Waymo is now exclusive L4 partner for Volvo!

[diplomat33]

This is big news!

"Waymo is now the exclusive global L4 partner for Volvo Car Group, a global leader in automotive safety, including its strategic affiliates Polestar and Lynk & Co. International. Through our strategic partnership, we will first work together to integrate the Waymo Driver into an all-new mobility-focused electric vehicle platform for ride-hailing services. "

Full blog:
Waypoint - The official Waymo blog: Partnering with Volvo Car Group to scale the Waymo Driver

The blog mentions ride-hailing but I think this deal does open the door to Volvo and Polestar to deliver L4 cars to consumers that use the Waymo software. This deal could allow Waymo to really scale up and deliver more FSD to more people.

 

[Yang_H]

Volvo is now owned by a Chinese car company whose owner is directly connected to Mr. Xi in China. and frankly I'd be concerned of the IP safety for Waymo.

 

[Bitdepth]

Things are starting to align. This was what was needed to fully realize this goal, doing piecemeal like a lot of car companies were trying to do was never going to work, at least for the first couple of years. You either integrate from the ground up or you don't do it at all.

Volkswagen going with Mobile Eye
Mercedes group have decided to go with Nvidia hardware and software
GM is doing their own hardware and software
Nissan, Volvo are going with Waymo
Tesla their own hardware and software

 

[Buckminster]

Things are starting to align. This was what was needed to fully realize this goal, doing piecemeal like a lot of car companies were trying to do was never going to work, at least for the first couple of years. You either integrate from the ground up or you don't do it at all.

Volkswagen going with Mobile Eye
Mercedes group have decided to go with Nvidia hardware and software
GM is doing their own hardware and software
Nissan, Volvo are going with Waymo
Tesla their own hardware and software

And BMW doing their own hardware:

And software:

 

[diplomat33]

Things are starting to align. This was what was needed to fully realize this goal, doing piecemeal like a lot of car companies were trying to do was never going to work, at least for the first couple of years. You either integrate from the ground up or you don't do it at all.

Volkswagen going with Mobile Eye
Mercedes group have decided to go with Nvidia hardware and software
GM is doing their own hardware and software
Nissan, Volvo are going with Waymo
Tesla their own hardware and software

Eventually, I think most automakers will have competent FSD. It won't be a matter of FSD or no FSD, but what distinguishes your FSD from the competition. It will be similar to smart phones now. You can get a smart phone anywhere from a variety of different makers and they all offer similar user experiences but some prefer the Apple OS and some prefer the Android OS. Similarly, we will be able to get autonomous driving from any automaker with similar user experiences, but some will prefer the Waymo FSD, others might prefer the Mobileye FSD, others might prefer the Tesla FSD etc... And I do think that eventually it will consolidate into just 2 or 3 main FSD groups.

 

[Pomato]

How much would it cost Tesla to add LIDAR? Maybe even just a basic one to avoid obstacles that the other sensors may have difficulty detecting.

This is just a layman's opinion, but it seems to me Tesla is making things a lot harder on themselves to avoid a potentially life-saving sensor that could cost as little as $100 or few hundred dollars per car. I read a basic LIDAR sensor for autonomous driving has already reached $100 each.

One vehicle fatality lawsuit could cost millions. One million dollars could buy 10,000 of a $100 LIDAR unit.

I can also imagine some regulatory bodies may allow specifically LIDAR based autonomous driving systems. The loss of sales and utility could outweigh the cost.

Wouldn't it be good for Tesla to explore LIDAR, see if they can drive down the manufacturing cost? Who knows, it might save them billions on software development costs.

 

[mikes_fsd]

How much would it cost Tesla to add LIDAR? Maybe even just a basic one to avoid obstacles that the other sensors may have difficulty detecting.

It would cost Tesla a full redesign of the entire FSD/Autopilot hardware suite, power, data, and then adding compute power for LiDAR specific tasks.

And ALL of that would be a complete waste as you must solve computer vision to reach L5 self driving - When you solve computer vision LIDAR is a parasite wasting your resources on the car.

 

[mikes_fsd]

Oh, so @diplomat33 we don't need to have computer vision solved now?
You're going to flip-flop again...?

Please do tell me how a L5 FSD works without computer vision solved? Specifically how they would get through a busy urban environment with traffic lights.

 

[Pomato]

It would cost Tesla a full redesign of the entire FSD/Autopilot hardware suite, power, data, and then adding compute power for LiDAR specific tasks.

And ALL of that would be a complete waste as you must solve computer vision to reach L5 self driving - When you solve computer vision LIDAR is a parasite wasting your resources on the car.

Let's say computer vision is solved. Is it really useless to have LIDAR to confirm the current path is obstacle-free in less than optimal conditions where cameras and radar might have issues? Even if FSD can go millions of miles between accidents, can't there be freak circumstances where LIDAR would help?

 

[mikes_fsd]

Let's say computer vision is solved. Is it really useless to have LIDAR to confirm the current path is obstacle-free in less than optimal conditions where cameras and radar might have issues? Even if FSD can go millions of miles between accidents, can't there be freak circumstances where LIDAR would help?

Possibly, maybe, could be, at what cost?

Waymo - with it's lidar - has promised us L5 FSD for a decade now....
Yet, here we are going into the second half of 2020 and Waymo still has done nothing even resembling what they promised.

How many lives is Tesla FSD going to save with its current sensor suite?

Karpathy said they get dozens of Auto-Emergency Braking events reported daily, where a pedestrian walks out in front of the car in some obscure fashion. That is ~12 ppl/day that were not hurt/killed by a distracted driver.

Humans are terrible at keeping their attention on the road or to their surroundings.

If the LiDAR based FSD is $10k more expensive, how many ppl will NOT purchase it, because of the financial burder? (event at $5k price difference, that is a lot of customers opting out)

 

[diplomat33]

Oh, so @diplomat33 we don't need to have computer vision solved now?
You're going to flip-flop again...?

Please do tell me how a L5 FSD works without computer vision solved? Specifically how they would get through a busy urban environment with traffic lights.

No, I am not going to flip flop. Of course you need computer vision. Don't be silly.

I've never said that you can do L5 without computer vision.

But I don't think you understand what "solved vision" means. Tell me, what does "solved vision" mean? Can camera vision be 100% accurate 100% of the time. How can vision be 100% accurate if the view is obstructed or if bad weather makes it impossible to see? And anything less than 100% accurate 100% of the time will mean that a camera-only FSD car will sometimes hit something.

And even if somehow you do get vision to be 100% accurate 100% of the time, what if a camera breaks? How can a camera-only FSD car continue to drive safely if it loses one or more cameras? It can't!

Waymo has vision for traffic lights, object detection, object classification, lane lines, driveable space and more. Waymo could drive with just camera vision. So why don't they? Why doesn't Waymo ditch lidar? Waymo does not ditch lidar because they understand that having both vision AND radar AND lidar makes for a safer FSD car because the combined sensors will work in more conditions and be more accurate and more reliable.

Mobileye also has vision that can do FSD by itself. But Mobileye plans to still add lidar with vision to achieve the reliability needed for safer FSD.

So it is not a question of just "solving vision". Yes, you need vision. But no single sensor can cover everything perfectly. Vision alone will not be 100% perfect 100% of the time so it is better to have multiple sensors that can work in different weather conditions and can increase the overall reliability if you want a safe and reliable FSD car.

 

[diplomat33]

Waymo - with it's lidar - has promised us L5 FSD for a decade now....
Yet, here we are going into the second half of 2020 and Waymo still has done nothing even resembling what they promised.

What?! I don't know what you were expecting but Waymo has made huge strides towards what they promised. They started with the little firefly prototype car and now they have solved the fundamentals of autonomous driving and they have L4 autonomous driving! They even have actual driverless robotaxis taking passengers all around Chandler, AZ. Tesla does not have that! But you are going to diss them just because they have not deployed L5 to every single city yet? WOW!

And when do you think Tesla is going to achieve L5? Traffic light and stop sign response still requires some confirmations. NOA misses exits. AP cannot make turns at intersections. AP cannot pull over for emergency vehicles. AP cannot drive on residential streets. AP cannot handle roundabouts. etc... Tesla is very far from L5. There is like a million things that Tesla still needs to solve to get to L5. You do realize that L5 means no driver interventions at all? No stalk confirmations ever? No tugs on the wheel to confirm driver attention? L5 car has to be able completely drive itself with no human in the car and handle ALL roads, ALL driving scenarios, ALL the time!

It's great that Tesla's AEB can avoid accidents and hopefully save lives. But you do realize that every modern car on the planet has AEB now and will brake to minimize collisions? Tesla is not unique to have AEB.

All those cases I mentioned, Waymo has already solved. So Waymo might not have delivered L5 yet but they are a lot closer to L5 than Tesla is!

 

[Knightshade]

All those cases I mentioned, Waymo has already solved.

As long as you're in one tiny, heavily HD mapped, suburb in Arizona where the weather is nearly always perfect.

And you have a human remote backup driver to step in since the car STILL needs help sometimes.

So I'm not sure "solved" means what you think it does here

 

[diplomat33]

As long as you're in one tiny, heavily HD mapped, suburb in Arizona where the weather is nearly always perfect.

And you have a human remote backup driver to step in since the car STILL needs help sometimes.

So I'm not sure "solved" means what you think it does here

But geofencing and mapping is part of their solution that made it easier to solve for the small area. The bottom line is that the cars can handle those situations without driver intervention in the small area.

 

[croman]

As long as you're in one tiny, heavily HD mapped, suburb in Arizona where the weather is nearly always perfect.

And you have a human remote backup driver to step in since the car STILL needs help sometimes.

So I'm not sure "solved" means what you think it does here

Waymo has been doing AI vision for longer than Tesla. It took Elon hiring AK to even get this far. Tesla is behind and their only hope at catching up is continuing to gather data through campaigns and heavily curate that data. Waymo, however, has progressed beyond vision and has implemented advanced driver policy and path planning. Tesla has barely started that.

Sensor suite aside, Waymo has accomplished more with their limited trial since it is a completely autonomous system and therefore has proven an ability to complete a variety of tasks Teslas cannot. Anyone that has used FSD (or AP) knows it is limited and dangerous unsupervised no matter the weather or location.

 

[Knightshade]

But geofencing and mapping is part of their solution that made it easier to solve for the small area. The bottom line is that the cars can handle those situations without driver intervention in the small area.

Unless you count the remote backup driver that still has to help sometimes.

But then the autonomy day demo showed Tesla can handle that too as long as you narrow the operational domain enough- without even needing the remote driver.

Waymo has been doing AI vision for longer than Tesla. It took Elon hiring AK to even get this far. Tesla is behind and their only hope at catching up is continuing to gather data through campaigns and heavily curate that data. Waymo, however, has progressed beyond vision and has implemented advanced driver policy and path planning. Tesla has barely started that.

Sensor suite aside, Waymo has accomplished more with their limited trial since it is a completely autonomous system and therefore has proven an ability to complete a variety of tasks Teslas cannot. Anyone that has used FSD (or AP) knows it is limited and dangerous unsupervised no matter the weather or location.

Waymos system works great if you want L4 with human-remote-back in a tiny part of the country after years and years of effort and a super expensive sensor suite.


Teslas system FSDs goal is that it's intended to work everywhere, in an affordable car people can actually buy. They're not trying to "catch up" to Waymo because "working a tiny heavily HD mapped area with perfect weather and still needing remote driver backup" isn't a place Tesla is ever trying to get to.

 

[diplomat33]

Unless you count the remote backup driver that still has to help sometimes.

But then the autonomy day demo showed Tesla can handle that too as long as you narrow the operational domain enough- without even needing the remote driver.

Waymos system works great if you want L4 with human-remote-back in a tiny part of the country after years and years of effort and a super expensive sensor suite.

Teslas system FSDs goal is that it's intended to work everywhere, in an affordable car people can actually buy. They're not trying to "catch up" to Waymo because "working a tiny heavily HD mapped area with perfect weather and still needing remote driver backup" isn't a place Tesla is ever trying to get to.

Yes, Tesla is trying to catch up because Tesla is still working on basic features like like traffic light recognition, speed limit recognition and making turns at intersections which Waymo has already finished. Waymo might be geofenced to a small area but the cars have all the features for L5. According to Waymo CTO, Waymo has solved all the fundamentals of FSD, they are moving on to more complex planning problems. They are just geofenced to make sure that the FSD works right in a small area first before expanding to more areas. L4 is a stepping stone to L5.

 

[Watts_Up]

Things are starting to align. This was what was needed to fully realize this goal, doing piecemeal like a lot of car companies were trying to do was never going to work, at least for the first couple of years. You either integrate from the ground up or you don't do it at all.

Volkswagen going with Mobile Eye
Mercedes group have decided to go with Nvidia hardware and software
GM is doing their own hardware and software
Nissan, Volvo are going with Waymo
Tesla their own hardware and software

Toyota ?????​

 

[Mo City]

Eventually, I think most automakers will have competent FSD. It won't be a matter of FSD or no FSD, but what distinguishes your FSD from the competition. It will be similar to smart phones now. You can get a smart phone anywhere from a variety of different makers and they all offer similar user experiences but some prefer the Apple OS and some prefer the Android OS. Similarly, we will be able to get autonomous driving from any automaker with similar user experiences, but some will prefer the Waymo FSD, others might prefer the Mobileye FSD, others might prefer the Tesla FSD etc... And I do think that eventually it will consolidate into just 2 or 3 main FSD groups.

Agree with you on consolidation but the big hurdle for everyone but Tesla remains mass distribution, even with these partnerships.

I doubt it happens, but if Tesla delivers legit FSD with the current sensor suite in 2-4 years, it will be a massacre. Otherwise, the pursuit of widespread FSD will remain a very frustrating affair.

 

[Knightshade]

Yes, Tesla is trying to catch up because Tesla is still working on basic features like like traffic light recognition, speed limit recognition and making turns at intersections which Waymo has already finished

Except, they haven't finished those.

If they did they wouldn't have to restrict them to a tiny suburb in Arizona with ideal weather, and still need backup remote drivers even then.

You keep ascribing far greater "finished" and "solved" claims to Waymo than there's any evidence to support.

. Waymo might be geofenced to a small area but the cars have all the features for L5.

If that were true- which it's obviously not- they wouldn't need to be geofenced.

According to Waymo CTO, Waymo has solved all the fundamentals of FSD, they are moving on to more complex planning problems.

If we're gonna just quote CEO claims directly contradicted by facts on the ground we'll be here for weeks citing stuff Elon said- don't think you wanna go down that rabbit hole

They are just geofenced to make sure that the FSD works right in a small area first before expanding to more areas.

And Tesla is still asking for confirmation on stoplights and signs to make sure it works right in ALL areas before removing confirmations.

You don't get to make excuses for why they don't really have anything "solved" unless everyone gets to.

L4 is a stepping stone to L5.

All levels are stepping stones to L5 depending on your approach.