On self-driving, Waymo is playing chess while Tesla plays checkers

We'll know Tesla is serious about robotaxis when it starts hiring remote operators.

See full article...

 

[balthazarr]

See, this is where I struggle. It's for the cars to manage the situation more than for first responders to learn how to co-exist with cars. If an officer is dealing with rapidly closing off a road due to a fatal collision that led to serious road hazards, they need to stop cars, make space for incoming emergency vehicles, and do it quickly. For officers to what... learn Tesla Sign Language? If an AV detects emergency lighting on the road, then it knows something is out of the norm. To me it's totally on the Teslas and Waymos of the world to sort this out.

Absolutely agree. Tim's example of EVs isn't really analogous, because all EVs use high-voltage and batteries - once they learn how to deal with one system, they can deal with them all.

Apparently there are 1800 numbers that LEOs, first responders, roadworkers etc. are "supposed" to call.... yeah, that's workable.

It it totally unreasonable to expect road users to learn 5, 10, 100, however many, systems for communicating/dealing with AVs... there needs to be one standard that all AVs must adhere to. It's safer, it's more workable, and it doesn't expect taxpayer funded workers to take time out to learn umpteen different systems for how to deal with these vehicles.

 

[thrillgore]

Tesla, Cruise, and Zoox are also under NHTSA investigation. My sense is they are investigating everyone to cover their bases. Why would this be relevant?

Someone on Lemmy posted a list of all the incidents by brand of self driving, and it wasn't even a contest: Cruise, Waymo, etc all came in under 15 or so reported incidents. Tesla had over 200. I'll try to find the report.

Speaking of incidents there was another high profile one this week. This technology isn't even half baked or outright unfinished, its garbage. EM doesn't want to fix this, he wants enough control of Tesla so they can't remove him from his personal fiefdom. The closest we'll get to a fix is 'Autopilot' handing it over to the driver to place blame.

 

[kirkhilles]

Excellent article, thank you! I know that there are some dedicated Tesla fanatics but I would imagine that most Tesla lovers are more realistic and most would probably like Musk to give up his CEO position.

My issue is that for BOTH companies, lack of data is not an issue. Both have available so much data that even adding 10 more years of data really won't improve the models. You have both extremes here with too cautious and too aggressive and I'm not sure either is going to be adequate. Avoiding interstates at year 15 (or whatever it is) of dev is concerning.

I have only one REAL question for Waymo. How many support people do you have for each driverless vehicle? How many employees are going through videos needing to "explain" changes to signs, roads, etc as well as assist the vehicles, fuel, maintain the vehicles, etc? That ratio would tell you everything. If you need 1+ person for every vehicle then this is all academic. You need to get to where you can have 10+ vehicles for every person and get to where you can add more cities quickly without having to do a mass hiring.

 

[Montezuma3000]

Off topic, but I see Waymo's remote intervention happening first for no-pilot cargo flights. And maybe even for some commercial flights as backup to pilots in the near future. Then removal of pilots in the far future.

Human beings are like the metaphorical frog in gradually heated water. Our memories are short and we get comfortable with the new reality more quickly than we think.

 

[itfa]

In one of the recent threads, I commented that Waymo have approached this with orders of magnitude more responsibility, and their use of multiple sensor types means they're not reliant on one modality that is subject to common and very predictable failure modes (eg. heavy inclement weather)... and they're still running into parked cars.

There is zero chance, absolutely ZERO that Tesla releases a safe, reliable robotaxi on 8 August. It's just the usual Musk blather (only this time designed to convince shareholders to again approve the ludicrous compensation package). Remember FSD has been coming "later this year" for over a decade now.

Come on now, full self driving is coming out next year.

In all seriousness, I don't know why Tesla hasn't tossed Musk out on his ass with how much he lies to investors.

 

[NAM37]

To me, long-term the difference is lidar. Waymo has it. Telsa doesn't.

 

[Tim Lee]

It's new and only involving incidents on their latest generation of cars?

The NHTSA notice cited 22 incidents going back to 2021. Here are the five most recent:

"While making a tight left turn from southbound [XXX] onto northeastbound [XXX], the Waymo AV made contact with the driver side of a parked passenger car. At the time of the impact, the Waymo AVs Level 4 ADS was engaged in autonomous mode, and a test driver was present (in the drivers seating position). Both vehicles sustained damage." (March 2024)

"The Waymo AV was traveling southbound on [XXX] in the center lane when the vehicle made contact with a white object that was less than 12" in height and located in the middle of the AVs lane. At the time of the impact, the Waymo AV's Level 4 ADS was engaged in autonomous mode." (February 2024)

"The Waymo AV was traveling south on [XXX] when it approached an automatic gate as it had started to close. The right side of the Waymo AV made contact with the door of the gate. At the time of the impact, the Waymo AV's Level 4 ADS was engaged in autonomous mode. The Waymo AV sustained damage." (February 2024)

"The Waymo AV attempted to enter the apartment complex parking lot entrance as an automatic gate was closing. As the Waymo AV passed the closing gate, the gate made contact with the rear passenger side corner of the Waymo AV. At the time of the impact, the Waymo AVs Level 4 ADS was engaged in autonomous mode. The Waymo AV sustained damag. (January 2024)

"The Waymo AV was traveling northbound on [XXX] in the far right lane when its front passenger side tire made contact with a rock in the road and immediately deflated. At the time of the impact, the Waymo AVs Level 4 ADS was engaged in autonomous mode. The Waymo AV sustained damage. " (January 2024)

Sounds pretty dangerous to me!

 

[iim]

Carry that anti- Elon garbage ars....what a joke this website has become. Political instead of technical...= garbage.

So you just want this site to be strictly focused on the technical and to ignore Elon’s bad behavior?

And if this publication should point out when Elon acts out that is considered by you to be anti-Elon?

 

[Jordan83]

They see Tesla’s FSD, with its capacity to operate in all cities and on all road types, as a more general technology that will soon surpass Waymo.

I sound like a broken record to myself, but I keep coming back to this - why is there no regulatory body signing off on the use of such software on public roads? Why is there not a mandatory minimum standard somewhere that must be met before anyone is free to release their self-driving software on public roadways, and all types of roads?

This isn't even specific to Tesla. I'm talking about for anyone. It just feels crazy to me that we seemingly let any auto company release any version of self-driving software on public roads, and our assurances as the general public that it works well enough is basically the company saying, "trust me, bro."

 

[enlightened.doggo]

I wouldn't expect Tesla to innovate on anything at this point. The business is a one trick pony they basically have one car model which they have optioned out, but their big hit was actually just reinventing gas stations and there are widespread quality issues with the cars themselves.

 

[balthazarr]

Come on now, full self driving is coming out next year.

In all seriousness, I don't know why Tesla hasn't tossed Musk out on his ass with how much he lies to investors.

Two reasons:

1) the Board are his cronies - relations, friends, etc.; and, more importantly
2) his lies have mainly worked to date - vastly inflating the stock value far above what the economic fundamentals suggest should be the value.

The bubble appears to be deflating, though. It will be very interesting - either way - to see the outcome of the vote for his compensation package/the move to Texas.

 

[Ten Wind]

Agree with the first point though a quibble with the second.

Robotaxis scale at the point where they become cheap enough and frictionless enough that it allows semi-urban folks to go from two or three cars to one or two, or urban folks who might have a car for once-a-week trips to drop them altogether. I don't believe we're anywhere near that, but it certainly scales more than the current taxi business.

My back of the envelope math says a taxi driver earns $20/hour and has $0.50/mile in car costs, let's assume 40 miles/hour of driving (which seems high), for $40/hour total expense. If you drop the labor cost there to $1-5/hour and ensure significant reliability ("When I push the button I will get a car within n minutes"), current consumer car behaviors can change.

For the urban people you mention I think car shares are doing a great job of that already where I live. There are EVs parked all over that can be unlocked with a smartphone and used for short or long trips at a reasonable cost. No need for ownership and much much cheaper unless you're driving all the time.

 

[Mechjaz]

[Tesla] will realize that safety

....yeah, maybe not so much. I increasingly think their early, famously high crash test ratings were less intentional than a happy accident of design.

 

[darwiniandude]

Tesla?
Hire??

 

[balthazarr]

Sounds pretty dangerous to me!

Maybe not, but some of the incidents seem pretty concerning - like the hitting a parked car, even when there was a driver behind the wheel. (Concerning in the sense that the technology is not "ready").

One can see why they limit the speed - but I think the incidents just highlight how difficult a problem autonomous driving really is, and how there's much further to go - even for Waymo which is - ahem - streets ahead.

 

[flunk]

Someone on Lemmy posted a list of all the incidents by brand of self driving, and it wasn't even a contest: Cruise, Waymo, etc all came in under 15 or so reported incidents. Tesla had over 200. I'll try to find the report.

You do need to factor in the number of Teslas out there vs Cruise and Waymo, massively more cars would mean massively more incidents. You'd really need the number on indents per road mile travelled to get a good idea of the relative safety of the different systems. I wasn't able to find good sources for this information for all 3.

 

[Tim Lee]

Maybe not, but some of the incidents seem pretty concerning - like the hitting a parked car, even when there was a driver behind the wheel. (Concerning in the sense that the technology is not "ready").

One can see why they limit the speed - but I think the incidents just highlight how difficult a problem autonomous driving really is, and how there's much further to go - even for Waymo which is - ahem - streets ahead.

You have to remember the denominator. Those five incidents correspond to millions of miles of driving—several human lifetimes. If you listed the accidents caused by a typical human driver over their lifetime it would look much worse.

 

[mmiller7]

I don’t see how Tesla’s “camera only” approach ever works. They need more sensors which Musk forced them to take out.

And “robotaxis” is not some liquid gold industry! It is a pretty niche business that “doesn’t scale” because the taxi business just isn’t that big.

I could see where it might scale better if you don't need drivers then maybe its cheaper to offer all the routes that aren't in a city effectively.......but no business will want to do that because the ROI is so much higher if you limit it to in the city.

Not even stuff like Uber/Lyft manages to serve outside the city. And if you do manage to book a reservation for a ride its typically cancelled on you shortly before (probably once the drivers realize it won't be quick to get the next ride).

 

[J.C. Helios]

Could Waymo have a Just Walk Out problem?​

This was running through my mind for most of the article, so I'm glad you addressed it. But I'm still left wondering if Waymo & Tesla are both playing checkers, and the shareholders will eventually get tired of it.

Of course I hope not!

 

[MGAubrey]

Sure, but neither are rail, so...

 

[balthazarr]

You have to remember the denominator. Those five incidents correspond to millions of miles of driving—several human lifetimes. If you listed the accidents caused by a typical human driver over their lifetime it would look much worse.

Fair point. But it seems many of these systems make boneheaded mistakes that virtually no human driver will make - that doesn't engender confidence, and when there are multiple such vehicles around, instead of one or two, how will that scale?

I'm probably sounding super negative - I'm not, really, I would LOVE, LOVE a real Level 5 AV... my life would change in so many ways for the better (as would that of my immediate family)... but, I think we're a long, long way from that, and idiots like Musk announcing robotaxis and FSD any minute now... it's irresponsible and, in Tesla's case, fraudulent that they've been selling vapourware for a decade or so now.

 

[mmiller7]

Absolutely agree. Tim's example of EVs isn't really analogous, because all EVs use high-voltage and batteries - once they learn how to deal with one system, they can deal with them all.

Apparently there are 1800 numbers that LEOs, first responders, roadworkers etc. are "supposed" to call.... yeah, that's workable.

It it totally unreasonable to expect road users to learn 5, 10, 100, however many, systems for communicating/dealing with AVs... there needs to be one standard that all AVs must adhere to. It's safer, it's more workable, and it doesn't expect taxpayer funded workers to take time out to learn umpteen different systems for how to deal with these vehicles.

Its also not always flashing lights on an emergency scene.

I've come onto some that you couldn't really see in advance because it was a large area and there was only 1 cop on-scene so the best they could do was drape some caution tape all the way cross the roadway...they had flares but they were obstructed by a hill and with the poor lighting nobody was seeing the caution tape until it was almost on top of them and then skidding to a stop.

I've also seen times where its an off-duty person responding to an emergency they happened onto and/or other "regular" citizens are the only ones on scene for a while attempting to render aid and maybe a few of them have flashlights to try and catch attention of oncoming traffic and point/wave them to change lanes (I was helping with this once at a night-time rollover crash I had a high powered flashlight on me to illuminate the scene while calling 911 as others kept the occupants talking and laying still, while a few others attempted to help direct traffic)

I've also seen (and called in) disabled vehicles with no working lighting (either battery died or collision destroyed the electrical system) blocking the middle of a thru-lane at night.

These vehicles will have to cope with all that stuff.

 

[ip_what]

Excellent article, thank you! I know that there are some dedicated Tesla fanatics but I would imagine that most Tesla lovers are more realistic and most would probably like Musk to give up his CEO position.

My issue is that for BOTH companies, lack of data is not an issue. Both have available so much data that even adding 10 more years of data really won't improve the models. You have both extremes here with too cautious and too aggressive and I'm not sure either is going to be adequate. Avoiding interstates at year 15 (or whatever it is) of dev is concerning.

I have only one REAL question for Waymo. How many support people do you have for each driverless vehicle? How many employees are going through videos needing to "explain" changes to signs, roads, etc as well as assist the vehicles, fuel, maintain the vehicles, etc? That ratio would tell you everything. If you need 1+ person for every vehicle then this is all academic. You need to get to where you can have 10+ vehicles for every person and get to where you can add more cities quickly without having to do a mass hiring.

Here at Waymo, we’ve reduced labor costs by replacing a $40,000/ year taxi driver with 0.5 data analysts making 30,000/year and a $150,000/year software engineer

 

[citizencoyote]

There is zero chance, absolutely ZERO that Tesla releases a safe, reliable robotaxi on 8 August.

Whatever Musk shows on August 8 (if anything) will be as much a robotaxi as the spandex-clad human was a Tesla robot.

 

[IncorrigibleTroll]

The Amazon robostores weren't good at anything, it seems. Hundreds of cameras and they still had that problem rate? They didn't have a good selection or prices. Even with fewer items than usual, they couldn't automate it?

I also don't know why they persisted on that when rfid already exists and does most of the job.

And I think that's telling for Tesla. Even if the stores were trying something that a human literally doesn't do (spy on shoppers from the ceiling with a high accuracy rate), they threw many more cameras at the problem than Tesla and had a simpler, less dangerous task with more scrutiny for tagging/modeling.

Yeah I’m still stumped as to why they didn’t go with a much simpler and more reliable RFID system for checkout-free purchases.

 

[Tolos]

Thanks for the detailed analysis, the self driving articles are some of my favorite content on ars.

 

[balthazarr]

I'm sorry, but as an engineer comments like "Tesla is several years behind Waymo" are really frustrating and utter rubbish. None of us know, and neither do the companies. They are taking radically different approaches to the problem, which is great. I have stock in both companies because one of them may succeed. Predicting one or the other is just "sports talk" at this point.

The problem is, yes, Waymo has better disengagement numbers in San Fran. But what about Duluth, MN? Or Tulsa, OK? I live in Austin and Cruise served Austin, Waymo is coming. However, I live in the city but 10 miles from downtown and neither company plans to come anywhere near me that I've seen -- Cruise only came out 5 miles. So while Waymo's numbers are much better in small, geographically bounded areas, Tesla's FDS can drive from Austin to Tulsa but not without a human ready to take over. So, Tesla needs more zeros in their reliability, Waymo needs more zeros in their geographic coverge.

Which approach is better? Waymo (and Cruise) depend on extremely detailed maps. Reportedly, they operate as a bit of a simulation where they us AI to map the vehicle and objects against a known, detailed, 3d world. This is a very software/engineering-centric approach and has a lot of benefits when it comes to validation. Tesla has pivoted to a completely neural-net / generative AI based approach and is not dependent on maps and works anywhere that road design and laws are reasonably consistent. Statistical AI approaches are harder to verify, but the real world is rather messy. Can you actually engineer a solution for all traffic conditions in the US? We don't know yet. It's a bit like the difference between Google Assistant or Alexa and ChatGTP. I trust Google Assistant to do math for me more than ChatGTP, but ChatGTP is far better at understanding "real world" text inputs.

Sorry for the negative tone at the start, but this is an absolutely fascinating field with some incredible technology being created. The fact that talented engineering teams are trying different approaches to solve similar problems is exactly what should be happening. Why does it have to be turned into team sports bashing one approach or the other?

The problem with Tesla's approach is the totally arbitrary and unnecessary limitation imposed upon them by Emperor Musk.

It's blatantly obvious that camera-only will never be as effective as a system that has other sensor inputs that work in even zero-visibility.

Why limit yourself to only visual, especially when there are so many situations where visibility is compromised?

Which system would you trust more - Tesla's vision-only needing to tell the difference between a light-grey fog-shrouded blob, and a lighter-grey fog-shrouded blob - or one that has radar and lidar that can tell that the light-grey fog-shrouded blob is actually an oncoming vehicle, and the lighter-grey fog-shrouded blob is an unmoving obstacle on the side of the road?

 

[balthazarr]

Whatever Musk shows on August 8 (if anything) will be as much a robotaxi as the spandex-clad human was a Tesla robot.

He may show nothing if he doesn't get his compensation package passed... apparently he's threatened to pack his bags and go if he doesn't get his way. Toddler tantrums are absolutely the right way to run a multi-billion dollar business, right?

 

[TheNewShiny]

We'll know Tesla is serious about robotaxis when it starts hiring remote operators.

Or rather, we'll know Tesla has fired Elon when it starts hiring remote operators.

 

[Dioptase]

For contrast:

My son is finishing up his 60 hours of driving experience with a learners license that requires a licensed passenger. If he passes the driving exam, he will then be, ahem, FSD.

His is one of the old school LLM AI (Large Language Model Adolescent Intelligence), so adding driving to his locomotion and speech functions has not been without issue. Had to provide negative feedback to the model after it attempted (succeeded?) driving in a ditch and damaging the vehicle.

 

[enlightened.doggo]

I know ARS and it's forum dwellers feel the need to feed on a constant stream of Tesla hate, but this one takes the stupidity taco. Waymo is ahead precisely zero, because it's tech is no more "driverless" than FSD. Moving the operator to a remote location doesn't change the fact that there's a human behind the wheel. The fact that even needs to be mentioned shows how far the author is straining to create this supposed superiority gap. He even mentions, as a throwaway line, that Waymo declined to provide any data on how often its remote human drivers have to intervene. With Tesla you have a mountain of publicly available info, via tons of youtube videos and writeups on social media.

So they both have humans driving, Tesla has some publicly available data on its effectiveness, Waymo has zero public data, author declares Waymo far superior because the human operator isn't in the vehicle which doesn't change the fact that Waymo's system is NOT "driverless".

Imagine lurking for a decade just to make a generic Tesla stock holder comment. You can know Waymo are self driving by just observing them over the years they have been on the road.

 

[sbradford26]

Yeah I’m still stumped as to why they didn’t go with a much simpler and more reliable RFID system for checkout-free purchases.

So there is a case-study from when Walmart tried to go all in on RFID tags back in the early 2000s. The large hurdle was getting suppliers on board with packaging things with RFID tags. I believe another hurdle back in the day when Walmart first tried to push it was that RFID tags were not as cheap as they are now.

For Amazon they are large company but those stores do not move the amount of product that Walmart does. So trying to strong arm suppliers to include RFID tags probably was not an option. So they tried the classic strategy to make up for a lack of appropriate hardware with fancy software. In some situations that can work but clearly not always.

But a classic as technology progresses story it seems like Walmart is trying to push RFID tags again.

Just a note I think RFID tags in products is a great idea, it is just the logistics of getting those tags into all of your products that is a tough hurdle to get over.

 

[pug fugly]

I get that ars loves to crap on musk (it gets oretty old), but an article like this not even mentioning that waymo is under a nhtsa investigation since a week or so ago is a little dishonest

When it looks like a toilet, smells like a toilet, and presents itself as a toilet, it gets crapped on.