Self-Driving Cars

It’s worth noting that while we don’t try and force technology solutions at large for people failing to pay attention in an unassisted cars we do run a lot of public awareness about how it’s bad and a crime.

We have PSA on drunk driving not on failing to monitor your lv 2 system properly.

We also do impose technology restrictions in some instances. Breathalyser ignitions get fitting if you do enough drink driving, situational limiter boxes also get imposed.

Here's my angle. As long as we can do it with reasonable safety, we should be pushing autonomy as fast and as hard as possible for all forms of transport, but especially cars. Why? Because we've gotten way too complacent about the costs and dangers of human drivers.

If 40,000 people died a year (and millions were hospitalized) to eating bad sushi, we'd be cracking down on that industry faster than you can say "California roll".

I am not against Waymo or Cruise's approaches. If they can make it work, great. However, I don't see what they're doing as being sufficiently scalable, not unless they can dramatically reduce the cost.

It's not just about solving the problem — for all we know there could be dozens of potential ways to accomplish that. It's about solving it in a way that replaces manual cars with automated ones as rapidly as possible.

Why? Because every year that passes with humans driving cars is one that contains countless human tragedies, and it's utterly shameful that we tolerate it to the degree that we do.

Let’s be clear here, the US tolerates it to an absurd degree, roads in other developed countries are much safer.

The issue you’re running into isn’t a self-driving issue, it’s a USA cultural issue. As a cultural grouping the US simply isn’t particularly bothered about road safety. It’s why you don’t have a national seatbelt law, it’s why you give out drivers licences without tests, it’s why your roads are in such disrepair, it’s why drunk driving is still culturally acceptable in many areas, and yes it’s why you aren’t properly pushing self-driving technology.

In fairness, almost every state in The US (with one exception) has a seatbelt law.

Drivers' licenses are also handled at the state level, and I'm not sure where you're getting that they're handed out without tests. Maybe for learner's permits? Those mean you can't drive on your own, and even then we might've had to take a short written test, I don't remember (it's been a while). But as that goes by state, it's going to vary.

Silas, the world doesn't just consist of the US and Europe. Globally, the annual road death toll is 1.35 million.

Apparently, it's easier to make cars that drive more safely than it is to get legislators to fix the laws, but let's worry about one windmill at a time. Or rather, let's accept that we can work on more than one thing at a time.

Edit: Europe had 19,000 road deaths in 2021. That number has been trending downward, which is fantastic, but it needs to be zero, or as close to it as possible.

If the EU wants to hold back autonomy as long as possible until it's "perfect", that's its choice.

Edited by Fighteer on Apr 10th 2024 at 2:46:17 PM

Last time I looked into it I remember finding at least one state that issues a learner permit once you’re of age and a full licence as long as you don’t get any infractions on your learner permit for a certain number of years.

Both companies have always used safety drivers in situations where they were not prepared to use their systems in level 4 mode. This has continued even as they began to operate level 4 vehicles as well. As mentioned, Cruise temporarily suspended its driverless program in late 2023 after an incident where a pedestrian struck by another car hit the Cruise vehicle and was dragged as the vehicle (accurately) identified a problem and autonomously pulled over to the side of the road.

To be perfectly clear: suspending the program is a good thing. The situation is an edge case, but real harm was done and halting driverless operation until a solution is implemented to prevent a similar incident from happening again is the right call. There are inevitably going to be setbacks on the way to commercially viable level 5 systems, but accepting responsibility and doing everything possible to address issues as they arise is the right way to handle it. I hope Cruise is successful in fixing their system and is able to get back on the road without this being a fatal blow to their business.

Waymo has continued to operate level 4 vehicles in limited circumstances, and in fact recently expanded its program to autonomously delivering Uber Eats orders in Phoenix.

My friend, you literally accused Waymo and Cruise of being biased in favor of lidar because they've invested in lidar. You do not get to pretend that I'm the one being disingenuous here. Your exact words were "We'll never achieve self-driving without lidar, says the company manufacturing lidar." I was pointing out the ridiculousness of this by reversing the argument and applying it to Tesla. "this sort of argument style" is literally the augment you advanced.

I have never claimed that it was necessary. I have in fact explicitly stated that it may be possible to do it eventually. I literally say this IN THE POST YOU ARE QUOTING. "So far the only proven level 4 systems use lidar for extremely obvious reasons. It may be possible to eventually achieve level 4 without it." If you cannot read the posts you are responding to then I'm not sure this conversation is worth having.

You can pontificate about human nature, or you can mitigate harm. Driver monitoring is a way to mitigate harm in vehicles with driver assistance systems. Driver assistance systems with driver monitoring are safer than driver assistance systems without driver monitoring. We're saying "Tesla could save lives with more effective driver monitoring". You're saying "but it's not Tesla's responsibility that people are misusing its systems!" Even granting you the point (we have argued the point of whether Tesla's marking about Full Self-Driving is misleading into the ground), that's not the point. Whether or not Tesla is responsible, Tesla could save lives with more effective driver monitoring. Thus, we think that Tesla should implement more effective driver monitoring.

In other words, you're arguing about morals, but we're talking about outcomes.

Tesla has implemented comprehensive driver-monitoring that improves with each release of the software. The X post in question laments, somewhat tongue-in-cheek, that the same is not available for manual drivers and would probably save quite a few more lives.

"for extremely obvious reasons" — This is biased language. I'm surprised you can't see that. What are the "obvious reasons"? If Tesla FSD can accurately determine distance and speed without these tools, how are they "obviously" necessary?

How about a break from Tesla?

Waymo Driverless Car Gets Stranded, Rescued, & Manually Driven | #Waymo Ride Along #25

The ride begins in Los Angeles in a USPS parking lot with a blocked-off exit. It's not shown how the car got in there; from the narration, it must have used the one-way entrance. Anyway, it gets stuck when it can't figure out how to leave and keeps driving in circles, to the increasing annoyance of the postal employees. The passenger contacts roadside assistance, which is unable to persuade the vehicle to drive against the marked traffic flow. Eventually a service driver has to come in and manually operate the car.

I am curious how other autonomy solutions would resolve this edge case. Would a human always need to take over because the car should never violate traffic flow on its own? Should it be possible for the human to issue verbal commands if the car gets stuck? Could the car have recognized that it was entering a dead-end parking lot and refused?

Anyway, the problem the car is encountering here has nothing to do with lidar, radar, or vision. It has to do with decision-making. This is a point I raised earlier, which got ignored. If we assume that all self-driving vehicles are capable of accurately positioning themselves in 3D space and understanding their surroundings regardless of what technology they use, then we go back to situational awareness and semantic problem-solving.

Edit: For the sake of completeness, the service driver takes the Waymo vehicle outside its HD mapping area during the manual drive. This a limitation of the system that I'm not sure can be adequately overcome without massive capital investment, further increasing the marginal cost.

Edited by Fighteer on Apr 10th 2024 at 9:51:13 AM

It's still pretty shit by industry standards. Consumer Reports gives Tesla a score of 3 out of 10 on "keeping driver engaged". The best rating they gave was 9/10.

They use lidar for range finding. Accurately determining the range between the car and various things around it is extremely important for a self-driving system. I don't think it's controversial to say that this is obvious.

Fighteer. For the love of god. I need you to understand this. I am not, did not, and do not believe that they are necessary. I explicitly said that they may not be necessary. And then I pointed out that I said that since you appear to have missed it the first time. This is now the third time I am saying it in my last three posts.

I don't know what else to do. You are arguing against things I did not say. You are arguing against things I said the opposite of. Please fucking stop.

Worth noting is that Consumer Reports scored Autopilot, which is Tesla's basic ADAS that comes with all vehicles, not Full Self-Driving, which would have been in beta at the time of publication. I am not personally aware of how many driver-engagement features are backported from FSD to Autopilot. If in-cabin camera monitoring is not one of them, that could explain the low score. FSD has moved almost entirely to the camera and still maintains the steering wheel nag mainly for compliance reasons.

Autopilot is five-plus year old tech at this point: completely obsolete by the standards of FSD. I've heard rumors about why Tesla doesn't replace it with a feature-limited FSD stack, but I don't know the full story there so will not speculate.

As for the rest, I'm afraid we're going in circles. You think I'm ignoring your points, I think you're ignoring mine. I'm simply going to stop engaging.

Edited by Fighteer on Apr 10th 2024 at 9:59:22 AM

ABC News Australia: Tesla owner who falsely claimed car was in autopilot before Melbourne crash pleads guilty

The driver fled the scene and later attempted to blame Autopilot. Tesla's vehicle telemetry data revealed that she had not used it at all, and had instead attempted to "beat the tram". Indeed, the car detected the pedestrian and sounded an alarm for a forward collision warning.

After two years of claiming her innocence, the driver has pleaded guilty and faces up to five years in prison for dangerous driving, plus ten for leaving the scene of an accident.

(Australia puts you away for ten years for "failing to stop"? Man, when can we get laws like that?)

Edited by Fighteer on Apr 24th 2024 at 11:52:57 AM

I wonder if Tesla could also bring a civil case for the false claims made blaming them?

I doubt it. Tesla would have to prove they were financially damaged by the comments made and more to the point she’s in prison, not exactly going to be able to pay off fines. They’d be throwing a lot of money away that could just be spent on a PR campaign.

I didn’t think Tesla could enable the driver assist programs in Australia, for much the same reason they can’t in the EU.

Tesla does enable Autopilot (the basic driver assist system) globally, and has sued drivers for making false claims about the company in the past. This has most notably occurred in China, which has aggressive laws regarding such matters.

It is apparently something of a cottage industry in China to raise public outcries against companies based on false product liability claims for clout, and its government takes a harsh stance on that. Lying to escape a criminal charge isn't in quite the same category.

The article doesn't mention whether Tesla intends to sue in Australia, but I'd imagine five to fifteen years in prison will settle things nicely. Man, why can't we do that? In many parts of the US, a reckless gets you a few hundred dollars fine and the judge buys you a beer. I guess hitting a pedestrian elevates the severity of the crime, but still...

Edited by Fighteer on Apr 25th 2024 at 8:57:51 AM

Interesting (if perhaps paywalled) piece by Matt Yglesias arguing that at this point self-driving cars are, in fact, underhyped (https://www.slowboring.com/p/self-driving-cares-are-underhyped). This bit stood out to me:

That was a potential impact that hadn't occurred to me, though I'm not sure how realistic it is in the short term...

In a word: peak times. That (and road space usage) is a pretty big argument against inefficiently reinventing mass transit via unsupervised taxi°. Plus "same cost as owning a car but you don't own anything" is both less cost effective than public transport per mile travelled per person AND a continued transfer of assets away from people. "Everything is for rent and nothing is owned if anything goes wrong" isn't great.

°Also a problem with completely driverless buses. Not fare dodging, damage etc.

I've discussed this subject several times in the thread.

Full autonomy — not merely driver assistance systems, no matter how fancy — available on a majority of vehicles would revolutionize land usage along with vehicle ownership. It's hard to understate just how big this would be, but a very brief list might include...

• End of private vehicle ownership for most people.
  • Vehicles are not assets for the majority of consumers; they are liabilities.
• You would summon rides on regular schedules or on demand.
  • For example, instead of getting in your car at 8 AM to go to work, you'd be picked up at your front door/apartment entrance/etc.
  • Regular users could pay a subscription for the service.
• The need for garages (and even driveways) in most single-family homes would disappear.
• Commercial parking would radically change. Instead of leaving your car there all day, you would be dropped off and then picked up later.
  • Parking lots would dramatically shrink in size. Areas previously dedicated to parking would be reconfigured for drop-off and pick-up lanes.
• Utilization rates of vehicles would increase enormously, meaning we'd need fewer total cars to serve the same population needs. This would confer substantial environmental savings by itself.
  • By balancing fees around demand, peak (rush-hour) traffic could be smoothed out — although we've heard that one before and it remains the major rational objection.
• Individual stress from commuting would likely be reduced significantly, a tangible public health benefit beyond merely lives saved due to reduced accident rates.
  • We might even see productivity increases as people can work in their cars.

As I said, for most people a car is not an asset. Between depreciation, interest, insurance, registration, maintenance, parking/garaging, fuel, and opportunity cost on capital, it is something you are obligated to have because you have no other reasonable way to get around, not an investment. Robo-taxi service should be substantially cheaper on net for almost everyone.

Obviously, many people utilize personal vehicles for business purposes, and they would continue to do that even if the cars themselves are autonomous. People who buy them for the purpose of remanding them to robo-taxi fleets could of course do so as well.

Edited by Fighteer on Apr 25th 2024 at 12:16:23 PM

Eh, at least were I live, bus drivers of standard city busses don't check if you paid your fare. The problem is, the larger the vehilice, the less driver salary factors into the operation costs, making automation less usefull. This is why there is little hype in automating trains, despite this being easier.

Edited by Smeagol17 on Apr 25th 2024 at 7:05:47 PM

However, as we are finding out in Switzerland and Germany, drivers can be a liability in terms of strikes, people not being around when they should be etc. It's not all about salaries.

Most peak demand isn't caused by consumer choice, remember? Increasing fees doesn't do all that much to stop demand. -Continues staring at peak rail fares.-

Because what the world needs is more landlords contributing nothing, while people have to work even more because now "I'm travelling" isn't a restriction. I'd rather have the environmental disaster than an even more dystopian future, but with overly complicated taxi systems.

I feel robo taxis might end up not being a huge game changer.

At the very least, the people who will get the most benefit out of them are people who are already less much less likely to own their own vehicle.

I can only speak to the US, but owning your own vehicle is a pretty major status symbol and I do think there is a difference between relying on a robotaxi network that will fail for reasons completely outside of your control and relying on a vehicle that you own and therefore have more influence on whether or not it fails.

Flash forward 3 or 4 generations and who knows. But making any sort of absolute predictions on that kind of time horizon is an exercise in futility. History is full of trends that seemed inevitable only to reverse course unexpectedly.

Yes, we know employers are shit. Blah blah. Autonomy doesn't change that; it just shifts some numbers around. "Accelerationist" arguments like this can go to hell as far as I'm concerned.

As for the fleet ownership thing, there may be some advantages to individuals buying cars to rent them out in the near term, but once the service becomes truly widespread, competition should drive down profit margins to the point where only large-scale operators can stay in the market. This is mainly because small operators wouldn't be able to take advantage of economies of scale in garaging, maintenance, charging, etc.

You can't have it both ways: exalt small business owners over large corporations but then sneer at "landlords". Anyway, governments would probably step in to regulate robo-taxi operation so that consumers don't get shafted.

A lot depends on how fast these vehicles can achieve market penetration. Right now it's just a tiny fraction of cars (cough Teslas) that could in principle become autonomous once the software is ready, assuming that their approach works as intended.

In the interim, people who own AVs would gain the direct benefits, either by using them themselves or seconding them to the growing fleet services. You're absolutely correct that the transformation would not happen overnight, but there is a future generation coming up who will never learn to drive.

Edited by Fighteer on Apr 25th 2024 at 2:27:12 PM

Edit: Hollered as a forum rules violation.

Anyway, what I said is that private rent extraction will eventually hit diminishing returns and become unprofitable against the scale advantage of large fleet operators.

Edited by Fighteer on Apr 25th 2024 at 1:00:53 PM