Waymo’s “commercial” ride-hailing service is... not yet what I hoped

[strangecosmos]

It took me a while to get around to watching this video in full. I think everyone who has an interest in autonomous vehicles should watch it. The dash cam footage gives a reality check on Waymo’s current capabilities:

You can’t just watch marketing videos and executive speeches and use that to draw your conclusions about the current state of the technology. We need better evidence than that.

I think all the problems shown in the video are solvable, and they may even be solved in a short timeframe. That would be great.

But we need to dispel the notion that Waymo’s vehicles can drive 5,000 miles without making a mistake. We’ve all been misled by the California DMV disengagements data because the Cali DMV rules don’t actually require all disengagements to be reported. Reporters, technologists, analysts — everyone — have been misled. It’s frustrating that this misapprehension went on for so long, and it’s sad that the driving skills of Waymo, Cruise, et al. aren’t as advanced as we thought.

It also seems unlikely that Waymo’s vehicles are ready to launch with zero human monitoring or intervention by the end of 2018, even in tightly geofenced areas. In fact, we know that Waymo will continue to have people either in the car or operating it remotely. And the dash cam videos help show why.

 

[electronblue]

But we need to dispel the notion that Waymo’s vehicles can drive 5,000 miles without making a mistake.

Nobody that I can see here is saying that, so this to me is a strawman argument.

What is the question is: which company is the most advanced. One measure is Waymo taking on third party customers even without safety drivers on board. That is a world first as far as we can say. Of course it can still make some mistakes but to establish leadership position isn’t about being mistake free it is about being better than anyone else at this time. One would have to ignore awful lot of other evidence to suggest Waymo is not first at this time.

If someone is a close second at this time it surely is not Tesla. It could be MobilEye with their Israeli taxi car.

 

[CarlK]

Waymo, and pretty much everyone else, are only running their fleets in certain small mapped areas with the goal of community transportation or ride hailing services. Only Tesla is working on "general autonomous driving" that it will eventually let you to buy a car that you can have it to drive itself anywhere in the country. No one else is even attempting to do that.

But we need to dispel the notion that Waymo’s vehicles can drive 5,000 miles without making a mistake. We’ve all been misled by the California DMV disengagements data because the Cali DMV rules don’t actually require all disengagements to be reported. Reporters, technologists, analysts — everyone — have been misled. It’s frustrating that this misapprehension went on for so long, and it’s sad that the driving skills of Waymo, Cruise, et al. aren’t as advanced as we thought.

You just need to take the DMV data with a huge grain of salt. It can be easily gamed. Just run the same mile over and over again you will get you car to have close to perfect record. Probably not Waymo but I would think some new companies will try to do this to gain credibility.

 

[electronblue]

Waymo, and pretty much everyone else, are only running their fleets in certain small mapped areas with the goal of community transportation or ride hailing services. Only Tesla is working on "general autonomous driving" that it will eventually let you to buy a car that you can have it to drive itself anywhere in the country. No one else is even attempting to do that.

I know Elon Musk says that but I’m not sure if it is true. I get the theory but I’m not sure if it is at all given. Many companies are aiming for Level 4 products around the world and certainly MobilEye with EyeQ4 is already doing generic HD mapping at scale from consumer cars.

Many autonomous players may start smaller in the sense that initial products are map constrained for practical reasons but most are aiming at a generic outcome. Remains to be seen if Tesla is in the end any faster to a generic car responsible driving product than these other roadmaps.

After all even Navigate on Autopilot is constrained to one country at first and it is not car responsible driving yet which would likely start even smaller due to state legislations and so forth.

It is that car responsible driving that is the tricky bit. A driver’s aid with driver responsible is another thing.

 

[CarlK]

I know Elon Musk says that but I’m not sure if it is true. I get the theory but I’m not sure if it is at all given. Many companies are aiming for Level 4 products around the world and certainly MobilEye with EyeQ4 is already doing generic HD mapping at scale from consumer cars.

Many autonomous players may start smaller in the sense that initial products are map constrained for practical reasons but most are aiming at a generic outcome. Remains to be seen if Tesla is in the end any faster to a generic car responsible driving product than these other roadmaps.

After all even Navigate on Autopilot is constrained to one country at first and it is not car responsible driving yet which would likely start even smaller due to state legislations and so forth.

It is that car responsible driving that is the tricky bit. A driver’s aid with driver responsible is another thing.

It's not just Elon says that. It is true by just looking at how it's set up. Did Tesla refuse to sell anyone FSD because of the location or intended use? What Mobileye or anyone else, including Waymo, is doing is NOT general autonomous driving. Your "general" drivers license should allow you to drive a car anywhere on any road that allows a car.

 

[Bladerskb]

It's not just Elon says that. It is true by just looking at how it's set up. Did Tesla refuse to sell anyone FSD because of the location or intended use? What Mobileye or anyone else is doing is NOT general autonomous driving. Your "general" drivers license should allow you to drive a car anywhere on any road.

So because Tesla sells something which does not exist (which is fraud) makes them better?

 

[CarlK]

So because Tesla sells something which does not exist (which is fraud) makes them better?

Cut this Tesla bashing off will you? Tesla is geared toward getting there (general autonomous driving) but no one else is. That was the point.

What your beloved Mobileye had been saying and doing, and its spin of the lacking of NN machine learning technology, are just laughable. It could only fool a few non-technical and gullible people by trying to confuse driver assistance to autonomous driving capabilities.

 

[Bladerskb]

If/when Traffic Jam Pilot is activated in Germany, it will only work on divided highways at speeds up to 60 km/h (37 mp/h), and will merely follow the car ahead of it, just like Traffic Aware Cruise Control (TACC). It will not automatically change lanes, or do anything more advanced than simply drive forward within a single lane at speeds up to 60 km/h.

So, is Traffic Jam Pilot more advanced than Navigate on Autopilot? Apparently different people have different opinions on this. I find automatic lane changes more impressive than slow TACC, even if the driver’s eyes are off the road.

Here's two close call under 37 MPH on divided highway (v9) and you can find dozens of these incidents on youtube and reddit.
Calling any "L3 Traffic Jam" system simply TACC and just following the car in-front is as disingenuous as it gets. But this is par for the course coming from you. I don't understand how ppl take you seriously anymore.

Close call with Autopilot today. This is why you should always pay attention! : teslamotors

Another close call with Autopilot today - merging truck not recognized : teslamotors

 

[electronblue]

It's not just Elon says that. It is true by just looking at how it's set up. Did Tesla refuse to sell anyone FSD because of the location or intended use? What Mobileye or anyone else, including Waymo, is doing is NOT general autonomous driving. Your "general" drivers license should allow you to drive a car anywhere on any road that allows a car.

I know what they sold to me too I am just not yet convinced they will deliver it any faster than others in a form where car would be responsible for the drive.

I do get Tesla’s approach is different and can surely result in a more versatile driver’s aid being available faster because others are focused on car responsible driving and doing that by piecemeal because of liability and legislation and all that. This will surely make for an interesting driver’s aid for us in the meanwhile.

But we have pretty much no visibility into when and how Tesla might offer Level 4+ car responsible drive or how global that might be when they do compared to others at that time.

 

[Bladerskb]

Tesla is geared toward getting there (general autonomous driving) but no one else is. That was the point.

"these cars should be able to go anywhere on the highway system"

Elon's general "highway system" is NAP and we know how well that works right?

So when Elon says a general full self driving is coming in 2019 we also know how well it will work.

"You know, I think we’ll get to full self-driving next year. As a generalized solution, I think. But that’s a ... Like, we’re on track to do that next year. So I don’t know. I don’t think anyone else is on track to do it next year."

NAP on highway sucks therefore NAP on surface street will suck too.
its not rocket science, its 1+1.

 

[acoste]

Tesla needs to change their mentality if they want to deliver some reliable product.
A quote from Glassdoor:
"The only con I can perceive is that Tesla is a high stress, high pace environment. "
People make mistakes under high stress. It also kills creativity. And the target is one of the toughest for the current technology.

When I was working in a young design team, we made aggressive targets on deadlines, worked our ass off and ended up with a demo that had tons of bugs. Later in my career I joined another and mature team, that had a low turnover rate, a high coverage verification process in place with proper documentation practices and now although it takes more time to get to a demo, than with the previous team, however we rarely have to do anything twice (or even more) and the product comes out sooner.
I'm glad for our quality checks in place. Even with this it is a bit stressful when it is released to the public. I can imagine what developers at Tesla feel when they release another software update with no time to check every function.
Many people don't understand that frequent updates are dangerous. This is not a cellphone app.

Watch out for the slow paced "sleeper" companies with less publicity.

 

[CarlK]

"these cars should be able to go anywhere on the highway system"

Elon's general "highway system" is NAP and we know how well that works right?

So when Elon says a general full self driving is coming in 2019 we also know how well it will work.

"You know, I think we’ll get to full self-driving next year. As a generalized solution, I think. But that’s a ... Like, we’re on track to do that next year. So I don’t know. I don’t think anyone else is on track to do it next year."

NAP on highway sucks therefore NAP on surface street will suck too.
its not rocket science, its 1+1.

Regardless if it's 2019 or 2021 Tesla is the only one that has a system set up to get there. On the other hand while giving it undeserved high praises you keep invading the question what Mobileye has now that could lead it to there is a year, two years or even five years. I asked you to provide evidences that Mobileye even had anything on NN machine learning before 2016 (they did not) but all you could do is keep on telling how good its vision chips will be. Vision chip is not machine learning and will NEVER get us to full self driving capability. One company might be slow in delivery but the other does not even have a valid plan to deliver. No those vision chips are not valid plans.

 

[fmonera]

If someone is a close second at this time it surely is not Tesla. It could be MobilEye with their Israeli taxi car.

As far as I know, the Israeli taxi car you talk about doesn't exist at this time. So if someone is a close second at this time it surely is not MobilEye. When the Israeli taxi car fleet is put in production and takes customers without human drivers, then we will see where is Waymo, Tesla and the rest.

 

[electronblue]

As far as I know, the Israeli taxi car you talk about doesn't exist at this time. So if someone is a close second at this time it surely is not MobilEye. When the Israeli taxi car fleet is put in production and takes customers without human drivers, then we will see where is Waymo, Tesla and the rest.

Yes it was a forward looking statement on my part. The car exists I believe but certainly not the fleet or the passenger hauling.

I acknowledge Waymo as the current autonomous leader with taking third party passengers. I do not think anyone else does that yet. MobilEye might be number two in the world (in the future) possibly which is what I was trying to say.

 

[CarlK]

As far as I know, the Israeli taxi car you talk about doesn't exist at this time. So if someone is a close second at this time it surely is not MobilEye. When the Israeli taxi car fleet is put in production and takes customers without human drivers, then we will see where is Waymo, Tesla and the rest.

Then again being able to run a taxi service in a small designation area is far from being able to sell you a car that you can use it anywhere in the world. Tesla might appear to be moving slowly but it's only because it's bar is set much higher than the rest. You might argue the bar is set too high but it will be much easier for Tesla just go back to make its car to work only in a small area (such as the Silicon Valley that has a lot of Tesla) than the other way around.

 

[electronblue]

Then again being able to run a taxi service in a small designation area is far from being able to sell you a car that you can use it anywhere in the world. Tesla might appear to be moving slowly but it's only because it's bar is set much higher than the rest. You might argue the bar is set too high but it will be much easier for Tesla just go back to make its car to work only in a small area (such as the Silicon Valley that has a lot of Tesla) than the other way around.

It is also true that piloting a globally ready system in a small area can still make sense — especially for liability and legal reasons with third party passengers. It is in itself not evidence of a system being locally usable only. It might simply be a logistical and legal choice.

Similarly aiming for a global solution does not guarantee the solution works even locally at all. It might not work well anywhere so it is not evidence of capability either.

I do think this comes back to whose strategy and approach we believe in. If Tesla started a small taxi service somewhere I am sure many would see that as a positive step and not indicative of their lack of general applicability.

In fact I fully expect Tesla to geofence their autonomous features too to enable a controlled ramp-up. Navigate on Autopilot is the first major step in that direction by leaving out rest of the world.

 

[CarlK]

It is also true that piloting a globally ready system in a small area can still make sense — especially for liability and legal reasons with third party passengers. It is in itself not evidence of a system being locally usable only. It might simply be a logistical and legal choice.

Similarly aiming for a global solution does not guarantee the solution works even locally at all. It might not work well anywhere so it is not evidence of capability either.

I do think this comes back to whose strategy and approach we believe in. If Tesla started a small taxi service somewhere I am sure many would see that as a positive step and not indicative of their lack of general applicability.

In fact I fully expect Tesla to geofence their autonomous features too to enable a controlled ramp-up. Navigate on Autopilot is the first major step in that direction by leaving out rest of the world.

Google started its self driving car development as a moonshot project with no particular object in mind but just to see where it will lead to. That's part of the reason why it chose to work with expensive Lidar and small number of test cars. Eventually it realized the ride hailing potential and formed Waymo. Although it (and Apple too) knew that approach will not lead to general autonomous cars and they did not even attempt to try that. Most others were just following the Waymo lead with or without the understanding of what it's good for or not good for.

Tesla on the other hand has always had general autonomous driving in mind. It of course could start with taxi services in small areas and get there sooner but that just does not fit the nature of its business. Not to mention it is positioned better than anyone else, Waymo, Apple, Nvidia or Mobilieye, to implement general autonomous driving. After all its business is to sell cars not just to sell services even though the later is also in the business plan now. That's why it started with cameras, which by that time is already very capable, instead of Lidar and put the system in every car whether you purchase the option or not. It's not saying it will not implement geofencing so it could get to some areas sooner but the advantage of able to have general autonomous driving technology will always be there. Agree with you we could very well see that happen for competitive reasons.

 

[kbM3]

I think this statement have been somewhat debunked already so lets not bring it in here.

Then let’s leave Mobileye and Tesla out of this thread also. I started commenting because it is completely disingenuous to compare anyone else’s future tech and demos... to Tesla’s existing system that’s for sale.

 

[verygreen]

Then let’s leave Mobileye and Tesla out of this thread also. I started commenting because it is completely disingenuous to compare anyone else’s future tech and demo’s... to Tesla’s existing system that’s for sale.

Well, Tesla certainly has a demo as well, even though it's already somewhat dated.

 

[kbM3]

Well, Tesla certainly has a demo as well, even though it's already somewhat dated.

Are you claiming that’s what people are doing? Comparing Tesla’s demo to others? That is not what I’ve observed.