What is your prediction re Waymo/Lidar?

[dqd88]

If Elon is right, and Waymo (et al) scraps Lidar at some point after they realize it's too expensive, etc...
What happens then? Do they write-off all that Cap-ex that went into Lidar? Do they start from the ground up with Vision? Do they call it quits? Do they try to acquire Tesla for autonomy? Or make a big investment into Tesla?

If Elon is right, it seems that Tesla may in a few years have a virtual monopoly on autonomy. But that sounds too good to be true. Thoughts? I feel like I'm missing something.

 

[heltok]

Rumor is that Alphabet have been developing their in house solid state lidar. I would guess that they at scale can get costs of this one down to $100-1000. Don’t think the costs of it will make or break their viability. The bigger question is how they will get access to enough long range electrical vehicles, how long time it will take for them to integrate their system into the vehicles and how much data they need to get a good enough system.

It should be noted that Google has vision also and the extra lidar data can be very useful to train vision even if they decide to scrap lidar. Maybe they have wasted some time, but they had plenty of time headstart and they have a lot of talented people working for them also.

 

[diplomat33]

I don't think Waymo will scrap LIDAR. They have an excellent hardware/software combo that works for what they are interested in: local robotaxis. So I think Waymo will continue to be a local autonomous taxi service for awhile. They will expand to more cities over time, eventually get rid of the safety drivers entirely but remain a robotaxi service of a few thousand cars. The cost of the hardware will limit them but they will be ok with it because robotaxis is their business model. Eventually, Tesla will get to autonomous driving with vision only and get it to all their cars. So Tesla will have a much larger fleet of autonomous cars.

Basically, I think Waymo will "win" the race to local robotaxis but Tesla will eventually "win" the race to mass market, cheap, autonomous driving. What about the big auto makers or Mobileye? Won't they win the race to mass market, cheap, autonomous driving before Tesla? Normally, yes, they would. But I think their big problem is their total lack of boldness. We are already seeing that playing out now. GM could have pushed Super Cruise further and had L3 highway and totally crushed Nav on AP. But instead, GM put SC on one luxury car and then cancels it. And we've seen auto makers put expensive LIDAR and hardware on a few test cars and test them over and over again but since the cars are not reliable yet as L4 autonomous, the auto makers are holding back and not releasing much to their commercial cars. Mobileye has great hardware and software but the auto makers are again unwilling to use it. Basically, I think the big auto makers will squander their lead and go in circles while Tesla will push ahead. It's the turtle and the hare paradox. The big auto makers are the hare that gets ahead but then goofs around so much that the turtle (Tesla) which is slow but consistent, eventually gets ahead and wins.

 

[acoste]

Even with all the progress in radar development car makers are still choosing lidar.

For example Audi late last year: Audi’s ATD Teams Up With Luminar

Or these guys:
"Over the last 18 months, Luminar has scored partnerships with Toyota, Volkswagen, and Volvo."
How 10 leading companies are trying to make powerful, low-cost lidar

Either they don't want to take any risk or they see the prices coming down.

"The bottom line is that while bringing lidar costs down will take a significant amount of difficult engineering work, there don't seem to be any fundamental barriers to bringing the cost of high-quality lidar down below $1,000—and eventually below $100."
Why experts believe cheaper, better lidar is right around the corner

 

[electronblue]

I don't think Waymo will scrap LIDAR. They have an excellent hardware/software combo that works for what they are interested in: local robotaxis.

I’m not sure that is the only thing they are interested in. It is where they start.

I would assume the licensing opportunities for a full self-driving suite are enormous, as with MobilEye, once they truly get there. Also heavy transport. How much of that eventually will use sensors and of what type remains an interesting and open question.

I agree with lack of boldness from the established car makers. It was the same story with BEVs, where Tesla obviously showed the way successfully. I just wonder if with FSD Musk made an uncharacteristic mistake and called it too soon with the AP2-3 suite.

 

[Eno Deb]

If Elon is right, and Waymo (et al) scraps Lidar at some point after they realize it's too expensive, etc...
What happens then? Do they write-off all that Cap-ex that went into Lidar? Do they start from the ground up with Vision? Do they call it quits? Do they try to acquire Tesla for autonomy? Or make a big investment into Tesla?

What a weird thing to say. Waymo has an excellent computer vision system that has been doing things for years that Tesla has yet to make work reliably (e.g. traffic light and sign recognition).

 

[electronblue]

What a weird thing to say. Waymo has an excellent computer vision system that has been doing things for years that Tesla has yet to make work reliably (e.g. traffic light and sign recognition).

Thank you, you voice of reason.

 

[Eno Deb]

Basically, I think Waymo will "win" the race to local robotaxis but Tesla will eventually "win" the race to mass market, cheap, autonomous driving. What about the big auto makers or Mobileye? Won't they win the race to mass market, cheap, autonomous driving before Tesla? Normally, yes, they would. But I think their big problem is their total lack of boldness.

Maybe. On the other hand, MobilEye is making quite a lot of money on their ADAS systems while they keep working on autonomous driving (see latest quarterly report by Intel), while Tesla is bleeding money on all fronts.

 

[electronblue]

Maybe. On the other hand, MobilEye is making quite a lot of money on their ADAS systems while they keep working on autonomous driving (see latest quarterly report by Intel), while Tesla is bleeding money on all fronts.

That would be endangered if MobilEye’s customers really become ”horse” makers as Elon Musk suggested — with Tesla’s impending Level 5 feature complete at the end of 2019 — so there certainly is some reason to pick up pace... If you believe Musk of course.

I’m not sure MobilEye or even the traditional automakers are as passive as some suggest but some concern to that effect is reasonable especally with the latter group.

 

[Eno Deb]

That would be endangered if MobilEye’s customers really become ”horse” makers as Elon Musk suggested — with Tesla’s impending Level 5 feature complete at the end of 2019 — so there certainly is some reason to pick up pace... If you believe Musk of course.

I’m not sure MobilEye or even the traditional automakers are as passive as some suggest

They aren't. The automakers are fully aware that the combination of networked car sharing services and full autonomy is a mortal threat to their current business model.

 

[dqd88]

What a weird thing to say. Waymo has an excellent computer vision system that has been doing things for years that Tesla has yet to make work reliably (e.g. traffic light and sign recognition).

What's a weird thing to say? I'm just going off of what Elon said on Autonomy Day. He was asked a question from an analyst along the same lines... ie- what will his competitors do regarding Lidar and he said that they will probably scrap it at some point. So I'm wondering if that does in fact happen... what happens next, ie- do they all switch to vision...

 

[Eno Deb]

So I'm wondering if that does in fact happen... what happens next, ie- do they all switch to vision...

It's not an either/or. Every single autonomous vehicle out there already uses computer vision.

 

[electronblue]

It's not an either/or. Every single autonomous vehicle out there already uses computer vision.

Yes indeed. That is something often forgotten: everyone already uses vision. I guess the more realistic point is that if Waymo relies on a much more expensive sensor suite (as they do even with their high-quality cameras), and someone makes it with a bunch of ”mobile phone” cameras instead, that could hinder the one with the more expensive suite. But then there is no proof of this either. Waymo has a careful approach for sure but any realistic comparison of their and Tesla’s abilities and trajectories of their abilities is hard from the outside, other than noting Tesla’s approach is certainly the less proven one so far.

The issues Waymo faces are not ones of perception anyway as far as we can tell. They already create a perfect 3D view of the world with their suite, which to drive in. I would expect they would fare just as well in any neighbourhood if they dropped their geofencing and we know they successfully test in winter conditioins too.

They seem to be working with the subtler cues of the world and the driving policy now, not sensing.

 

[renim]

why is LIDAR considered a more expensive sensor suite? there seems a real track This compact and cheap lidar could steer small autonomous vehicles to get LIDAR down to $10. or to put it another way, to reduce the price of LIDAR down until its just another fashion accessory in the vehicles head lights assembly. today it might be $700,in a few years it will be $200-$250.

so long term, it is LIDAR for all, even if the car is not autonomous.

 

[electronblue]

why is LIDAR considered a more expensive sensor suite? there seems a real track This compact and cheap lidar could steer small autonomous vehicles to get LIDAR down to $10. or to put it another way, to reduce the price of LIDAR down until its just another fashion accessory in the vehicles head lights assembly. today it might be $700,in a few years it will be $200-$250.

so long term, it is LIDAR for all, even if the car is not autonomous.

If you refer to me, I mean Waymo’s sensor suite overall is more expensive. Their cameras alone are much more numerous and higher quality and more expensive than those of Teslas and they have several Lidars and more radars around the car.

So the amount of sensors alone means they are more expensive.

Of course nothing suggests Waymo couldn't operate with lesser or cheaper sensors if they wanted to. They have simply settled on this suite. It will get cheaper over time but their sensor suite amibitions still seem much higher quality and thus more expensive than those of Tesla’s.

 

[renim]

Ok, i will speak plainly to you.

Waymo is looking at taxi, maybe 150,000 km per year, vs tesla looking at personal car 15,000 km per year. The taxi had better not be same cost/class sensor suite as personal vehicle. It must be better or at least more expensive.

Sooner or later the upper management will need to front before a jury and explain their decisions.

 

[electronblue]

Ok, i will speak plainly to you.

Waymo is looking at taxi, maybe 150,000 km per year, vs tesla looking at personal car 15,000 km per year. The taxi had better not be same cost/class sensor suite as personal vehicle. It must be better or at least more expensive.

Sooner or later the upper management will need to front before a jury and explain their decisions.

I agree that is Waymo’s thinking and I agree their suite looks quite robust and includes multiple redundancies.

That said isn’t Tesla also aiming for a robotaxi fleet with their sensor suite. A very different kind of sensor suite with much less cameras and redundant sensors than Waymo but they say they will be operating the Tesla Network as a robotaxi in 2020 somewhere (presumably in the U.S.) with that suite.

I guess it is somewhat ironic that people always talk of Lidar when they talk of Waymo compared to Tesla but in actual fact a Waymo has more cameras than any Tesla has.

 

[CarlK]

If Elon and Karpathy, and the ex-Google guy Levandowski, are right the issue is not just Lidar but big data for NN training. Again if what Elon said was right simulatoin would not work. Not sure how Waymo could solve that issue. It's not easy to make an about face when you're this deep into a program. They could put all blame to Levandowski though.

Did anyone notice those Lidar people are awefully quite on this? Not even one Lidar guy has something to say about this? How could you not clarify when the technology you are relying on to attract customers and investors got publicly denounced like this? Even Nvidia came out did a weak protest on the chip side when they were not that badly depcited.

 

[electronblue]

Did anyone notice those Lidar people are awefully quite on this? Not even one Lidar guy has something to say about this? How could you not clarify when the technology you are relying on to attract customers and investors got publicly denounced like this? Even Nvidia came out did a weak protest on the chip side when they were not that badly depcited.

Quite likely because very few people in the industry take Elon seriously on FSD. If you don’t take it seriously why say anything, you keep developing your thing. He will need to show much more to change that perception created by the past.