn't call it a cop out in this regard just yet, although the FSD sale and price raise far before it has been realized, and under the systematic promise that it's coming in two years or next year since 2015, is an issue nonetheless. It is still explicitly unready and in a trial phase, so the driver attention requirement is understandable. (The perpetual beta label on all of the other notoriously barely functional features such as Summon or Autopark is indeed a cop out of a different nature, though. I find it hard to fathom how systematic underdelivery flies so well with the Tesla community)

Its current utility is what I'd question, then. It's not full self driving giving the lack of responsibility and frequency of error. With some basic knowledge of autonomous driving algorithms and the CEO's recent tweet that confirms they're far from video in, control out, I'm confident the control side of FSD doesn't rely on a learning algorithm or massive amounts of data, but rather more traditional motion planning coupled with traditional control methods. So what's the purpose of the beta? Testing the success rate of a critical algorithm on real people? With some additional minor feedback on points of failure, which is to be implemented manually, thus making the utility of making it widely available to the public questionable?

EDIT: I withdraw the motion planning comment. It's possible that it is indeed data-driven, making the benefits of a large fleet in varied scenarios clear. But because mistakes are still manually reported and of concerns regarding the quality of data, I'm still skeptical of the safety of their Beta approach. Regardless, Tesla's success, also given the benefits of their unique approach, would be damned interesting and certainly great for Tesla enthusiasts, regardless of Waymo.
 
Last edited:
You had previously proposed a scenario with a system that is "safer than human driving" while requiring "the driver to watch the road", and that's that with which I disagreed ;)

I'm thinking of that fuzzy area where FSD is safer than human driving but not by a lot. So it is safer but maybe not safe enough where Tesla would remove the driver. For example, if FSD was 10% safer than human driving, that would be safer but would Tesla remove the driver? Probably not. So I could see a scenario where FSD was 10% safer than human driving and would still require the driver to watch the road.
 
  • Like
Reactions: Microterf
That sounds like level 3 autonomy.[/QUOTE
.

If I can be in the front seat and legally be distracted so long as I am available to take over in a minute or so when FSD tells me, then that will drive huge interest in FSD and I'd find that super valuable.
Doesn't have to be robotaxi capable since the vast majority of people won't go into the robotaxi business anyway
 
  • Like
Reactions: diplomat33
Krafcik is really cool. I asked for a recap of his talk during the Future of the Car conference and he replied and did a quick recap on Twitter of his main talking points.

tmV8Yjt.png


jUOBHfC.png
 
Way to prop up that weak straw man. :rolleyes:

There's very little overlap between Waymo and Tesla so it being first doesn't matter--

- A few Tesla owners might be owner/operators of a ride sharing service, if and when it becomes available from Tesla. I won't be adding my Tesla to any ride sharing service.
- No one can buy a Waymo car to be an owner/operator.

Anyone can get a ride from their friend, spouse, a taxi service, Uber, Lyft, ....etc... right now. Waymo will be another service provider. Some people might choose Waymo over the current available options, but that doesn't matter to me as a Tesla owner with FSD.


I actually don't think Tesla fans should be too upset. Tesla produces great cars. I think most Tesla fans really love AP. And even if FSD is not fully driverless and requires driver attention, it could still provide a lot of benefits to Tesla owners in terms of added convenience and added safety. Imagine a system that does require the driver to watch the road but that can basically drive you anywhere, anytime, is safer than human driving, is cheap to add to cars and is available on all Tesla cars. That's essentially what Tesla seems to be heading towards with FSD Beta. That would still be a huge win for Tesla IMO. It would crush anything that other automakers offer on their vehicles right now.

As I see it, at least in the short term, it looks like there won't be any 1 company that wins FSD. Rather, we will have different companies offering different types of FSD. Waymo and Cruise will offer driverless ride-hailing in limited areas. Tesla will offer a driving system that requires driver attention but can drive anywhere, anytime. Other automakers might offer L3 that offers hands-free driving but only on highways under specific conditions. I think those different business models will coexist.

Also, even if Waymo did "win" by offering driverless robotaxis in all US cities, there would still be a place for Tesla because Tesla offers very compelling EVs. I think there will still be folks who will want to own a Tesla especially with the FSD package.


This is a weak straw man.

All the Tesla fans getting upset that their Lord Musk might be beaten to it by Waymo.

To be fair I supposed Tesla probably will be pretty badly screwed when someone does get there first. Musk has bet too much on FSD being soon and being first.
 
  • Love
Reactions: mikes_fsd
If I can be in the front seat and legally be distracted so long as I am available to take over in a minute or so when FSD tells me, then that will drive huge interest in FSD and I'd find that super valuable.
I agree Level 3 is extremely valuable for drivers. Many are stuck in traffic an hour or more each day, it would be a huge benefit to free up that time to read and answer e-mails or even just watch a movie. Audi announced their Level 3 "Traffic Jam Pilot" on the A8 three years ago, but withdraw it due to legal issues (and perhaps also poor performance, no way to know). The A8 was the first production car with lidar. Today we see high end models from Mercedes, Toyota, Honda, BMW, Volvo, etc. adding lidar. Many are promising Level 3 where legalization is nigh -- so far Japan and Germany(rest of Europe to follow). You can bet GM will follow suit with Supercruise, and of course China won't want to fall behind.

For all their talk about skipping "dangerous" Level 3, Waymo could sell this type of capability to OEMs. But for now it's mostly Mobilieye (Intel) and NVIDIA processing with lidar from Valeo, Luminar, Innovision, Ibeo, etc. It would not surprise me to see regulations evolve to require lidar and/or high res radar. Lobbyists from companies who sell 99% of cars carry a lot of weight.
Doesn't have to be robotaxi capable since the vast majority of people won't go into the robotaxi business anyway
Shhhh, don't tell Tony Seba that :)
 
Bonus clips with some cool info.


A couple interesting points:
- Yielding to emergency vehicles with visualization
- 3 point turn in dead end parking lot with extra cones
- Self-cleaning of camera/lidar dome

The cleaning routine is really cool IMO. Waymo has a solution for keeping the sensors clean.

We also get a great up close view of the lidar dome. We can see there are cameras placed in pairs all around the dome for 360 degree coverage.