The customer is always wrong: Tesla lets out self-driving car data when it suits(theguardian.com)
theguardian.com
The customer is always wrong: Tesla lets out self-driving car data when it suits
https://www.theguardian.com/technology/2017/apr/03/the-customer-is-always-wrong-tesla-lets-out-self-driving-car-data-when-it-suits
55 comments
While I agree that Waymo is in the lead, I disagree they're two orders of magnitude ahead of the competition. If you look at Cruise, they are clearly making significant progress each month. If you look at their last month available, they have 6 disengagements for 2284.35 miles, or 2.62 disengagements per 1,000 miles.
Waymo reports 0.2 disengagements per 1,000 miles, which is clearly better - but only by 1 order of magnitude. And if you look at how quickly Cruise is making progress, I'd suspect (by as much data as we can glean from these reports, which is limited) that Cruise will be roughly on par with Waymo in <2 years.
Waymo reports 0.2 disengagements per 1,000 miles, which is clearly better - but only by 1 order of magnitude. And if you look at how quickly Cruise is making progress, I'd suspect (by as much data as we can glean from these reports, which is limited) that Cruise will be roughly on par with Waymo in <2 years.
> And if you look at how quickly Cruise is making progress, I'd suspect (by as much data as we can glean from these reports, which is limited) that Cruise will be roughly on par with Waymo in <2 years.
Assuming Waymo is standing still. The last 20% take a lot longer than initial 80% in the machine learning business. The last 2% literally takes a decade.
Assuming Waymo is standing still. The last 20% take a lot longer than initial 80% in the machine learning business. The last 2% literally takes a decade.
If Waymo stands still, I suspect Cruise will be there in ~1 year. And right now Waymo is in the throes of a managerial clusterfuck, which is why most people left.
What does this have to do with the article?
>Tesla said it doesn’t show anyone the logs themselves, only a description of the logs that it distributes to media when it feels the integrity of its product has been impugned.
>“Drivers don’t have access to this data to defend themselves if they need it,” the Swiss driver said. “So this data is 100% to the disadvantage of the drivers.”
This is scary stuff and is one of the main things that spooks me about Elon's view of the future. He's pretty much the anti-Stallman. My devices log everything I do in a blackbox fashion and when I have a conflict with these companies they can use this logging against me. I can't get these logs and even if I could how could I properly interpret them for my defense. Look at how Toyota had teams of experts from all over the world going over their recent acceleration issues and still had a hard time making heads or tails of it.
How can a consumer like myself ever hope to have these kinds of resources if there's a conflict?
>The Guardian could not find a single case in which Tesla had sought the permission of a customer who had been involved in an accident before sharing detailed information from the customer’s car with the press when its self-driving software was called into question.
This is inexcusable. Executitves at Tesla shouldn't be making these calls, the consumer should.
I think IoT and smarteverything is going to have us re-evaluate our relationships with the companies we buy from . There's a fairly major power inbalance in their favor and they are not shy about using it.
>“Drivers don’t have access to this data to defend themselves if they need it,” the Swiss driver said. “So this data is 100% to the disadvantage of the drivers.”
This is scary stuff and is one of the main things that spooks me about Elon's view of the future. He's pretty much the anti-Stallman. My devices log everything I do in a blackbox fashion and when I have a conflict with these companies they can use this logging against me. I can't get these logs and even if I could how could I properly interpret them for my defense. Look at how Toyota had teams of experts from all over the world going over their recent acceleration issues and still had a hard time making heads or tails of it.
How can a consumer like myself ever hope to have these kinds of resources if there's a conflict?
>The Guardian could not find a single case in which Tesla had sought the permission of a customer who had been involved in an accident before sharing detailed information from the customer’s car with the press when its self-driving software was called into question.
This is inexcusable. Executitves at Tesla shouldn't be making these calls, the consumer should.
I think IoT and smarteverything is going to have us re-evaluate our relationships with the companies we buy from . There's a fairly major power inbalance in their favor and they are not shy about using it.
Personally, I'm okay with this in the short term. Electric vehicles are still at a fairly press-sensitive stage of development. And there's definitely skulldoggery that ICE competitors could manufacture if they wanted to.
Or even crazy customers. See: https://www.tesla.com/blog/when-life-gives-you-lemons
After they're accepted and the new normal? Absolutely agreed that the balance should be shifted toward the customer.
Or even crazy customers. See: https://www.tesla.com/blog/when-life-gives-you-lemons
After they're accepted and the new normal? Absolutely agreed that the balance should be shifted toward the customer.
Wow.
A friend of mine in china told me "What do we need civil rights for right now. It's more important that we become a world leader. We can always get our rights back later"
Once you set a new normal, it's the new normal, and people get used to it. Remember that time we went back to a strong 4th amendment? Or less surveillance? or ...
If electric cars can't survive with negative press, they shouldn't survive. Ends don't justify means.
Once you set a new normal, it's the new normal, and people get used to it. Remember that time we went back to a strong 4th amendment? Or less surveillance? or ...
If electric cars can't survive with negative press, they shouldn't survive. Ends don't justify means.
Remember the time the United States was doing terrible damage to the environment (noon in Pittsburgh, strip mining, coal ash in the rivers), then we became a world superpower, then the EPA was given teeth to clean up the environment?
History suggests your friend might have a point. Might. There are likely good counter-examples to consider.
History suggests your friend might have a point. Might. There are likely good counter-examples to consider.
This is the usual argument.
Here's the problem: There's always another thing.
always.
In the history of the world, we go from thing to thing to thing, always teetering on the edge.
So if you always say "hey, just doin it for this crisis, we'll go back to normal later", you'll be doing it forever.
EIther we make it without that, or truthfully, we aren't goign to make it as a species over time.
So if you always say "hey, just doin it for this crisis, we'll go back to normal later", you'll be doing it forever.
EIther we make it without that, or truthfully, we aren't goign to make it as a species over time.
How would we have industrialized without massive amounts of coal and oil-fueled power?
Sometimes there are stages you must go through in order to reach the other side (e.g. an excess energy future where we can afford to research and pump out solar panels).
Sometimes there are stages you must go through in order to reach the other side (e.g. an excess energy future where we can afford to research and pump out solar panels).
Civil rights and pollution have almost nothing in common, especially when we're talking about an oppressive autocracy that has huge incentives to keep civil rights down.
Worse, the EPA exists because of grassroots democratic action. What grassroots right do these Chinese people have? Much, much less than Americans in the 60s.
Worse, the EPA exists because of grassroots democratic action. What grassroots right do these Chinese people have? Much, much less than Americans in the 60s.
While I agree civil rights and pollution have almost nothing in common, civil rights and consumer rights are also pretty unrelated. The larger umbrella category encompassing them is "Pro-individual vs. pro-collective policy," which I assumed we were discoursing under because you brought up civil rights in the context of consumer rights.
Under that umbrella, one can certainly discuss tradeoffs a society makes between "Better for the individual" vs. "better for the collective" to move to somewhere the society wants to be economically. Where did "civil rights" stand when the US was employing children to operate textile machines or had no 40-hour work-week protections?
Under that umbrella, one can certainly discuss tradeoffs a society makes between "Better for the individual" vs. "better for the collective" to move to somewhere the society wants to be economically. Where did "civil rights" stand when the US was employing children to operate textile machines or had no 40-hour work-week protections?
I don't think electric car makers should be coddled like this. What exactly is your worry with negative press?
That an incumbent with substantial influence uses their position to suppress a disruptive technology that's an emerging danger of erroding their business model?
When an infant alternative is competing against a fully-developed and monetized legacy product, I'm okay with some coddling (e.g. the EV subsidies). You have to get to the point of "Is it likely this new thing will be more beneficial to society when equivalently developed?"
Otherwise we get stuck optimizing a local maxima.
When an infant alternative is competing against a fully-developed and monetized legacy product, I'm okay with some coddling (e.g. the EV subsidies). You have to get to the point of "Is it likely this new thing will be more beneficial to society when equivalently developed?"
Otherwise we get stuck optimizing a local maxima.
>That an incumbent with substantial influence uses their position to suppress a disruptive technology that's an emerging danger of erroding their business model?
That's already been happening since the industrial age, and since we haven't plateaued, your fear of a local maxima doesn't look convincing. Would you agree that it is only fair that the onus is on you to explain what is different about this scenario?
I see it as a good thing because by the time the disruptive tech does take over, the new tech has undergone a cycle of refinement to the point where it goes from being "great", "oh this is nice" to being simply unavoidable.
Also, why is the incumbent suppressing a disruptive tech necessarily a bad thing? Since they're an incumbent they could also be in a position to evaluate the tech and know that there are certain downfalls to that approach. I'm fairly certain that as a reasonable person, you would agree that most disruptive tech never pans out and the companies die out.
> You have to get to the point of "Is it likely this new thing will be more beneficial to society when equivalently developed?"
Well but the the government picks winners and losers and uses tax payer money for that. What if there is this other great tech that even better and it gains some prominence 4 years after the government decided to subsidize EV? Do we now abandon all EV subsidies? Subsidize both?
That's already been happening since the industrial age, and since we haven't plateaued, your fear of a local maxima doesn't look convincing. Would you agree that it is only fair that the onus is on you to explain what is different about this scenario?
I see it as a good thing because by the time the disruptive tech does take over, the new tech has undergone a cycle of refinement to the point where it goes from being "great", "oh this is nice" to being simply unavoidable.
Also, why is the incumbent suppressing a disruptive tech necessarily a bad thing? Since they're an incumbent they could also be in a position to evaluate the tech and know that there are certain downfalls to that approach. I'm fairly certain that as a reasonable person, you would agree that most disruptive tech never pans out and the companies die out.
> You have to get to the point of "Is it likely this new thing will be more beneficial to society when equivalently developed?"
Well but the the government picks winners and losers and uses tax payer money for that. What if there is this other great tech that even better and it gains some prominence 4 years after the government decided to subsidize EV? Do we now abandon all EV subsidies? Subsidize both?
> since we haven't plateaued
I'd argue ICE is the ultimate example of that. I'll toss in monoculture, fertilizer/herbacide-powered farming; centralized banking (in the sense of Chase, not the Fed); commute-mandating urban planning; hydrocarbon-powered grid energy production; von Neumann architecture; electrical computing; and nuclear reactor designs to the list.
My argument is not that these things aren't efficient. Indeed, I'm saying they're so hyper-optimized that they're nigh impossible to unseat.
Maybe fusion power production is feasible, maybe it isn't. But where is the very substantial investment to see going to come from if it can't immediately compete against existing solutions?
And barring that investment appearing, we (as a planet) end up stuck on a technology that's globally suboptimal and impossible to improve further.
> Also, why is the incumbent suppressing a disruptive tech necessarily a bad thing? Since they're an incumbent they could also be in a position to evaluate the tech and know that there are certain downfalls to that approach.
I feel history is definitely on my side that large companies with a focused product are more likely to attempt to torpedo viable alternatives than incubate and help them. (E.g. MS/Lotus, MS/Netscape, MS/OSS, GM/Streetcars, GM/EV1)
I'd argue ICE is the ultimate example of that. I'll toss in monoculture, fertilizer/herbacide-powered farming; centralized banking (in the sense of Chase, not the Fed); commute-mandating urban planning; hydrocarbon-powered grid energy production; von Neumann architecture; electrical computing; and nuclear reactor designs to the list.
My argument is not that these things aren't efficient. Indeed, I'm saying they're so hyper-optimized that they're nigh impossible to unseat.
Maybe fusion power production is feasible, maybe it isn't. But where is the very substantial investment to see going to come from if it can't immediately compete against existing solutions?
And barring that investment appearing, we (as a planet) end up stuck on a technology that's globally suboptimal and impossible to improve further.
> Also, why is the incumbent suppressing a disruptive tech necessarily a bad thing? Since they're an incumbent they could also be in a position to evaluate the tech and know that there are certain downfalls to that approach.
I feel history is definitely on my side that large companies with a focused product are more likely to attempt to torpedo viable alternatives than incubate and help them. (E.g. MS/Lotus, MS/Netscape, MS/OSS, GM/Streetcars, GM/EV1)
Then don't buy a Tesla. There are plenty of cars that log data in a user-friendly fashion, and plenty others that don't log at all.
You have choices.
You have choices.
How can I find such cars? Where do I look for the information regarding data collection policies of a specific car? Reading page after page of verbose legalese of licence agreements and trying to interpret them for hidden meanings is not practical. How do I, as a privacy-concoius person, find a car?
You personally: yes.
Then if you want, you summarize your findings and publish them so others don't have to (and if you're particularly savvy, you monetize that publishing. ;) ).
Then if you want, you summarize your findings and publish them so others don't have to (and if you're particularly savvy, you monetize that publishing. ;) ).
This will have to change next year, at least for European customers, when the EU General Data Protection Regulation takes effect. The GDPR grants everybody a right to data portability for personal data:
https://en.wikipedia.org/wiki/GDPR#Data_portability
The mobility data they collect is regarded personal data because it indirectly allows to identify the driver.
https://en.wikipedia.org/wiki/GDPR#Data_portability
The mobility data they collect is regarded personal data because it indirectly allows to identify the driver.
Will they continue to have European customers once that regulation takes effect?
The Summon incident certainly was rather sleazy. Tesla's press release made it sound like a long, complicated series of steps were involved that could only be triggered intentionally. In reality, all it took was the car thinking the user double-tapped Park rather than single-tapping it, and then them not noticing and cancelling in time. Not only that, according to other owners there was sometimes a several second lag before message and warning sound to appear, meaning he could well have been outside the car by then.
This has bothered me. Tesla has all the data, but will only release it when it benefits them. It is easy to control the public debate when you are the gatekeeper of facts.
I just go surf youtube if I want to real results of AP2 and AP1 unfiltered. Some of them are downright scary. There is not going to be a sudden jump to good autonomous driving.
Also remember that AP1 wasn't even created by Tesla. That technology belonged to Mobileye, who is now owned by Intel. And it is public that Mobileye was the party who ended their relationship; Tesla was pushing their tech beyond what Mobileye deemed safe.
Thus, AP2 was born.
Tesla is a surprisingly scummy company when you start digging into some of their behavior.
Thus, AP2 was born.
Tesla is a surprisingly scummy company when you start digging into some of their behavior.
Elsewhere on HN, someone posted this video:
https://www.youtube.com/watch?v=UZ1XLqc5IUg
This was my comment reply:
Wow. That guy is going nineteen miles per hour on a simple road in daylight and the car literally crossed to the wrong side of the road in nine seconds!
That's an absolute disaster.
This was my comment reply:
Wow. That guy is going nineteen miles per hour on a simple road in daylight and the car literally crossed to the wrong side of the road in nine seconds!
That's an absolute disaster.
As much as I would love to have a self-driving car(and I would!) I honestly shake my head at anyone saying that we will have fully autonomous vehicles in the next 5-10 years. Unless the roads are fully standardised and beacons placed every few meters, this is literally the image recognition problem all over again. In the 60s scientists thought that with the power of computers they will "solve" image recognition completely within a few months. It's 2017 and our best of the best algorithms and neural networks are stupidly bad at things they haven't been specifically trained to recognise, and even then they make catastrophic mistakes. And then we want to put essentially this tech on a car, where our roads are bad, signs are faded or straight up wrong, where you can have any sort of weather conditions obscuring the view and/or lidar, plus the whole thing has to be so reliable that you can put it on a consumer product......within 10 years? If anything, we will get a better version of Tesla Autopilot within 10 years, but full autonomy? Try 50 years.
Humans aren't all that great at driving safely and consistently either. Autonomous vehicles just have to be better than the average (not terrific) driver, not perfect; and progress recently beggars the progress in previous generations. I'm placing my money on sooner rather than later in this case.
You see, this is where I disagree.
"better than a human" was never a valid defense against a machine failing and killing someone. Why would it be different with autonomous cars?
Let's say we have a fully 100% autonomous car, where you don't have to pay attention to the road. It crashes, injures or kills someone - who is responsible?
If a human did that, then we call it "an accident" and your insurance company foots the bill. With an autonomous car, sure, an insurance company will pay for the damages - but they will demand that a problem is fixed or they won't insure those cars anymore.
So anything less than 100% perfect is a huge liability for manufacturers, and we are very very far off 100% perfect.
I cannot imagine a world where autonomous cars have minor/major accidents but we just brush it off as "meh, whatever - humans have a higher rate of accidents so it's fine".
Or maybe think about it this way. Introduction of autopilot in airplanes resulted in a major improvement in safety, the number of plane crashes has literally plummeted(sorry) after its introduction. But it hasn't eliminated crashes entirely - and every time there is a crash, Boeing/Airbus will ground their entire fleets to investigate what went wrong, if there is a perceived fault then thousands of planes have to be updated before they are safe to fly.
Now bring that analogy over to cars - small, personal transportation vehicles, where profit margins really aren't that fantastic. Which manufacturer will risk releasing a product where they have to keep updating and maintaining the autopilot for the lifetime of the car, and for which they are liable every time it fails? Not to mention that planes are held to ridiculous safety standards, while cars are not(so far, I haven't heard an answer to what happens if a bird shits on the lidar and you don't clean it - it's an instrument with almost chirurgical precision which requires similar level of maintenance, and I know people who buy a car and then don't even wash it once in 3 years of owning it).
I'm not saying we won't get there. I'm saying that I don't see any chance that this technology will be ready in 5-10 years, and if anyone is foolish enough to sell an "autonomous" car in the next 10 years, they will set back our progress massively because the car will fail, and there will be public and legislatory backlash.
Let's say we have a fully 100% autonomous car, where you don't have to pay attention to the road. It crashes, injures or kills someone - who is responsible?
If a human did that, then we call it "an accident" and your insurance company foots the bill. With an autonomous car, sure, an insurance company will pay for the damages - but they will demand that a problem is fixed or they won't insure those cars anymore.
So anything less than 100% perfect is a huge liability for manufacturers, and we are very very far off 100% perfect.
I cannot imagine a world where autonomous cars have minor/major accidents but we just brush it off as "meh, whatever - humans have a higher rate of accidents so it's fine".
Or maybe think about it this way. Introduction of autopilot in airplanes resulted in a major improvement in safety, the number of plane crashes has literally plummeted(sorry) after its introduction. But it hasn't eliminated crashes entirely - and every time there is a crash, Boeing/Airbus will ground their entire fleets to investigate what went wrong, if there is a perceived fault then thousands of planes have to be updated before they are safe to fly.
Now bring that analogy over to cars - small, personal transportation vehicles, where profit margins really aren't that fantastic. Which manufacturer will risk releasing a product where they have to keep updating and maintaining the autopilot for the lifetime of the car, and for which they are liable every time it fails? Not to mention that planes are held to ridiculous safety standards, while cars are not(so far, I haven't heard an answer to what happens if a bird shits on the lidar and you don't clean it - it's an instrument with almost chirurgical precision which requires similar level of maintenance, and I know people who buy a car and then don't even wash it once in 3 years of owning it).
I'm not saying we won't get there. I'm saying that I don't see any chance that this technology will be ready in 5-10 years, and if anyone is foolish enough to sell an "autonomous" car in the next 10 years, they will set back our progress massively because the car will fail, and there will be public and legislatory backlash.
We get used to things that seem unsafe but are safe very quickly according to history. Elevators, for example. When I was young they generally had human operators, but now they're all autonomous, have been for a while, fail sometimes here and there but nobody seems hugely exercised about it.
That would make logical sense, yes. But humans aren't logical, and regulators won't approve autonomous vehicles until they make accidents exceedingly rare.
Also, even the worst driver I know drives better than the Tesla in that video. Seriously, watch the video if you haven't already. I think Google/Waymo are really going to be the winners here.
Also, even the worst driver I know drives better than the Tesla in that video. Seriously, watch the video if you haven't already. I think Google/Waymo are really going to be the winners here.
The logging is also the primary reason why I'm not getting a Tesla.
my 17 Volt reports my driving even without my solicitation. Imagine my surprise the first time I looked at what it had and it showed exactly where I had accelerated too hard or braked too hard; all times showing of electric drive and such to friends.
Needless to say I did not sign up for their related insurance program which sends your driving information to interested insurance companies.
my big concern isn't these affiliates, its the government being able to subpoena such data since its not in my possession and there for not protected by any rights I have.
Needless to say I did not sign up for their related insurance program which sends your driving information to interested insurance companies.
my big concern isn't these affiliates, its the government being able to subpoena such data since its not in my possession and there for not protected by any rights I have.
Is there a transmitter you can disconnect? Or maybe stick a baby faraday cage on?
GM vehicles have OnStar which is their communication mechanism.
Pretty much all new cars have it. They are hardly unique in that sense.
It's a major factor for me as well. I'm even leery of these always-connected infotainment systems in modern cars.
It seems like there's a large volume of information in the logs. I mean, if Tesla can tell whether the driver has his hands on the wheels there's probably a ton of other details.
It makes me wonder how practical it would be to use these logs. I hear about so-called "black boxes" in cars (not just Tesla), but where are the tools to decipher the data? How much can you trust that the logs reflect reality?
It makes me wonder how practical it would be to use these logs. I hear about so-called "black boxes" in cars (not just Tesla), but where are the tools to decipher the data? How much can you trust that the logs reflect reality?
I'm skeptical of the "hands on the wheel" thing. They can't actually sense hands on the wheel, what they sense is torque on the steering wheel. I often have my car nag me to put my hands on the wheel when I already have them there, and I'm just not applying enough torque with them.
It's possible that the threshold for the nag is just higher, and that the torque sensor is so sensitive there's no way it could possibly have a false negative here. But we just don't know.
It's possible that the threshold for the nag is just higher, and that the torque sensor is so sensitive there's no way it could possibly have a false negative here. But we just don't know.
I wonder if there's enough play in the steering column that you could send generated micro torque amounts back up to the steering wheel, then see if they're resisted.
You might be able to do it just by using road vibration that comes through the steering column. I think it's quite possible to detect hands with pretty good sensitivity in this way, it's just that the evidence suggests it's not actually that sensitive.
Good idea. It's definitely an interesting problem and seems like it would be really relevant to the safety side of an autopilot.
E.g. if there are no hands on the wheel, autopilot should probably do whatever it can to prevent a likely accident it sees
E.g. if there are no hands on the wheel, autopilot should probably do whatever it can to prevent a likely accident it sees
It might be difficult, but that shouldn't impact the right of drivers (if any) to try.
I agree, and it's not just drivers that would be able to make use of such "black box" log data but also insurance companies and law enforcement.
The thing is it would require A LOT of cooperation from Tesla. Once Tesla starts sharing logs, they have to support it and can no longer change it at a "whim" for internal purposes.
Perhaps most importantly, Tesla would also have to defend the accuracy of the logs when they go against other observations. Just because something is in a log, doesn't mean it correct, accurate or true.
I can understand why they would exert a mighty effort to avoid giving out logs.
The thing is it would require A LOT of cooperation from Tesla. Once Tesla starts sharing logs, they have to support it and can no longer change it at a "whim" for internal purposes.
Perhaps most importantly, Tesla would also have to defend the accuracy of the logs when they go against other observations. Just because something is in a log, doesn't mean it correct, accurate or true.
I can understand why they would exert a mighty effort to avoid giving out logs.
> The driver said he believed Tesla had violated Swiss data protection law by declining to send the raw data
Raw data might not help anything if there are no instructions on how to interpret it. There are two ways of hiding data: by not showing at all and by showing it in a flood of useless data, where it can't be seeing except if you know what you are looking for (and when there is only raw data and no manual, well, you can't make a good use of it).
Yes, having some data is better than none, but I don't expect it to be that helpful for a while.
Raw data might not help anything if there are no instructions on how to interpret it. There are two ways of hiding data: by not showing at all and by showing it in a flood of useless data, where it can't be seeing except if you know what you are looking for (and when there is only raw data and no manual, well, you can't make a good use of it).
Yes, having some data is better than none, but I don't expect it to be that helpful for a while.
The other way to look at it is the difference between not having data, which makes it impossible to interpret, and having data, which makes it somewhere between easy and very difficult to interpret.
Having that full fidelity raw data would be a gold mine, even without instructions. I'm sure the hacker community would reverse engineer it within months.
This is the second story that's somehow made it on the front page of hn today that biasedly talks down Tesla and particularly autopilot. All this on the day that Tesla takes over Ford in marketcap. This is not a coincidence. I'm getting the feeling that there is definitely some astroturfing going on here. Instead of talking about this historic day for Tesla and the EV market in general, we're talking about an AP accident from a year ago and some nonsensical study navigant did. Seems like a concerted effort to steer the conversation
"lets out data when it suits"
that is rich coming from a journalist.
that is rich coming from a journalist.
This is for real self-driving. Level 2 semi self driving systems like Tesla's old "autopilot" don't have to report, so we don't get data for production vehicles.
From the DMV reports, Google/Waymo is performing about two orders of magnitude better on self-driving disengagements than anybody else reporting.[1] They're up to 5000 miles between disengagements. No disengagements on freeways. That's one of the few objective measures we have in this business. Road and Track, and Top Gear, aren't getting self driving cars to test yet.
[1] https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/disen...