Waymo is geo fenced and needs remote control. This is an expensive lab experiment.
True, but you have allowed yourself to get sucked into a simplistic level 2 vs 4 contrast, Tesla does make large numbers of end to end trips without error or interruption. Lots of film. Lots of varied environments. Also so far with enough trips with errors of some kind, sometimes errors of taste, that no one is trying to call it level 4. Waymo manages good enough in cities to call itself level 4. But the contexts are totally different. Only with the Austin trial will we see how it does in city only driving with the latest software. Maybe in 6 months they are still trying to get a reasonable percent of completed trips in cities. Maybe concentrating on cities that long will get them an acceptable error rate. But remember that is what it is .. not some perfect level, but a difference in error rate.
Point being that they are not different kinds of driving but a different error rate. You are also confusing the comparison between Waymo in cities and Tesla everywhere.
1 Like
Um, Tesla Level 4 is, how to put this delicately?
Nowhere.
Waymo is Level 4 in five cities, I think? Limited, surely, but the comparison between 5 and 0 is (checks calculator) infinite.
I have to admit to being amused to see people saying that companies which are actually providing driverless transport - both of humans and of freight - are dismissed as irrelevant, while Tesla, which promised both and (so far at least) has delivered neither, is defended to the max.
Odd.
3 Likes
Cruise, the autonomous vehicle company backed by General Motors, achieved Level 4 (L4) autonomous technology with its Cruise Origin vehicle, which was designed for fully driverless operation in specific conditions, such as urban environments, without manual steering controls. In February 2022, Cruise became the second service provider after Waymo to offer driverless taxi rides to the general public in San Francisco, operating at L4 in a limited geographic area. However, following a series of incidents, including a pedestrian collision, California suspended Cruise’s driverless permit in October 2023, and the company halted its robotaxi service. By December 2024, GM stopped funding Cruise’s robotaxi efforts, shifting focus to integrating the technology into advanced driver assistance systems for personal vehicles. So, while Cruise successfully implemented L4 technology for a period, it faced significant setbacks and is no longer operating L4 robotaxis.
1 Like
It’s not that no one’s “trying to call it Level 4.” Tesla’s system isn’t Level 4. At no time, and in no environment, is it safe to allow Tesla’s system to operate the car without full and active monitoring by a human in real time. That’s a fundamentally different characteristic of their system than a Level 4 AV, which can be allowed to drive the vehicle without a human involved in real time.
I think you’ve allowed yourself to be sucked into thinking that there’s not much real difference between L2 and L4 systems but for the error rate. But the two have massively different characteristics when operating. Ice isn’t just very cold water that’s a few degrees colder still - at some point, when it’s cold enough, it takes on completely different attributes that water doesn’t have.
3 Likes
Labels of l2 and l4 are less important than reality. Tesla FSD is very advanced and getting better every day. They make 35,000 cars every week.
Waymo and Uber are toast.
1 Like
The reality is that Tesla’s FSD isn’t advanced enough to be operated without realtime human supervision. The little information we have (crowd sourced data and observations from the Austin open beta) show it being orders of magnitude away from being able to operate autonomously. There’s no evidence that it’s “getting better” fast enough to be a viable AV product anytime soon.
The difference between L2 and L4 isn’t just labelling. It’s functionality. An L4 system can do something fundamentally different that an L2 system cannot. It doesn’t matter whether Tesla makes 35K cars a week if they’re only operating at L2, because that level of driver assistance isn’t sufficient to serve in an AV fleet.
2 Likes
They said this about Cruise before it was shut down. Disaster.
They said what about Cruise? That it was a Level 2 and not a Level 4 system?
2 Likes
They actually are different types of driving, and in context of this thread are L2 and L4 are regulated differently. Recall the claim in the original post in this thread:
The problem is that when it comes to AV ride hailing, all those L2 miles in cities and highways all across the world plus five bucks will get you a latte’.
Regulators want to see documented L4 miles, and thus far Tesla hasn’t submitted any that I’m aware of and hasn’t even started the permit process in any jurisdiction as far as I know (Texas does not require special permits for AV ride hailing).
A reasonable person would conclude that puts Tesla’s end of the year goal at risk, especially given they don’t have a product yet.
2 Likes
No. Cruise was labeled L4 just like Waymo is. Disasters.
But that doesn’t mean Waymo will be a disaster, any more than it will mean that Tesla will be a disaster when it gets to L4. And that doesn’t address any of the deficiencies in Teslas system - even if Waymo were to fail, Tesla could still fail to achieve autonomy as well.
2 Likes
Your whole argument is Waymo is labeled L4 so is superior.
Cruise was labeled L4 and got shut down. It was a mess.
Just like Cruise, Waymo is lidar based, expensive and bloated tech that does not scale and on life support with help of its daddy. Just like Waymo, all manual interventions were remotely operated.
There were many “lidar based” promising companies that are shut down and died.
1 Like
The argument is that Waymo actually has L4 and Tesla doesn’t.
2 Likes
I think you misunderstand the argument.
I think it’s entirely possible that Waymo fails. I think the most like outcome is that both Tesla and Waymo fail to achieve any significant commercial success in the robotaxi business in the next five or so years. The tech just isn’t there yet.
I talk about Waymo because all your criticisms of Waymo are accurate - they just also apply to Tesla. The tech is bloated and expensive and doesn’t seem to work well enough to support a solid business model. But that’s also true of Tesla. Their bloat and expense is on the software side and (for now) on the personnel side with their in-car staff, rather than hardware - but Tesla’s systems are also still super-expensive, costing billions and billions of dollars.
You can’t have a viable robotaxi service without at least L4 technology, because there’s little economic benefit if you still need an attentive driver in the car. And while there’s little solid information about Tesla’s overall performance, what there is strongly suggests they’re not really close to that yet. They can do L2 almost anywhere, it seems, but every indication is that they’re several years away from the “once in 100K miles” without someone in the car that they’ll need in order to have a real robotaxi.
Waymo’s approach may not work, but that won’t mean Tesla’s will, either.
2 Likes
No. Waymo and Tesla’s approach is very different. Waymo’s approach is similar to many autonomy companies that are dead including Cruise. They overloaded on sensors and their software is bloated. It has to be geofenced with the HD maps.
Tesla’s approach is completely different. It is the ONLY one with that approach.
Last I read, Waymo is beginning to overtake (market share) traditional uberlyft in some areas. That may be one measurement of commercial success.
4 Likes
According to the articles in Goofy’s post, Aurora Innovations is operating driverless Class 8 trucks on TX public roads.
AUR uses lidar, radar, vision cameras and training simulators.
AUR provides the software, Continental AG provides the trucks.
Various IC vendors provide the ICs
ralph owns some AUR cause it appears the tech just might be nigh.
1 Like
So? That doesn’t mean their approach will work. So far, they also have spent billions and billions of dollars on their robotaxi program - and like Waymo (and Cruise before them), it’s resulted in only a limited deployment of geofenced robotaxis. They haven’t even yet reached the point where they can remove the in-car humans, despite being five years further down the road on tech than when those other systems entered testing phase. And all available (though limited) data suggest that their system isn’t anywhere close to sufficient to serve as a robotaxi.
Different, but with no better results thus far - and nothing to indicate their results will be better going forward.
Yes. I’m not saying technology for autonomous driving isn’t ready - I’m saying technology for autonomous driving for an economically successful taxi service isn’t ready. Taxi service presents issues (technological, operational, and economical) that are different than with trucking. Especially long haul trucking.
2 Likes