Would you ride in Elon's robotaxi

[clyde]

Originally Posted by rumatt

I'm not sure if you misunderstood what I was saying, or you're just being difficult.

On the off chance that it's the former, I'll clarify one more time. I believe it's safe to use "Because I'm still 'driving' and I can override it any time I don't like what it is doing."

But you're a smart guy and I know you know exactly what I'm arguing. So unfortunately, I'm pretty sure it's the latter.

Honestly, I did not think that was an answer to "what convinced you to turn over your responsibilities to the car?" but okay. In that case, I guess my question wasn't specific enough and that's my bad.

To borrow from your cruise control analogy, what makes you believe you will be able to override it any time you don't like what it's doing? Because Musk tells you you can? But, whatever. You've made it clear where you are.

On to cruise control...

Traditional cruise control exists to do one thing: maintain a set vehicle speed. Full stop. Its inventor was a blind guy who did not like how his lawyer was jerky on the gas for no apparent reason. He wanted his lawyer to have an easier time holding a constant speed. (Google "Teetor cruise control" for a more nuanced version.)

Autopilot exists to do many things such as choose and then adjust vehicle speed, position, and heading as it sees fit on a continuous basis. It exists, according to its chief proponent and his company to be a means to an end; that of fully autonomous driving.

Traditional cruise control has two potential points of failure: runaway acceleration and an inability to disengage that could create a need for the driver to recognize the failure condition and take significant corrective action, likely with significant time afforded, to avoid catastrophic results.

Autopilot has, um, I have no idea how many potential points of failure, but I'd wager "a fuckton" (not sure if SAE or metric) covers it. I suppose they all boil down to improperly responding to stimulus (taking action where none was necessary, taking the wrong action, or taking no action at all when it should have) with no time for the driver to recognize and respond before catastrophic results.

If traditional cruise control works as designed, it will drive a vehicle into something unless the vehicle runs out of fuel and coasts to a stop first.

If Autopilot works as designed, it can drive you where you want to go or tell you to take over if it thinks it can’t.

Successful (let me know if we need to define “successful”) use of traditional cruise control requires the driver must make continuous manual steering inputs, manually apply the brakes to reduce speed, and therefore also requires an attentive driver.

Successful use of Autopilot requires the driver to make physical contact with the steering wheel once every 1-5 minutes depending on environment and speed (most recent info I found, but perhaps not up to date? https://electrek.co/2018/06/11/tesla...g-hands-wheel/ ) while the driver does whatever else that may or may not include paying attention to what a driver normally pays attention to.

If Autopilot doesn’t belong in a conversation about full autonomous driving, I completely fail to see how injecting traditional cruise control fixes that. Traditional cruise control is barely Level 1 and Autopilot is Level 2 certainly, and arguably quasi-Level 3 (save for Tesla saying it’s not) and Level 3 is fully, if conditional, autonomous driving.

I don’t see how pressing cruise control as a comparison point to Autopilot it as you have will help you make your point. They exist to do totally different things and their regular operation have vastly different requirements.

But, you seem to see them as equals, I’ll play along to help things move along.

As for my own use of traditional cruise control…

I used cruise control for the first time in 1986. At that point, cruise control had been available on production cars for nearly 30 years. My assumption today, perhaps faulty, is that if there were significant issues with the safety of cruise control systems by 1986, they would have been identified and addressed through court cases and/or regulation due to the volume of time and use of the system (see JST's tort post up thread) in nearly all possible environments by people in nearly all or nearly all parts of American (and many non-American) society. At the time, I doubt I gave its safety more than a passing thought if that much.

In those first few years of driving, I suspect that almost every time I used cruise control, I presented less of a risk to myself and others compared to the times I was not using cruise control given how I often drove back then.

“Almost every time.”

A time that doesn't qualify was in 1988. I had the cruise control set while traveling southbound on the NY Thruway (I-87) in Saugerties, NY, fell asleep, drifted through the right shoulder, woke up, overcorrected, and spun through the Number 2 and Number 1 lanes into the median before crashing into the guardrail separating me from northbound traffic. There are many ways that could have turned out so much worse than it did that I can't count them all. No one got hurt. Only my car and a couple reflector posts and sections of guardrail owned by New York were damaged.

Since then, there’s been another 30+ years of cruise control use. I’m guessing significantly more use than the previous 30 years with cruise control making a nearly universal switch from the option sheet to standard equipment by the early 1990s. More importantly, over the past 30 years there has been an ever increasing portion of the driving segment of society that has never known the absence of cruise control. It’s just been there and part of the regular driver environment no matter what car one may be sitting in with the only difference being where the buttons are.

This compares to the early cruise control using population that was probably of a higher income and education level with some years of previous driving experience. That’s not unlike the current Autopilot using population. On the other hand, many things related to safety and risk tolerance or perceptions of acceptable risk are different today than they were in the early days of cruise control. Back then, everyone smoked, we used leaded gas, scoffed at helmets, and barely tolerated seat belts being present in our cars. Even when we knew better, we acted like we didn’t. I’m probably digressing.

In 60ish years of traditional cruise control use, documented failures of the system to operate as designed and intended (dismissing cases where it won’t engage) are so rare it is hard to find any cases at all. Society, particularly American, has also become so much more litigious in the past 30ish years that even if there was a semi-solid reason to argue against the safety of traditional cruise control hadn’t been explored or tested by the late 1980s, it’s all but inconceivable that it would not have been tested by now—and been successful if even dubiously meritorious.

So, today, why do I believe using traditional cruise control is safe to use? 60 years of demonstrated global use of safe cruise control operation.

[clyde]

Originally Posted by rumatt

Nice edit

[rumatt]

what makes you believe you will be able to override it any time you don't like what it's doing?

I'm using my judgement based on my experiences, and the types of scenarios where I choose to use it.

What happens if the car suddenly swerves hard left and/or does not let me override it? The same thing that happens if the cruise control accelerates out of control and/or does not let me override it. I crash. True statements of any automatic driver control.

On to cruise control...

I didn't ask for a comparison of the differences between cruise control and autopilot. I asked only about cruise control and why you felt it was safe to use.

I never said cruise and autopilot are equivalent. They obviously are not. Autopilot is a much more extreme form of cruise control. But the analogy points out the flaws in your line of questioning. If you can't answer them about cruise control, they're not very good questions, right?

Your answer seems to be

60 years of demonstrated global use of safe cruise control operation.

So the obvious followup question is: what about the first 1, 5, 10 years of use? Who told you it was safe then? If you didn't know it was safe why did you use it? Why did you endanger the lives of my family?

And if years of experience by end users is the measure we use to define safety, how are we supposed to get there without using it?

[clyde]

Originally Posted by rumatt

I'm using my judgement based on my experiences, and the types of scenarios where I choose to use it.

What happens if the car suddenly swerves hard left and/or does not let me override it? The same thing that happens if the cruise control accelerates out of control and/or does not let me override it. I crash. True statements of any automatic driver control.

Originally Posted by rumatt

I didn't ask for a comparison of the differences between cruise control and autopilot. I asked only about cruise control and why you felt it was safe to use.

I never said cruise and autopilot are equivalent. They obviously are not. Autopilot is a much more extreme form of cruise control. But the analogy points out the flaws in your line of questioning. If you can't answer them about cruise control, they're not very good questions, right?

Autopilot is not a more extreme version of cruise control. It's an entirely different thing. I explained why I don't think they are similar enough to warrant comparison and why I don't accept your analogy as valid. Because it is not. But, whatever.

So the obvious followup question is: what about the first 1, 5, 10 years of use? Who told you it was safe then? If you didn't know it was safe why did you use it? Why did you endanger the lives of my family?

As I described, I did not use it in the first 1, 5, or 10 years it was available. I waited about 30 years. Even then, when I did use it, I nearly killed myself and very easily could have killed anyone nearby. That that was not the result was dumb and fortuitous luck. Hopefully, if you live another 30 years, you won't have a similar story to tell. As I also described, I didn't give the safety of cruise control any thought at the time, much like it appears you're giving no thought to the safety of using Autopilot today.

And if years of experience by end users is the measure we use to define safety, how are we supposed to get there without using it?

It can be a measure, sure. But I'm not saying that's the only way. Not at all. For some reason, Musk and Tesla thought it was safe enough to unleash on the world a few years ago and also safe enough to add all the things they've added since. What were those reasons?

I would like to think they have a sound reason (ideally "reasons" plural) to believe it's safe like some combination of simulations, real world data gathered under controlled circumstances, maybe other things. I'd like to think that, but I don't. I'm highly skeptical for a few reasons: They

• won't show us their data
• have given incomplete and misleading data about the subject to the government wrapped in substandard self-interested analysis
• they regularly mislead, deceive, and lie to the public and government agencies
• they operate beyond the bounds of law and regulation (see SEC and labor violations for starters)

It's all about their self-interest whatever effect it may have on the rest of us. They're kind of too Trumponian to be trusted in my view. Of course, 60 million people in the US think he's a truthteller, so

Would anyone be all that surprised if Musk went on a "fuck you world" Twitter rant and confessed to unleashing Autopilot and enhanced features to drive sales, media interest, and goose its stock price without any underlying reasonable reason suggesting or showing it was safe? Would anyone have trouble believing it? More directly, would it prompt you to reevaluate your use of it? Given what you've told us in this thread, the answer is no.

None of that means Autopilot isn't safe enough to use. We just don't have any way to know. You made your decision why you want to believe them. ZBB made his. JST saw all the same things you did and made a different decision.

Of course, there's a point where if you're going to advance anything you need to take risks. The choice is to take those risks blindly, like many Tesla drivers are doing (and, unfortunately, taking those of us that must shared the roads with them along for their joyride) or refusing to take those risks without an opportunity objectively evaluate those risks and thereby make an informed choice about what risks we're willing to take and under what circumstances.

Once you cross that threshold and make it available for general use, opening the data about how, when, and where Autopilot and competing technologies are used and, perhaps more importantly, how, when, where, and under what circumstances it fails would let us make informed decisions about whether the technology is safe enough to keep using widely or it needs more development work.

There was a time when the only choice for just about everything was to just start doing things in public and see what happened. Today, there are still some things we don't have much choice about but to turn them loose in public and see what happens, but there are a lot of things where we have better choices available to us to give us a fair preliminary sense of what's likely to happen and have debate about what constitutes acceptable risk in difference scenarios related to those things before public introductions.

[robg]

Clyde - how do you feel about adaptive cruise control systems?


Sent from my iPhone using Tapatalk

[clyde]

Originally Posted by robg

Clyde - how do you feel about adaptive cruise control systems?

Assume you're referring only to systems that work at controlling speed (throttle, brakes, and/or shifting only and no steering or other functions).

Kind of mixed, I think, but haven't given it much thought until recently. On the comfort level spectrum between traditional cruise control and Tesla's Autopilot, closer to traditional cruise, but still decidedly undecided.

Two big things tend to give me comfort with it. One is that it's a limited and incremental advance that acts only on vehicle velocity, no other axis of motion. The other is that it has a substantial length of use in the real world (about 25 years) without major issues.

OTOH, the earlier systems were much simpler than today's with many fewer points of failure. Today's adaptive cruise control system are often tied in with other systems that affect more aspects of the vehicle's behavior than just velocity and those systems are not well proven as independent systems...and we're tying all these different things together? I'm not fond of that.

[robg]

Originally Posted by clyde

Assume you're referring only to systems that work at controlling speed (throttle, brakes, and/or shifting only and no steering or other functions).



Kind of mixed, I think, but haven't given it much thought until recently. On the comfort level spectrum between traditional cruise control and Tesla's Autopilot, closer to traditional cruise, but still decidedly undecided.



Two big things tend to give me comfort with it. One is that it's a limited and incremental advance that acts only on vehicle velocity, no other axis of motion. The other is that it has a substantial length of use in the real world (about 25 years) without major issues.



OTOH, the earlier systems were much simpler than today's with many fewer points of failure. Today's adaptive cruise control system are often tied in with other systems that affect more aspects of the vehicle's behavior than just velocity and those systems are not well proven as independent systems...and we're tying all these different things together? I'm not fond of that.

I see. It seems your primary concern is with electronic systems taking unintended and potentially harmful actions. Given that nearly all modern cars have electric steering and electronic throttle systems (in addition to computerized braking systems for stability control, and automatic braking), it seems that nearly any new car would be a cause for a concern, no? AEB systems alone are prone to false positives leading to slamming on the brakes unintentionally.

Then there’s the issue with any modern safety system of lulling people into complacency and over reliance (blind spot monitoring and backup sensors even have this issue for example)

[rumatt]

Originally Posted by clyde

Autopilot is not a more extreme version of cruise control. It's an entirely different thing.

Cruise, adaptive cruise, and autopilot all have one thing in common - the human is watching and responsible for the end result. You're glossing over this like it's a minor detail. It's not.

If you had to lump everything into two buckets, autopilot is definitely closer to cruise control than it is to full self driving. Removing the human from the equation is, um, kind of huge.

Yes, autopilot automates more thus creates more potential risk than cruise. How much risk is too much? Where do we draw the line? This would make for a very interesting discussion, but it appears you have already drawn the line and put things into two buckets - a) "safe enough to not worry about" and b) "risking the well being of your family". It's impossible to have a constructive discussion when you start it with that premise.

As I described, I did not use it in the first 1, 5, or 10 years it was available. I waited about 30 years.

It certainly makes sense that someone who didn't want to use cruise control also wouldn't want to use autopilot. I have no problem with that.

it appears you're giving no thought to the safety of using Autopilot today.

Blatantly false. I've stated twice that I use it only where I feel comfortable that I can safely override it before anything bad were to happen. Isn't thinking about safety an obvious prerequisite of that statement?

For some reason, Musk and Tesla thought it was safe enough to unleash on the world a few years ago and also safe enough to add all the things they've added since. What were those reasons?

Why do you keep asking Elon to tell you how safe it is? Would you believe them if he did? I don't buy that there is anything he could tell you that you would accept. Your answer regarding cruise control was not "Because Ford told me". It was years of experience in the hands of real users.

Musk is already in the process of creating the only answer you would ever accept anyway - a long history of use in the real world.

[clyde]

Originally Posted by rumatt

Musk is already in the process of creating the only answer you would ever accept anyway - a long history of use in the real world.

This clearly headed into an immovable object at high speed. Call me when we get there.

[clyde]

Originally Posted by robg

I see. It seems your primary concern is with electronic systems taking unintended and potentially harmful actions. Given that nearly all modern cars have electric steering and electronic throttle systems (in addition to computerized braking systems for stability control, and automatic braking), it seems that nearly any new car would be a cause for a concern, no? AEB systems alone are prone to false positives leading to slamming on the brakes unintentionally.

Then there’s the issue with any modern safety system of lulling people into complacency and over reliance (blind spot monitoring and backup sensors even have this issue for example)

Primary concern is the free introduction of systems that transfer control of the vehicle from human to machine without any societal agreement on what is "safe enough" or any common understanding of what the systems are capable of, what they're incapable of, where their edges are, what their error rates are, what happens when they fail, etc compounded by manufacturers hiding their specific knowledge and understanding of these details from the public and regulating agencies.

For the fourth(?) time in this thread, I have no reason to believe these systems are unsafe. I'm completely open to the possibility these systems are significantly safer than humans driving without these systems in place, but I also have no reason to believe that's the case.

That's what I'm asking for. That's what I want: An objective source of data or information and validation that suggests they're safe enough to use. Instead, all we have are outlandish marketing claims from manufacturers and their leadership that have demonstrated histories of being misleading, deceptive, and outright lying when it suits them. Unfortunately, that's all we have and more unfortunately, that's more than enough for a lot of people.