Would you ride in Elon's robotaxi

[equ]

I haven't been able to read most of this thread, but it is a topic that I find fascinating and important.

I've used cruise control now and again, really only if my right leg - which happens to be on the side of my L4-L5 affected nerve - needs some resting on very long trips. So basically never.

I tried the JGC's auto cruise control a bunch of times, never warmed up to it. It gave three following distance settings and seemed to work "for the most part". Its blind spot sensors were more useful to be honest.

The Macan Turbo did not have distronic but it DID have lane keeping assist. Simply atrocious. Some semblance of steering, but generally a terrible check-the-box execution. Cannot be used on any real curves.

My general thought on this topic is that any solution between "full human driving" (and i consider manual/automatic/electric about the same for this purpose, let's not worry about spark plug advance and older things) and true autonomous driving (what is it, level 4?) where you climb into a lounge-car and mix drinks and pay ZERO attention, no override ever, highly troublesome. The definition of these two endpoints are clear, with clear responsibility lines. The middle is murky as it could go any which way. There is a high order infinity (uncountable) number of ways in which semi-autonomous driving could end up. It also doesn't play well with human nature; we WILL stop paying attention as semi-tested technology begins to fill the void and it WILL be catastrophic. This is not a feature that should be oversold. That has been my beef with Tesla from years ago. I was pissed they called it "autopilot" back in the day and then further pissed when it's called "full self driving". Beyond irresponsible.. It might be a very good car in other respects, but the general AI problem of autonomous driving is nowhere near solved and to continually and intentionally suggest the opposite (over the years: coast-to-coast, robotaxis, "FSD" etc), while the product benefits from the attention and the stock price benefits from the seeming achievement is reckless and perhaps criminal.

It will also not be solved just by cameras and image processing. We WILL need radar/lidar and I assume gyroscopic sensors to "feel" the car and the road. So the hardware is incomplete as well.

[robg]

Originally Posted by equ

I haven't been able to read most of this thread, but it is a topic that I find fascinating and important.

I've used cruise control now and again, really only if my right leg - which happens to be on the side of my L4-L5 affected nerve - needs some resting on very long trips. So basically never.

I tried the JGC's auto cruise control a bunch of times, never warmed up to it. It gave three following distance settings and seemed to work "for the most part". Its blind spot sensors were more useful to be honest.

The Macan Turbo did not have distronic but it DID have lane keeping assist. Simply atrocious. Some semblance of steering, but generally a terrible check-the-box execution. Cannot be used on any real curves.

My general thought on this topic is that any solution between "full human driving" (and i consider manual/automatic/electric about the same for this purpose, let's not worry about spark plug advance and older things) and true autonomous driving (what is it, level 4?) where you climb into a lounge-car and mix drinks and pay ZERO attention, no override ever, highly troublesome. The definition of these two endpoints are clear, with clear responsibility lines. The middle is murky as it could go any which way. There is a high order infinity (uncountable) number of ways in which semi-autonomous driving could end up. It also doesn't play well with human nature; we WILL stop paying attention as semi-tested technology begins to fill the void and it WILL be catastrophic. This is not a feature that should be oversold. That has been my beef with Tesla from years ago. I was pissed they called it "autopilot" back in the day and then further pissed when it's called "full self driving". Beyond irresponsible.. It might be a very good car in other respects, but the general AI problem of autonomous driving is nowhere near solved and to continually and intentionally suggest the opposite (over the years: coast-to-coast, robotaxis, "FSD" etc), while the product benefits from the attention and the stock price benefits from the seeming achievement is reckless and perhaps criminal.

It will also not be solved just by cameras and image processing. We WILL need radar/lidar and I assume gyroscopic sensors to "feel" the car and the road. So the hardware is incomplete as well.

I agree. Requiring humans to attentively monitor their vehicles is the biggest flaw with all semi-autonomous systems. Even people who are professionally trained to do it have trouble (Uber crash in Arizona for example)

[John V]

According to a close friend who was brought in to an autonomous vehicle development lab in California to work in their simulation, that is the big problem. When people are required to be in the loop to react to situations that the autonomous system cannot handle, they almost universally react too late and often with a response that makes the situation worse.

[rumatt]

Originally Posted by John V

According to a close friend who was brought in to an autonomous vehicle development lab in California to work in their simulation, that is the big problem. When people are required to be in the loop to react to situations that the autonomous system cannot handle, they almost universally react too late and often with a response that makes the situation worse.

I feel pretty confident in my ability to control the vehicle when I'm using autopilot (again, only using it in places with nice buffers) but who knows, I could be wrong. And I don't doubt that some people will take it to an extreme and not pay attention at all because they feel the car is in control.

But once you go down the path of "the human might not do their job", well - they're already not doing their job. That article claims that depending on the time of day, 60-80% of drivers use their cell phone at least once while driving.

I would guess the risk of someone killing me goes something like this (greatest to smallest)

1. Drunk driver
2. Distracted driver (Texting, etc)
3. Tired drivers (NHTSA says 795 deaths in 2017 and I know someone who fell asleep and killed someone)
4. Assholes driving aggressively and swerving in and out of lanes.
5. Using immature auto-pilot while distracted (texting etc)

[big gap]

6. Using autopilot responsibly
7. Human driving responsibly

I assume 6 is worse than 7, but people do make mistakes - and there are some pretty shitty drivers out there, even when they're trying their best. How long will it take before 6 is no longer worse than 7? (on average - I'm not trying to beat the world's best driver)

Bottom line - if I want to stay up at night worrying about what might kill my family, I'm not going to be spending much time thinking about #6.

[ff]

rumatt and I are in complete agreement.

[John V]

I never said anything about autopilot being a bigger danger than drunk, distracted or tired drivers or people driving like assholes.

In my mind it's similar to the football helmet effect, where by providing people with a safety net, their level of engagement in dangerous activities increases. To be clear, I'm not accusing *you* of doing that, Matt. But based on the accidents that have already happened involving autopilot, it's clear some people see it as an excuse to "safely" do other things besides paying attention to driving. And the technology is clearly not ready for that yet.

[rumatt]

Originally Posted by John V

I never said anything about autopilot being a bigger danger than drunk, distracted or tired drivers or people driving like assholes.

In my mind it's similar to the football helmet effect, where by providing people with a safety net, their level of engagement in dangerous activities increases.

Understood and agree.

My point is that the bar for what is allowed on the road is already absurdly low. It's a shit show out there.. We allow selling high HP sports cars to teenagers and watch them crash at significantly higher rates than other drivers. Why? Because it's fun and they want to.

Autopilot is nothing compared to half the shit we allow on the roads, and has a (theoretical) up side.. It feels like a strange place to selectively focus ones rage and fear of safety.

But yeah, discussing it in the context of "hmmmm how should we regulate this shit?" does make a lot of sense.

[John V]

Originally Posted by rumatt

Understood and agree.

My point is that the bar for what is allowed on the road is already absurdly low. It's a shit show out there.. We allow selling high HP sports cars to teenagers and watch them crash at significantly higher rates than other drivers. Why? Because it's fun and they want to.

Autopilot is nothing compared to half the shit we allow on the roads, and has a (theoretical) up side.. It feels like a strange place to selectively focus ones rage and fear of safety.

But yeah, discussing it in the context of "hmmmm how should we regulate this shit?" does make a lot of sense.

The bolded portion is not a good argument for allowing something that professes to be somewhere between a driver aid and autonomous operation. I agree again that there is tons of shit that is very dangerous out there. People make extremely bad decisions that endanger everyone, and we largely do nothing about it. I see cars in the parking lot every day with tires so bald I'm not sure how they don't drive off the road in the lightest rain.

Because that shit exists doesn't mean we should be okay with blindly introducing features touted as "self driving" when they are clearly not, nor have they been exhaustively tested by any unbiased organization to have sufficient safety controls in place. Again. I trust *you* to make intelligent decisions about how and when it's safe to use. But 99% of the driving population? No. Especially with how Tesla has recklessly marketed that particular feature.

[clyde]

Originally Posted by rumatt

Autopilot is nothing compared to half the shit we allow on the roads, and has a (theoretical) up side..

https://gizmodo.com/of-course-someon...ing-1834638212

Originally Posted by jalopnik

Of Course Someone Shot a Porn Video in a Tesla Driving on Autopilot

It was only a matter of time of course. There’s now a porn video of two people having sex in a Tesla while it was driving on Autopilot.

The upside you had in mind?

[Josh (PA)]

Originally Posted by clyde

https://gizmodo.com/of-course-someon...ing-1834638212


The upside you had in mind?

to be fair... have you never gotten road head? I'd think doing it on autopilot is probably safer.