Waymo outperforms comparable human benchmarks over 7M+ miles
waymo.com
waymo.com
> In August, a Waymo at an intersection “began to proceed forward” but then “slowed to a stop” and was hit from behind by an SUV.
> In October, a Waymo vehicle in Chandler, Arizona, was traveling in the left lane when it detected another vehicle approaching from behind at high speed. The Waymo tried to accelerate to avoid a collision but got hit from behind.
It’s worth noting that all 3 of these incidents involve a Waymo getting hit from behind, which is the other driver’s fault even if the Waymo acted “unexpectedly”. This is very very good news for them.
Of course, anecdotal evidence isn't a very high standard. Thankfully, statistics on this sort of thing are tracked. Statistically, the Tesla self-driving features reduce accidents per mile. They have for years now and as the tech has progressed the reduction has grown as the technology has matured. So statistical evidence also indicates that the person you are asking the question to is also uninformed.
What is probably happening is that it makes for good clickbait to involve Elon and Tesla into discussions. Moreover, successful content online often provokes emotion. The resulting preponderance of negativity, especially about each driving accident Teslas were involved in or caused, probably tricked them into misunderstanding the reality of the Tesla safety record.
While that is the claim, I've never seen an independent analysis of the data. There are reasons to believe that Tesla drivers are not average. I don't know if what claims are true, which is why I want independent analysis of the data so that factors I didn't think of can be controlled for.
I know YouTube videos aren't always representative of reality, but there are some videos of these cars randomly driving extremely slowly in very busy intersections which might be a contributing factor to getting rear-ended, even if it's not Waymo's "fault" from an insurance perspective.
Most human drivers I observe are taking huge chances every single day. I see them drive at speed around corners they can't see around fully, driving far too closely to cars front, using their phone etc. They get lucky but it's only a matter of time before they have accidents. The three incidents recorded here are simply due to three humans' luck running out.
Nah, I'm jumping straight to this conclusion: if you hit something in front of you that didn't leap out in front of you at the last second, you fucked up. The object that you hit was erratically slowing and speeding up? You should have left more room to allow for the unpredictability. It was raining, can't see, the roads are slick? Leave more room in front of you.
Yes, I way too often don't do that, either. But if there's something in the road ahead of me, and I hit it? Man, there are few scenarios where I can claim that there was nothing I could do. And in the case of a full-sized vehicle in front of me, I don't care how erratically it's driving, don't run into the back of it.
It's not causing accidents in these cases. The humans are.
So while your statement isn't wrong, it's also not always pragmatic in the real world.
If not, then you already recognize the probability is less than 100%, so that has to be baked into your statement.
The only comments saying that are...yours. If your argument is so lacking that you need to argue in bad faith, perhaps it is best to not bother at all.
Or perhaps you are not aware of the proper following distance (and therefore, part of the problem about which you complain). Two car lengths (EDIT: seconds, not car lengths; oops) is the general advice.
>Two car lengths is the general advice.
No. It's speed, roadway, and car dependent. Two car lengths isn't even sufficient at 25mph let alone at 70mph.[1] Which all goes to show how poorly people tend to think about these things and quickly resort to overly simplified mental models.
[1] https://one.nhtsa.gov/nhtsa/Safety1nNum3ers/august2015/S1N_A...
Moreso that my post has a mistake: two second following distance, not car lengths. Brain fart on my part; apologies for causing you to have to find a URL. But as general rules go, that increases the distance as speed goes up. No, it doesn’t account for everything, but good enough for most circumstances.
Indeed - this is the 70mph is slower than 50mph thing I mention.
Roads like that are currently operating at bursting point. There are incidents and accidents every single day and constant police presence is required to unblock them. If you alleviate congestion, more people use the road. They just go back to bursting point. In other words, it's utterly insane.
Can you imagine if there were accidents on railways or in the air every day? Imagine the scandal if train operators were found to be unsafely squeezing more trains on to the line that it could handle. Roads are stressful, inefficient and shit. Enforcing a safe stopping distance and pricing journeys accordingly, like trains, is where we want to be.
I don't disagree about "pricing journeys accordingly," but there are many reasons why this is difficult in practice in the US. Going through those points is a bit of a digression from my main point. Namely, that there are pragmatic tradeoffs that have to be considered. I'm consistently taken aback by the amount of "simple" solutions people advocate on HN and sometimes I wonder if it's due to software developers constantly working with the abstract rather than the concrete.
Like most real-world engineering, there is a cost-benefit balance. Could we design an interstate highway system that allows everyone a 10 car buffer? Sure. Would we like how much it costs, the effects on the environment, etc.? Probably not.
As an aside, your comment seems to go against HN guidelines by taking the least charitable interpretation of the comment.
You might have forgotten the ultimate rule of road safety: Everything is a tradeoff. Safer is only sometimes better. Otherwise the speed limit would be 10mph, because it's quite a safe speed.
You do not have to raise the average speed of travel very much to make up the theoretical loss due to increased spacing
Moral: support public transit.
The only time it can be not your fault is if the slow moving car switches lanes in front of you before there is time to stop.
His actual statement was, "if you rear-end someone that you are following, it is your fault..."
If someone abruptly pulls in front of you, you weren't following them.
Related to this is glare from the sun or artificial sources. I lived in a small city with antique-style globe lamps on Main Street. The veiling glare made pedestrians invisible, and even if you knew about the glare and watched for pedestrians, you would still be surprised when they became visible halfway across the street in front of you.
I'm nearly certain that I'm alive today because I drove as defensively as something like a Waymo. One day as I was approaching an intersection where I had a green light, I saw a car approaching the same intersection from the cross street at a speed that they couldn't have possibly stopped for their red light.
I instinctively slowed down suddenly and that car, as I predicted , ran the red light at high speed and turned just a few yards in front of me.
If I had been more tired, hasty, or it had been darker out, I might not have seen it or reacted in time. A fully autonomous and defensive driving system wouldn't get tired or hasty, and lidar can see fine at night.
And yes, I might have gotten rear ended, but that's a far better outcome than getting t-boned by a car going 65mph.
If you had just been an average driver. Most people are terrible at defensive driving.
Mostly because people misunderstand what defensive driving means.
It doesn't mean "go slow". It means always maintain a safety margin between your vehicle and others, adjusted for speed, road/weather conditions, and relative direction of travel.
Those are the kinds of objectives a self driving system can very reliably achieve, unlike a human.
If Waymos were having a seriously increased rate of non-fault crashes, that would still be a safety issue, even if every crash was ultimately a human's fault.
As another commenter mentioned, the fact that the Waymo detected that a vehicle was approaching it from behind at high speed and tried to accelerate to get out of the way is super impressive, and I'm not sure even most good human drivers would have been able to do that.
I also wonder how it fares in a Kobayashi Maru scenario. I chose to cream a construction cone because the guy in the left turn lane went straight. (Admittedly, I think he didn't realize he was in the left turn lane.) I could see the cone wasn't actually protecting anything, could a car do so?
Thus reminds me of highways when you can see a back up a quarter mile up. I tap the brakes about 5 times before letting off the gas to cut speed by 20% before actually even braking,which again is preceding by taps on the brake.
So. I don't think the majority of these cases were even tailgaters. I would love to see the data of what the following distances were, rate of deceleration, how much space there was for the lead car to avoid the obstacle. Ade these Waymo's cars just out there brake checking people? Are they waiting to the last moment before doing emergency brake mode?
At the end of the day, drastically changing speed is up there for dangerous behaviors, human drivers go to great lengths to avoid it and signal the intent. Turning right or left off of a 50 mph road- better believe I'm signaling that turn 2000 feet away and decelerating in stages and incrementally. With luck even that vehicle 4 cars back in the convoy is going to have lots of time.
> The safety driver unwittingly turned off the car’s self-driving software by touching the gas pedal. He failed to assume control of the steering wheel, and the Pacifica crashed into the highway median.
Why are these not counted? Are they really looking at their car crashes, or just autonomous driving software being in control during those car crashes?
Maybe they want to argue the software is safe, but that doesn't change the fact that I'd still be scared of getting into that car.
[1] https://qz.com/1410928/waymos-self-driving-car-crashed-becau...
This is missing the point. Just because we accommodate certain types of errors by humans, that doesn't mean should or would tolerate them if the same errors were made by machines. (Or vice-versa, for that matter.) The standards we hold machines and humans to can be very different, and that should be expected.
As a rider, you can't touch the Waymo steering wheel or pedals, which eliminates the cause of the accident you referenced.
That might make sense but this is obviously a little tricky to implement safely.
It would be interesting to know how often humans have to take over from Waymo, and also how often humans who’ve taken over from Waymo get into accidents.
I mean, it is not really realistic, but hypothetically one could imagine a driving strategy that has a low expected number of crashes, but puts the car in an easily detectable, strategically bad position, and then just dumps that position on the safety driver. Of course I’m sure Waymo wouldn’t do anything like that because it is, like, the most obvious bad-faith strategy which would be caught on a rigorous review and would be ruinous for the company.
I do not trust any data in 2023, and 2024 will be worse.
In his defence, I did get to the hospital well in advance of my appointment. It reminded me of the old Sega game Crazy Taxi.
However, most ticket issuing in London is by automated cameras, and Taxi drivers know where all the cameras are.
It was extremely uneventful - in a good way - while navigating urban traffic, road construction, unprotected left turns, etc., and felt (subjectivity alert!) a lot safer than many of the rideshare drivers I've ridden with over the past 8 years in LA.
I would definitely do it again and would feel safe putting a family member in one.
Why this car, and not all cars? If I fail to assume control of my steering wheel, I will also crash.
The short answer is because this isn't a deterministic "if". The probability matters too.
The thing is in a normal car you're forced to be alert all the time. With autonomous driving, 99%+ of the time you have nothing to do. Humans simply cannot pay as much attention all the time when they're not actively forced to. It's much easier to lose attention (drowsiness, chatting with people, etc.) than if you're physically already driving.
And moreover, regaining control of a car requires some context switching time that isn't there when you're already in control.
If my car is going to disengage and at a random point (whether due to my fault or otherwise), I'd rather just be in control the whole time.
But it's not at a random point. If you push the disengage button, it disengages and tells you it's disengaged. I don't understand why you keep mischaracterising what happened.
I'm not mischaracterizing anything. To the driver who dozes off and presses this completely unintentionally, this did happen at a random point. He certainly didn't intend or expect it to happen.
You have to realize, dozing off in a situation that's boring 99%+ of the time is a human thing. If you design your car such that a driver is prone to pressing the disengage button unintentionally when he dozes off, you get to share the blame when that happens. It's not something reckless like DUI where you get to put all the blame on the driver for it.
That's not the only perspective, though. It's possible to doze off when driving a normal car, and that doesn't suddenly become the car's fault, despite from the driver's perspective, he/she didn't mean for it to happen.
> If you design your car such that a driver is prone to pressing the disengage button unintentionally when he dozes off
But again - how do you know it's been designed that way? This still seems like editorialising without any additional info.
I recently rented an even later-model Malibu that only had collision warning auditory alert. Better than nothing, but I'm surprised cars are still made without automatic braking.
In the EU, at least, since May 2022, all new cars do have automatic emergency braking, along with intelligent speed assistance; alcohol interlock installation facilitation; driver drowsiness and attention warning; advanced driver distraction warning; emergency stop signal; reversing detection; and event data recorder (“black box”).
Other features like eCall – a built-in automated emergency call for assistance in a road accident – have been mandatory since March 2018.
That's far better than the situation in the US, where private services like Tesla, GM with OnStar, or Ford with "Sync with Emergency Assistance", which have no limits on data collection.
Shame. I've never had it work really reliably in any car, it's a feel good but mostly shit. Even more so when it's not even hooked into cruise (many cars will provide a shortcut to copy the sign's speed into the cruise or speed limiter, but far from all of them).
The Obama administration (2015) was able to successfully negotiate with and convince most major car manufacturers to voluntarily agree to start making new cars with automatic emergency braking. Their agreement stipulated that all new cars must have it by 2022 [1]. But this negotiated agreement is why we started to see some new car models include it post 2015.
The tl;dr is the Obama administration basically said "look, if y'all don't agree to these proposed minimal standards, we'll get congress to pass a law that is more strict. So the companies decided to take the agreement now to de-risk themselves from having to comply with potentially more stringent requirements in the future).
[1]. https://www.nhtsa.gov/press-releases/us-dot-and-iihs-announc...
Granted, I'm sure this will improve over time. But for the past 5ish years, all my experiences with auto-braking have been dangerously negative.
A system that takes driving too close and turns it into something more dangerous is not a good thing.
Yes, but...there is something else to be said here. One of the things we have evolved to do, without necessarily appreciating it, is to intuit the behavior of other humans through the theory-of-mind. If AVs consistent act "unexpectedly", this injects a lot more uncertainty into the system, especially when interacting with other humans.
"Acting unexpectedly" is one of the aspects that makes dealing with mentally ill people anxiety-producing. I don't think most of us would automatically want to share the roads with a bunch of mentally ill drivers, even if, statistically, they were better than neurotypical drivers. There's something to be said about these scenarios regarding trust being derived from understanding what someone else is likely thinking.
Edit: the other aspect that needs to be said is that tech in society is governed by policy. People don't generally just accept policy based on statistical arguments. If you think that you can expect people to accept policies that allow AVs without addressing the trust issue, it might be a painful ride.
I've got some bad news for you.
1. https://www.nimh.nih.gov/health/statistics/mental-illness
It remains to be seen if autonomous driving systems are actually safe. But if the other driver does something that is safe, there's then an onus on the first driver to have accounted for that.
I don't disagree, but in order for the first driver to "account" for the actions of the second, they have to have some reasonable ability to predict what that driver will do. That gets us back to the theory-of-mind question.
If you don't maintain a safe following distance for your speed, you are the one creating the dangerous driving environment. Tailgating is worse for both traffic flow and safety.
Tailgating is bad, regardless of if people brake check or not. If automous vehicles are what it takes to get you to stop tailgating and follow at a safe distance, then that is just an added bonus.
Why, for example, do you think trainers post "Student Driver" stickers on their cars? It's because it signals the driver may be more unpredictable and people (rightly) tend to give them wide berth. You're essentially advocating that everyone treat everyone else (and every robot) as a student driver. That's fine for a dichotomous safety mindset, but other people would prefer to recognize the tradeoffs with that approach.
Or maybe you're just deliberately bent on misunderstanding my point, I can't read your mind :)
What is this nonsense? The safe following distance is determined by how fast you can stop, not by who is driving the vehicle you are following.
> Why, for example, do you think trainers post "Student Driver" stickers on their cars? It's because it signals the driver may be more unpredictable.
No, it signals they have less experience and are more dangerous drivers. When it comes to driver predictability, student drivers tend to be far more predictable than the adult, overconfident drivers. I've never seen a student driver roaring past me in stopped traffic on a shoulder, or floor the gas to pass me through a light because they didn't want to turn in a turn only lane, or any of the other unpredictable things I see on a regular basis from experienced drivers.
> You're essentially advocating that everyone treat everyone else (and every robot) as a student driver. That's fine for a dichotomous safety mindset, but other people would prefer to recognize the tradeoffs with that approach.
I think that if more driver treated the people around them as student drivers, our roads would be a lot safer.
I know that if people followed at a safe distance then we would have fewer traffic jams.
Edit: You also seem stuck on the idea that Waymo unsafely unexpectedly brakes more often than human drivers, yet that isn't clear to me from the data we have. Indeed it seems like the opposite is true from the data.
So when you're driving, do you somehow know the braking distance of every car and reaction time of every driver around you? You don't, and since their braking distance is needed to know your own braking requirements, you have to use heuristics. Maybe your heuristic is "assume everyone will cram on the brake, full tilt, at any time." But, that is not a pragmatic solution given our current infrastructure. We don't have the road capacity for everyone to drive that way. So we make tradeoffs. Part of that tradeoff means anticipating what other drivers will do and adjusting accordingly. Naturally, this will trade some safety for other things we value. That is the reality of the world we live in. You seem to be advocating for something else. The OP was that we might struggle to apply such heuristics without a theory of mind to guide us.
We probably just disagree on the student driver vs. overconfident drivers. I feel like I'm pretty good at anticipating aggressive drivers, and I fear them must less than the super-tentative driver that tends to put other people at risk. But unless you have data, we're just talking about subjective opinion here so it's not really worth delving into further.
>I think that if more driver treated the people around them as student drivers, our roads would be a lot safer.
Sure. But again, it doesn't really fit with the world we live in. Should we all, in general, drive more defensively? Sure. But I doubt our infrastructure will allow for 25+ car lengths between vehicles that the NHTSA recommends, so we're stuck making some tradeoffs.
I agree on the data point. I'm not making strong claims about safety. I'm making claims about uncertainty. One thing that is clear (and I've advocated elsewhere) is that we don't have good data (in part, because companies get to share only what they want in many cases), which makes uncertainty greater.
25+ car lengths between vehicles that the NHTSA recommends
Note: 75 mph is 120 kmh
You can (usually) follow a large semi a bit closer because its braking distance is longer than yours.
But because of reduced visibility you can end up with “revealed brake checkmate” where the semi swerves into the next lane because a vehicle is stopped in the lane, which you then need to swerve or hit.
It’s more effective to let your initial point stand and let the discussion run its course. You’re working against yourself now.
More following distance => less throughput. There is already a math problem at play for how to get millions of single occupancy to and from a given location in a day. Which is to say, gridlock occurs today - it is a fact. Increasing following distance by 2 to 5x would potentially be quite bad for otherwise heavy but flowing traffic. Eg: NYC, it is known for good throughput. Light turns green, everyone picks up speed and keeps tight following distances. Anything less and that city would lock up more so than it already does.
Part of the issue is just complexity and variety. 2x following distances on most highways would likely be only a good thing. But in each and every circumstance, not necessarily good, perhaps disaster.
I think it can fun to be the driver that smooths out the shock waves, so I do largely agree with your point. Though, all things being equal, double following distance and that will double the length of a traffic jam. Longer traffic jams are more impactful - it takes longer to get through them & the longer traffic jam distance is more likely to cause back-ups on on-ramps and in turn to the surface streets leading to those on-ramps. Which is to say, a higher prevalence of grid-lock.
The AI crowd is right to say that with 100% AI cars, you can have smooth and fast traffic with impossibly tight following distances. I do believe you need some impractical necessities to accomplish that though.
You keep making the same argument over and over even though I've repeatedly explained that it does not match our understanding of how traffic jams form. Traffic jams are caused by braking, especially hard braking. Tailgating increases the need to brake hard if the person in front of you brakes or someone needs to merge. Following closely does not increase throughout, it is simply bad driving with no upside.
> I'm pretty good at anticipating aggressive drivers, and I fear them must less than the super-tentative driver that tends to put other people at risk.
Wow, if you read the behavior I descrived as "agressive driving" then we have completely different standards. I was describing reckless driving that blatantly violates the rules of the road.
Stop using excuses for your bad behavior and you might even be able to become a better driver.
>Traffic jams are caused by braking
Guess what one of the main issues has been in AV...nuisance braking. So much so that they suppress safety-critical actions to avoid unwarranted braking.
I also think you're confusing the proximate causes of conjestion for the root causes. Traffic conjestion is a load sharing problem. Tailgating is, in part, a symptom of inadequate capacity. You are advocating a solution that exacerbates it by reducing carrying capacity.
To reiterate (yet again) we aren't in disagreement about whether slowing down, or increasing following distance, will increase safety. It will, but also, that's not the point I was making. I'm just saying that is a superficial understanding of the problem and you aren't accounting for the tradeoffs. Those tradeoffs are the reason your proposal misses its mark.
Why are we still debating something that has been answered by science? Stop assuming your intuitions are correct and look at what the science actually says about how follow distances affect the maximum capacity of a road.
I'm willing to bet that's around 10 times what you were considering as a safe following distance in your head, and probably still 5 times more than what you were picturing for the safe distance behind a brake checking AV.
As more and more cars default to this safely, however, it’ll start to equal out.
That said, if you routinely tailgate the driver in front on the assumption that nothing will go wrong then you've chosen to accept the consequences when an otter runs out in front of them (hey I didn't see that coming either) and they suddenly limit brake and you're sitting in their back seat. Or a steering tie breaks in a classic car coming the other way, which hits the car in front of you head on and now they're stationary.
The question of how much additional caution (in terms of lower speed limits, longer following distances etc.) is optimal in terms of overall QALYs is, I feel, vastly under-considered and under-discussed.
Funny story, I actually "crashed" in a hot air balloon as a kid, when the hill we were landing on had a draft running up over it that caught the balloon after the basket had touched down, and dragged us along the ground sideways for a good quarter mile.
This is half remembered from a Snowmass balloon rally conversation.
Basically if on the same level the one on the other’s right has the right of way.
But a plane in distress has right of way over the balloon.
I don’t think you can make this strong of a claim with the available data. Likewise, someone can’t make strong claim in the opposite direction. The best data I’ve seen is from NTSB investigations, and it clearly shows some very dangerous behavior. But it’s just a snapshot of data.
And I think you’ve taken away the wrong point. It was about unexpected behavior, not “driving with a death wish.”
The ability of people to consistently miss/twist the point to fit their own predetermined viewpoint is tiresome.
Indeed - so how did systems with a "death wish" enter into this discussion? Well, it turns out it was by you, about which you said "...maybe not so good on a public road."
In this light, your complaint about others twisting the point seems rather ironic.
It was the way the pilots described the unexpected behavior of the RL model.
If I related a story about ChatGPT where someone said, "It wrote like it was drunk" would you insist that I'm saying ChatGPT actually imbibes in alcohol before coming up with a response? I think you might be missing the point of an analogy. This tends to correlate with people with dichotomous thinking, which is also part of the thread and the difficulty with people understanding the intended point.
The part you missed was that the "maybe not so good on a public road" was about breaking "best practices," not about acting with a death wish. The intent was to underscore, yet again, was that unexpected behavior is sometimes beneficial in a wartime environment where you want to keep the other party guessing and off-balance, but not beneficial in a public safety domain. Again, twisting an argument to make a preconceived point rather than reading it as it was actually written. It's hard to get someone to understand a point when their biases are hell-bent on deliberately not understanding it.
And you chose to quote it, and to make the point that it is not the sort of thing we want on public roads - as if there's the slightest hint in the extensive testing of these cars that the apocalyptic scenario you chose to introduce to this discussion was anything but hypothetical and hyperbolic FUD.
And then you had the effrontery to chide dcow for responding to your claim, as if they were taking the discussion in this absurd direction. You've chosen to repeat yourself here, so I will say again that there is considerable irony in what you are doing - irony in the sense of statements being made that display a lack of self-awareness.
> And you chose to quote it, and to make the point that it is not the sort of thing we want on public roads - as if there's the slightest hint in the extensive testing of these cars that the apocalyptic scenario you chose to introduce to this discussion was anything but hypothetical and hyperbolic FUD.
At some point there were numerous reports of cars using driving assist steering into the dividers at highway offramps. I believe there were in fact some some real accidents that happened this way. That definitely was unexpected behaviour and I would argue one could even qualify it as the car acting like it had a death wish so I don't think the OP statements can be qualified as hyperbolic FUD.
Now you're going to argue that these were early incarnations of non AV systems (despite the name), but I think they do illustrate how the systems can behave in unpredictable (and dangerous) ways when they encounter novel situations. That's why I commend waymo for not following the hype and keeping the environment they operate very restricted.
Well, yes I am, because, as you realize, it is a valid point! I will also add that, I have, myself, made the point that we have to be careful because complex systems have bizarre failure modes, so just because, in everyday circumstances, these systems may appear to function 'sensibly', we cannot simply trust them to be 'sensible' outside of the scope of testing.
I am, however, a believer in the relevance of empirical evidence, and I also agree (with some caveats) that human-driver performance is a valid basis for establishing whether, and to what extent, autonomous vehicles may be permitted on public roads. We are discussing the publication of some results of Waymo's extensive testing, and I stand by what I wrote in my first response here: none of these three cases involved the Waymo car behaving in ways that are not that uncommon among human drivers, and our theory of mind does not make us nearly-infallible predictors of what another driver is going to do.
I agree with you that Waymo is taking the right approach here (and, FWIW, I regard Tesla's two-faced stance as unethical.)
I doubt I would have objected to Bumby's invocation of a death wish in response to the past events of which you speak, but in response to Waymo's test results and the points I made about them, I think 'hyperbolic FUD' is justified. At some point, a rational person has to make accommodations when things have changed and arguments based on old data lose their relevance.
This bypasses the entirety of the trust argument originally stated by leveraging the very point (statistics) it cautioned against. (As stated elsewhere: "So all the bleating about statistics may be necessary, but not sufficient, to get wide-scale adoption of AVs on public roadways.")
>none of these three cases
Regardless if you think the decision should be based on statistics alone, my further point is that an n=3 sample size is not adequate to make strong claims. Add to it that Waymo only reports the data they want to[1], there is a reason to be careful about making any claims.
Here's how that scenario plays out when we rely on organizations to self-report safety data in a competitive environment, from my experience. Even though it may start with the best of intentions, cost and schedule pressure builds. Now items that would have otherwise been reported as safety incidents are now classified with vague terms like "test anomalies" and essentially buried. They will still report safety metrics, but now it's, at best, incomplete and misleading. Until some event that's egregious enough forces the company to be more transparent.
[1] https://spectrum.ieee.org/have-selfdriving-cars-stopped-gett...
We know they aren't fully transparent. We also know that other well-known and well-funded AV developers have very bad practices that are highlighted when they are forced by regulators to be transparent. While that isn't a smoking gun against Waymo, it should give us pause and make one question a naive perspective in favor of a skeptical one.
You made two points right next to each other, about sample size and bias, and the sample size point was invalid.
When called out on that, you don't get to move the goalposts and say that the second one was "the point" and the other person is "missing the point".
Given the protracted nature of this thread, I get why it's confusing. I have a couple sub-threads that make those two points separately, and was simply trying to show how both are related. If one isn't aware of the broader context of the discussion I was making, I understand why it may seem like goalpost moving. But in reality, I was deliberately trying to tie the two related posts together because (IMO at least) they are related.
If your absurd claim that Waymo's trials provide just three relevant data points is not part of "the point", then why did you make it? It does not give us any confidence in the proposition that "the point" has been well thought-out.
Furthermore, "the point" keeps shifting: recently it shifted to raising doubts about the provenance of the data whipped up from a six-year-old article. At this point, I feel that a quote is appropriate: "the ability of people to consistently miss/twist the point to fit their own predetermined viewpoint is tiresome."
And in the vein of trying to steel-man your position, I gave the comment to ChatGPT to see if it, too, considered the central point a claim about AV having a "death wish." Here's what it said:
"The statement highlights that some AI tasks, while effective, may deviate from conventional practices. An example is given where the Department of Defense (DoD) employed a company to train a dogfighting simulator using reinforcement learning (RL). Pilots were surprised by the simulator breaking established best practices and behaving recklessly, akin to a pilot with a disregard for safety. The implication is that while such behavior might be advantageous in a military context, it may pose risks or be unsuitable in civilian settings, such as public roads. The statement underscores the need to carefully consider and tailor AI applications to specific contexts and objectives."
So it seemed to recognize that the central point is that "the behavior" in question is "breaking established best practices" and that the "implication is that while such behavior might be advantageous in a military context, it may pose risks or be unsuitable in civilian settings". There's probably some irony in the fact that AI did better at a reading task.
Really the only thing left is for you to take a flight to SF and watch the Waymo cars drive. Or ride in one if you dare.
This dispute is not over a difficult or subtle issue: all the people who have responded in this thread see clearly the obvious and unequivocal reading - and, in your own example, ChatGPT also does! It has identified the tacit subject of the sentence "Possibly good in war, maybe not so good on a public road" as the specific military system, with behavior explicitly described as like having a death wish, that is the only topic of the preceding sentence. For one thing, the phrase 'possibly good in war' makes no sense in the reading you are trying to pass off: why, out of nowhere, did the needs of the military appear? And unexpected behavior is not something desirable in general in military systems, any more than elsewhere - it would take very special circumstances and a specific sort of behavior for that proposition to even be entertained. We can see, therefore, that dcow was right to respond 'your comparison is RL dogfighting?...' and 'Modern AVs are not driving like they have a death wish by any stretch of the imagination...'
Oh, and next time you invoke ChatGPT's response, include the prompt, verbatim, like this for example:
Prompt: In the statement 'That’s true, but also one of the selling points of some AI tasks. As a non-hypothetical example, the DoD hired a company to train a software dogfighting simulator with RL. What surprised the pilots was how many "best practices" it broke and how it essentially behaved like a pilot with a death wish. Possibly good in war, maybe not so good on a public road.', what is being called ' not so good on a public road'?
Response: In the statement, the phrase "not so good on a public road" refers to the behavior of the software dogfighting simulator trained with reinforcement learning (RL). The implication is that the simulator, which exhibited behavior contrary to conventional "best practices" and behaved like a pilot with a "death wish," might not be suitable or safe for use in a public road scenario. This suggests a concern about the potentially risky or unpredictable behavior of the AI system in a real-world, civilian setting such as driving on public roads.
What we have in this discussion is a motte-and-bailey fallacy, as we can see in your response to my first post here, which was:
None of these three cases involved the Waymo car behaving in ways that are not that uncommon among human drivers, and our theory of mind does not make us nearly-infallible predictors of what another driver is going to do. Your objection becomes essentially hypothetical unless these cars are behaving in ways that are both outside of the norms established by the driving public, and dangerous." Your reply, in outline, goes like this:
> That's true...
Here we are in the motte, where you nominally accept that the relevance of your concern, which is not unreasonable in itself, is constrained by the extensive testing that has been performed by Waymo so far...
> ...but...
Here we enter The bailey, where we are supposed to turn our attention to an unrelated system, which was found, on testing, to have alarming unexpected behavior. The bailey has become a place where Waymo has been curating the data to the point where we simply don’t have the slightest idea whether there’s dangerous behavior outside of the human norms lurking in Waymo cars.
It has also become a place where all of Waymo’s extensive testing has produced just three data points against this view. I must say that it seems generous of you to concede even three, if Waymo is curating the data to the extent you imply.
This is where you are bypassing the point about how much faith we can put in the data, because we don't have the full results of their extensive testing. We only have the results they are willing to share.
As an analogy, I have a relative who loves to talk about all the times he's won money gambling on the craps table. I almost never hear any information about his losses, unless they are couched to say how much more money he's won later. So would you say I can conclude he's an expert gambler who should quit his job to play craps full time, or do you think there might be some human bias in reporting going on that I should be skeptical about?
At least the idea that Waymo is hiding data that would reveal the vehicles as being too dangerous to be on public roads is not quite that wrongheaded, but it implies that Waymo is doing this on a massive scale that cannot possibly succeed in the long run. It is not clear, for example, how it could hide information about similar or worse incidents from the insurance companies, and a story about your gambling relative is not changing my mind.
This is your dichotomous thinking again. I am not making a point about it being "safe" or "unsafe" as a dichotomous choice. I'm saying we don't have good data to make a claim one way or another. That's also different from saying, "we have no information at all." When you combine that with the fact that we have other evidence that RL models can result in unpredictable behavior, it should give us pause.
Insufficient data + priors about unpredictable behavior = uncertainty of performance
The through-line of my entire point is that uncertainty erodes trust, and trust is necessary for wide adoption in the public sphere. It isn't a hard concept if you can lay your bias and dichotomous thinking aside to consider it. Waymo seems to agree; although the PR headline is about 7MM miles traveled, the report says:
>"the required ADS VMT to establish statistical significance ranges from tens to hundreds of millions of miles, and the fatal outcome requires hundreds of millions to billions of miles of driving are needed."
So unless you have better data to show, you've demonstrated nothing to make me change my mind.
You don't get a thread of 173 posts (44 by you) over 6 days and counting (4 in the last day) over the anodyne position you are now professing. That time has been spent dealing with all the bogus stuff that you are now tacitly disavowing, such as the bizarre three data points claim that you brought up again as recently as your previous post.
> ...change my mind.
You have been tacitly changing your mind throughout, though, of course, you will not admit it.
It certainly comes across like either you're too hung up on a position to read critically or you got fooled by a PR piece without knowing how to properly interpret it. I guess that's a win for the SV hype machine.
On the one hand, I can understand that it is difficult to keep track of everything you have said throughout your twists and turns, but on the other, it seems particularly important one to keep this one in mind as, if verified, it would be far more damning of Waymo than everything else you have said, combined!
I've replaced your word "it" to make the point as clear as possible.
>"And you chose to quote it, and to make the point that" [unexpected behavior] "is not the sort of thing we want on public roads"
Or to put it different terms, the sometimes unpredictable behavior resulting from RL may be a feature on the battlefield but a bug on public roadways.
I still stand by that point. And, yes, that means dcow also missed that point. There's nothing hypothetical about bringing about real-world case studies of autonomous behavior based on RL models. We've been through this so many times now that I'm coming to the conclusion you may be arguing in bad faith or you get incredibly distracted by certain terms that it inhibits reading comprehension.
>"And you chose to quote it, and to make the point that" [unexpected behavior] "is not the sort of thing we want on public roads"
Ah, so we're playing the "I didn't say what I just said" game now. Here's what you actually wrote:
"That’s true, but also one of the selling points of some AI tasks. As a non hypothetical example, the DoD hired a company to train a software dogfighting simulator with RL. What surprised the pilots was how many “best practices” it broke and how it essentially behaved like a pilot with a death wish. Possibly good in war, maybe not so good on a public road."
Anyone with the slightest familiarity with language will find no credibility in your claim that in this comment, the thing that is being called "possibly good in war" but "not so good on a public road" is anything other than the one explicitly-mentioned aspect of the system that you have specifically chosen to present as an example. Furthermore it was completely reasonable for dcow to respond "modern AVs are not driving like they have a death wish by any stretch of the imagination" after you chose to make the above statement.
With your latest response, you continue to put more weight on an anecdote about this unrelated system than you do on the extensive empirical evidence from testing the actual system that is the subject of this article.
They're telling you what you wrote, not what you intended to write.
They're correct.
> The "explicitly mentioned aspect" is the unpredictable behavior.
No, it was breaking best practices and behaving like it had a death wish. That's something that has some overlap with being unpredictable, but is not at all the same thing. You can have a predictable death wish, even.
> maybe your biases made you infer more than what was actually said
I don't think I have much bias here and I agree with them. Also consider that the person that writes something is biased to think the communication was clearer than it actually was.
> And not to beat a dead horse, my secondary point has consistently been
It's neat to have a correct secondary point, but it won't make your primary point correct, and people don't need to add a disclaimer of "while your secondary point is fine" every time they criticize your primary point.
Now I'll concede it's possible that RL can lead to finding new best practices. But that's not the case in the story relayed. It is only a good practice for the scenario because it was an unpiloted aircraft because it otherwise puts the pilot at too much risk (which is why the pilot say it was acting "as if" it had a death wish. He was anthropomorphizing it.) That still means it's bad practice when the goal is to save human lives.
But that's not the same as being unpredictable. Those are orthogonal complaints, and nothing in your quote suggests that the "death wish" flying was more unpredictable.
And we already know that these cars act super cautious, the opposite of acting like they have a death with.
So that quote has no useful information that we could transfer to the car situation.
What I am saying here should be clear to someone such as yourself, who is invoking our theory of mind in his claims: I am making a distinction between what you are now saying you meant all along, and what other people will recognize as having been your intent when you first wrote the passage in question. We cannot present proof, but nevertheless we know, beyond reasonable doubt. Your explanation does not pass the sniff test.
Wouldn't the most important safety question be "whether Waymo’s technology makes fatal crashes MORE likely?" Why assume Autonomous Vehicles have to perform better?
Waymo less fatalities == Win
Waymo same fatalities == Draw
Waymo more fatalities == Bad Waymo
Also, we've accepted and made legal frameworks around the concept that people sometimes kill each other with cars. Robots killing people with cars does not benefit from this carve out.
I am neither, hence I will not trust any system that does not perform at least as well as I do. The bar is much higher than the average.
...which is a pretty big fallacy and, if implemented as law (which I am working to make sure it is not), will get a lot of people killed.
Literal drug addicts, people driving under the influence of alcohol, and people driving distracted are driving right now. Even AV that performs at the level of the median driver will massively reduce the number of those people actually making driving decisions, and therefore directly reduce road deaths.
Why? Because most accidents and deaths are caused by drivers performing at the low end of the spectrum. In the US, a full third of accidents are caused by drunk driving - which would be virtually nonexistent with AVs ("virtually" because the AVs aren't going to be perfect and are going to need manual intervention).
The bar is literally the average, because if everyone drove average, deaths would plummet.
And, unlike the American public, it's massively easier to improve AV driving behavior than average human driver behavior. A system that starts out eliminating the lower end of driving performance while improving over time is a deal too good to turn down if you care about human life.
A significant proportion of the current fatal accident rate is based on incidents which involve driving so bad that humans implicated are severely punished for it, or at least banned from driving for a period.
Why would anyone set the benchmark for a commercial driving company lower?
It remains to be seen if autonomous driving systems are actually safe.
Alternative example:
It remains to be seen if microplastics will cause higher rates of cancer.
It remains to be seen if organic peanut butter will cause lower rates of cancer.
A good indicator of their next move is their previous move.
There's a good reason for this. It's because the human can be interrogated into what was going through their mind whereas many ML models cannot. That means we can't ascertain if the ML accident is part of a latent issue that may rear its ugly head again (or in a slightly different manner) or just a one-off. That is the original point: a theory-of-mind is important to risk management. That means we will struggle to mitigate the risk if we don't "hyperanalyze" it.
(Linking to the web.archive version because the graphics are better / more understandable when in the context of some of the text)
Chapter 6 is the most relevant here, but it's all a thought provoking story.
Cars can do unexpected things for good reasons, as the AI did in this case.
> "Acting unexpectedly" is one of the aspects that makes dealing with mentally ill people anxiety-producing.
Wat.
>"In August, a Waymo at an intersection “began to proceed forward” but then “slowed to a stop” and was hit from behind by an SUV."
>"Wat."
Are you saying you don't understand why unexpected behavior causes anxiety? It's a pretty well documented effect, from rats to humans.
> > "Acting unexpectedly" is one of the aspects that makes dealing with mentally ill people anxiety-producing.
> Wat.
The "Wat" probably refers to the fact that this seems unrelated. Dealing with mentally ill people is anxiety-inducing because they act unexpectedly... so what? Lots of things are unexpected. They shouldn't all be drawn into the analogy that says "well, that thing, plus a load of other things, can induce anxiety, therefore that thing should be tarnished with the same brush as all of those things."
People drive in unexpected ways all the time. Of all the criticisms to level at them, "the thing you're doing plus a load of other stuff can induce anxiety" probably isn't top of the list.
So to put a finer point on it, people need to acknowledge that public trust is necessary to wide-scale adoption of AV tech. Plenty of psychological research shows how we aren't intuitively wired to understand statistics. So all the bleating about statistics may be necessary, but not sufficient, to get wide-scale adoption of AVs on public roadways.
It wasn't just the word anxiety; that's a mischaracterisation, and another psychology-adjacent misstep with "hung up on". Your point wasn't great, or at least was very poorly articulated, and that's what caused me to "miss the real point". You seem to have clarified it with this comment, so my challenge to your original point seems to have helped.
Would also be nice to see all incidents, not just injury incidents to see what kind of unexpected "mistakes" these cars are making
I think you're understating it.
Mentally ill people act far off the cuff. If I'm walking outside, people can behave unexpectedly (but within the parameters of behavior that doesn't make me anxious).
Imagine: stopping instantly to bend over and tie one's shoes or to look at a storefront. Taking up multiple spaces on the sidewalk. Dropping an item that causes a loud noise. All of these are unexpected movements that require a reaction.
However, if somebody screams about CIA conspiracies or has very erratic mannerisms, that would create more anxiety.
So, apparently their point is that AI behavior on the roads might be more jarring than normal jackass human behavior on the roads.
I'm saying that your leap from "the car in front is behaving unexpectedly because you don't have all the information it has" to "a bunch of mentally ill drivers" is such a complete non-sequitur that it's deserving of a Wat [1].
Why do mentally ill people cause some people anxiety? Because they may behave erratically.* In other words, we have a lot more difficulty ascertaining what they are thinking, and by extension, predicting their behavior. The same can be said for the general public sentiment towards AVs.
* I also understand "mental illness" is a blanket category and I'm not using it as a stigmatizing term. It may be stretching the analogy too much, but it's simply a proxy for "I have a hard time predicting what's going through this persons mind"
I imagine the collision speed was quite low since they just left the junction. There's a reason things like following distance and looking in front of you while driving™ exist.
Depends on the size of the tree branch.
But also, being hit from behind at low speeds was a pretty common thing in early testing in Mountain View, when they were using the Google name on cars. That there are only a handful of incidents reported in this report means either the software has gotten better at communicating its intent to other drivers, or the driving public is aware that Waymo cars are way more cautious --- if they only make lane changes and unprotected turns by engraved invitation and everyone knows it, that's fine too.
In the early days, it seemed like it might be appropriate to install a 1979 regulation 5 mph rear bumper on these cars, because they'd likely get hit often enough.
There's at least one other alternative: selection bias. Since there is no standard industry definition, companies are allowed to not report many incidents.
>"Waymo, on the other hand, ran complex computer simulations after each disengagement, and only reported to the DMV those where it believed the driver was correct to take charge, rather than being overly cautious."[1]
So it may not be reported unless it meets Waymo's (non-independent) selection criteria. I think most people can at least recognize there is a potential conflict of interest when objective reporting isn't required.
[1] https://spectrum.ieee.org/have-selfdriving-cars-stopped-gett...
Our thinking fast reflexes are to pretty much avoid any obstacle whether living or dead. Hopefully our thinking slow brain has time to do a proper evaluation before doing anything drastic.
[1] https://www.kqed.org/news/11801138/apple-engineer-killed-in-...
[2] https://abc7news.com/tesla-autopilot-crash-driver-assist-cra...
Depends on the size of the tree branch.
As a thought experiment: Let's put you in a driving simulator, and ask you re-drive the same scenario 10,000 times. In each scenario, we randomly change the size of the branch -- both weight and volume. Repeat the same test with 100 other drivers. Repeat the same test with 100 different AV algorithms. I am sure each driver would have their own definition of "branch too large to safely drive over".
What concrete point are you trying to make?
being hit from behind at low speeds was a pretty common thing in early testing in Mountain View, when they were using the Google name on cars.
From driving near the things, they're very cautious and sometimes start to move and don't (I saw one try to make a lane change for about a mile on el camino before it was able to, turning the blinker on and off the whole time).
>> in early testing in Mountain View, when they were using the Google name on cars.
> Early testing? From what perspective.
You know, before they switched the name to Waymo.
On the road, sometimes I feel like I'm the only one following the rules. :/
ML having completely unpredictable failure modes is like the entire case against putting them anywhere. What would you call a vision system that mis-identifies a stop sigh because of a couple unrelated lines painted on it, other than "unpredictable"?
If humans can't do that, the solution is probably more automation.
The Uber fatality from years back showed that the software used "action suppression" to mitigate nuisance braking. The idea that that would be acceptable on a safety-critical software application should give us pause to consider that more automation is the knee-jerk solution.
Now, couple that with: 1) a general approach of "when in doubt, brake" and the hacky workaround to suppress braking if it's occurring too much, and you've got a bad nuisance braking problem that's primed for a safety incident.
But again, this is disjointed from reality. I'd agree that if we all left the recommended 4.5 seconds of stopping distance, we'd all be safer. But that's not how our roads were built and that's not how humans drive. You're tilting at windmills here to make a point that doesn't need to be made because it doesn't apply to reality.
Oh, it emphatically is how our roads were built. Most miles in the states were still laid down in an era where cars didn't do more than 45. As for how humans drive... I'll direct the audience's attention to the 3,700 per day road fatality rate in the United States. Even if they brake check more than human drivers, an automated vehicle being forced to follow its programming to maintain a healthy follow distance may very well save lives, especially in an ecosystem where they are the dominant vehicle on the road as opposed to human operated vehicles.
I'm not scared of mentally ill drivers. I'm scared of rich 16 year olds. I'm scared of drunk drivers. I'm scared of drivers sitting so low they can't see most of what is happening around them. I'm scared of drivers having seizures while driving (my mom was hit a while ago by a man who lost his license due to seizures, and still refused to stop driving). I'm scared of drivers who drive without a license because "fuck them, I drive when I want to". I'm scared of people mixing up gas and break pedals (just got hit by one), in cars which can go 0-60 in 2.6 seconds weighing 6000lbs.
This specific problem is just as much a design flaw as it is a PEBCAC issue.
You have two very similar pedals that perform polar opposite functions right next to each other, and they are both operated by the same foot.
I'm surprised this isn't a bigger problem.
man who lost his license due to seizures, and still refused to stop driving
Another one, suspended license, got drunk, and decided to play pinball with the car my fiance was in, causing lifelong injuries.
You can't get away from those. America is just a fun place.
You can't get away from those.
If these tragedies you describe keep happening (as described), and the fault you identified points directly to the problems of American drivers, I find it shocking you suddenly describe it as "this one issue". Either it's as problematic as you described or it's not.
And if one believes that, e.g. self driving cars will do a lot to address it in the next decade, they especially might not want to incur the costs of the alternatives.
I've seen people reverse up a highway onramp.
How will AI drivers navigate these 'cultural' differences? Will they insist on following the letter of the law (presumably, for liability) or will they adapt to local practices and potentially lower their overall collision rate with humans driving poorly. Interesting near term question.
The unexpected is, by definition, what was not predicted.
"all the time" means it is not unexpected. Humans do mistakes and omissions that you've subconsciously learned that humans do, so you can adjust to deal with that.
The difficulty with AI is that it behaves like a completely alien being, making errors that are alien and unpredictable to a human driver.
For example, just because someone has their turn signal on and is slowing down doesn't mean they're going to turn.
From my experience with LLMs, AIs even when called to be creative are fairly consistent, even if at first they seem to be producing output that’s more creative than human norms.
Perhaps a self driving car's unexpected behavior is on the "too defensive" side while human drivers are on the "suddenly do something wildly dangerous" but this is easier to account for as a driver rather than more difficult to account for.
You wouldn’t expect someone to suddenly accelerate from a stop while they have a red light, so you don’t typically look both ways when driving through a light that has already been green for a while when there are cars on either side waiting.
I’m not saying Waymo cars do that; just an example of expected unexpected vs unexpected unexpected behavior.
Also sometimes I can intuit that the driver in front of me is probably going to do something dumb because of other minor things I've observed about their driving style.
"Acting Unexpectedly" can often mean following the actual laws and general guidelines for safe and/or defensive driving. I would hazard a guess that sometimes doing the intuitive thing is, in reality, unsafe and/or against the law. If the car does this in 99% of circumstances, and still gets rear-ended, who is really the problem here?
I go by feel. If it feels much safer to speed, I do. Otherwise I'm in the right hand lane, usually a safe stopping distance behind a big rig for aero gains.
The unfortunate thing is people take your propensity to maintain a safe follow distance as an invitation to cut aggressively in front of you, potentially across multiple lanes.
It sucks all around, but I don’t think AVs are the solution and I definitely don’t trust any company producing AVs is delivering a product that does what it claims. Waymo doesn’t acknowledge in this about the part where their vehicles have human operators take control when the car doesn’t know what to do. I assume they’re not including that data. That’s going to skew the result. Even if the car is super safe in most driving conditions, ignoring what is arguably the least safe conditions the car can be in in your data analysis is fucked and intentionally dishonest.
2. Most of the time it is a false perception. You only notice the people whizzing past you, not the people driving the speed limit along with you because they never pass you (a variant of survivorship bias). This of course depends on where you are, but most people in fact do drive around the speed limit.
Counterpoint: Google Maps thinks you can drive an 86-mile trip from from Springfield, MA to Albany, NY in 80 minutes, a route which is patently impossible if you're driving the speed limit; you cannot drive 84 miles on I-91 in 76 minutes, an average just over 66 mph, on a road with no segments 65 mph segments, without exceeding the speed limit.
https://www.google.com/maps/dir/Springfield,+Massachusetts/A...
But also, this claim just doesn't stand up to scrutiny. Cars take up a lot of space. You can't just drive through them, you have to take affirmative action to change lanes and pass them. You'd find yourself behind people driving the speed limit pretty often, which you will notice.
That is no counterpoint. I am talking about people, not Google Maps. Also, the standard freeway speed limit is 65 MPH AFAIK, so I-91 seems like an exception that Google Maps is not accounting for (and perhaps human drivers as well).
> You'd find yourself behind people driving the speed limit pretty often, which you will notice.
I assume you're implying that you're always driving above the speed limit and you and you're saying you don't find yourself passing people. There could be many reasons for this. You could be in one of the exceptions that I mention, or you could be driving at times and/or along roads without many people. You could also be selectively remembering things; you are much more likely to remember passing people if the incident is frustrating and less likely if it is not frustrating.
You cannot average 66 mph without going over 65 mph, and the states involved don't post speed limits over 65 mph. You cannot get this result with by a simple oversight under the assumption that speed limits represent the speed people actually drive.
Consider what inputs might lead Google Maps to such a conclusion.
Do you care about justification, or do you care about being safe?
intuit the behavior of other humans through the theory-of-mind
If AVs consistently act "unexpectedly"
And I've been to MANY places in the US where people drive terribly and completely unexpectedly. I've seen people put on a left turn signal then go into the RIGHT lane. I've seen someone pass on a one lane overpass, where the car in front was towing a heavy load
What's good about these cars is they will ALL drive consistently, and the more people get used to them the easier it well be to drive with them because you will know how waymo cars react.
One of the first thing my driving instructor told me was: never trust anyone on the road except yourself. That’s why you see people waiting for a car with the indicator on to visually hint they really intended to turn.
Also, only the 2nd item is even consistent with Waymo behaving unexpectedly (we're not given enough info to know why it stopped). In the first item, the "unexpected" thing is the branch, not the behavior (stopping), and in the third Waymo's behavior didn't contribute to the accident at all -- instead it nearly avoided it despite the other car's bad driving.
The responsibility was 100% his because of "an unsafe lane change".
Since all of these accidents happened in the US, is the driver that hits from behind normally responsible for the accident? (For a moment, let's exclude predatory behavior where the front driver is doing something toxic, like intentionally pump-breaking on a high speed road to induce a hit-from-behind accident.)
How so? If human drivers did unexpected stuff like brake for no reason, we'd have a lot more accidents.
I think this just highlights how much better humans are at cooperating on the road compared to automated systems.
Traffic disruption with Waymos is UNDOCUMENTED and a real thing.
The Ars article linked to the Waymo blog post [1], but the underlying paper is at [2] via waymo.com/safety . A lot of folks are assuming this wasn't corrected for location or surface streets, but all of the articles do attempt to mention that. (it's easier to miss in the Ars coverage, but it's there). The paper is naturally more thorough on this, but there's a simple diagram in the blog post, too.
[1] https://waymo.com/blog/2023/12/waymo-significantly-outperfor...
[2] https://assets.ctfassets.net/e6t5diu0txbw/54ngcIlGK4EZnUapYv...
- Safety drivers prevent incidents
- Safety drivers cause incidents
Since the human+robo isn't the intended deployment mode, the only reason to count its numbers would be if there wasn't statistical power any other way. And that's no longer true.
In my perhaps naive view a safety driver failing to act and a robo driver should be identical.
Comparing robo-only miles has the advantage of not considering incidents like this, and also not considering incidents where the safety driver prevented harm. It also prevents accusations of "you're claiming human Waymo drivers caused these accidents that you've omitted and are lying" because then they're counting 100% of miles driven without humans in the car.
From looking at past Waymo data, you get similar numbers (a few more incidents, but a lot more miles driven) with safety drivers included. They're just noisier and more subject to dispute.
I'm sorry, but a dog running out in front of a car is not a "traffic fatality" by any conventional sense of the world, and it's certainly not an at-fault accident for the driver.
At this point, I believe you are trolling.
At this point I believe you are trolling and disinterested in discussion that might be critical of Waymo’s brand and safety.
We've previously had accident rates vs. total miles disclosed. Now, we have a new, better number (for reasons stated above). Of course, it's always good to look at a variety of measures.
> discussion that might be critical of Waymo’s brand and safety.
When someone uses the word "fatality" to obscure that they're talking about a not-at-fault accident where a dog runs in front of a vehicle and the dog dies, it convinces me that the person I'm talking to is not an honest counterparty in discussion.
- Most people reading "fatality" assume that it means a person died. This was not true of the incident. Repeatedly using the word "fatality" here is intentionally misleading. I have not heard the word "fatality" used in accident reporting to describe the death of an animal before.
- This was not an at-fault accident. It was not possible to avoid hitting the dog. Talking a lot about this incident doesn't provide very much information about Waymo's safety. (Of course, humans get in accidents that would be impossible for them to avoid, too-- all else being equal such accidents should be included on both sides of the analysis).
- The analysis didn't include this incident, because Waymo made a very rational choice to calculate using no-safety-driver miles and incidents. This isn't malice; it doesn't change the numbers much (because incidents and miles both go up when you include safety driver miles) and the numbers are higher quality and less confounded without including safety driver miles.
Not sure why you're focusing on this event. They're talking about miles with no safety driver.
Self-driving technology will overtake average human ability with regard to safety within a decade, but the biggest hurdle will be public acceptance. The AI will not make the same kind of mistakes humans make. So while the aggregate number of accidents will be (likely much) lower without a human at the wheel, the AI will make deadly mistakes that no human would make, and this will terrify the public. A intuitively predictable crash will always be scarier than one that makes no sense to our minds. The only way self-driving tech will ever succeed is if the AI can be limited to the same kinds of mistakes humans make, just fewer, and that's a VERY hard technical nut to crack that I do not believe will be solved anytime soon.
That said, I still believe that the ubiquity of cars is inherently a problem, human operator or no. If we put more effort into self-driving busses and autonomous trains—which have regular schedules, routes, and predictable speeds—I think we would see much greater dividends on our investment and far fewer "unintuitive" errors. Our collective fixation on cars blinds us as a society to this option unfortunately. More cars just clog up the road even more, demand more parking, and otherwise monopolize land use that could be more productive otherwise. More idling/circling driverless cars adds to the blight rather than relieving it. We need to transport more people between points in higher density, not lower, and cars are the lowest density transportation options available.
Of course, when a collision does happen and damages have to be paid, the injured party will of course start advocating for full liability even if they previously had no issues engaging in the risky driving themselves. Which is why this is not reflected at the polls when voting for a politician who would promise cracking down hard on moving violations, with things like cameras and increased police stops.
It's the cars. Autonomous or not, cars are the problem.
But just like health care, we keep choosing and voting for the absolute worst option because it's what we're used to and fear-mongers selling us a story about how any change will inevitably lead to a socialist hellscape. Fear: it's a hell of an addictive drug even as we witness our own obvious decline.
For what it's worth, I don't actually shop at the supermarket closest to my house that is within walking distance because they don't offer good prices.
There are some software decisions that worked well up to a certain point, but simply won't scale reliably to the current level let alone what you need for the foreseeable future. At some point you have to bite the bullet and refactor. Refactoring takes skill, and you need the right folks for the job, but it will get harder and more expensive the longer you wait. It typically doesn't have to be done all at once. Just fix as you go and stop following the older patterns. Have a plan and work toward the goal one step and a time. One commit at a time. One stroad at a time.
Folks may believe in the whole "personal freedom" with a car, but how personal or free are you sitting in bumper to bumper traffic? Cars = Liberty is a pernicious lie.
Even for cities of normal size, the total distance across means you would not want to make it a regular event to shop on the other side of the city. So if we add transit, and increase density you can find there are more places you within a reasonable range for whatever activity even though you lost the freedom of the car.
Of course if your activity isn't shopping - something a city excels at - but instead camping far away from other humans: then a car means freedom. When the activity you want to do is something a city excels then doing it via transit should mean even more ability to do it and thus more freedom. Of course this freedom via transit only works out when there is great transit and high density. Getting there is often difficult.
When in Manhattan, the subway can 100% beat a car over the length of the island. (Car can beat the busses going cross town though unless of course you add parking back to the equation.) Getting to Queens seems to be faster on the subway than taking a car across the bridge too.
D.C. and San Francisco are both towns I'd also rather take mass transit than drive.
We have excellent public transport, but it is slowly becoming too expensive for the municipal budget. Given that the city is historically not compact (there are either old industrial brownfields or rivers with adjacent floodplains that are unsafe for residential buildings), trams and buses need to cross kilometers of mostly uninhabited territory before reaching dense parts of the city again. Of course, that costs money in fuel or electricity, extra wear and tear on the vehicles, plus the polycentric character of the city does not allow for a simple network of lines meeting downtown. You need more of a triangle.
And there is approximately nothing that can be done about it. The floodplains are dangerous to build in, the rust belt of brownfields would be too expensive to redevelop, the economy of the city is far from stellar and won't support any extensive redevelopment anywhere; we are already losing population, though not dramatically so.
Even the sprawling suburbs are dense enough to support greats transit - but since they don't have great transit everyone drives creating a death spiral that is hard to break out of.
> and you have to use an autonomous car
The idea being you use this hypothetical level 5 autonomous vehicle instead of the ones that let you be a dipshit.
Anyway, we already revoke peoples’ licenses for drunk driving and we do it in the areas you’re referring to.
It will feel more dangerous and untrustworthy, and that has always been more than enough to kill a promising solution in this country.
If we can overcome that and actually follow the numbers instead of our guts, we won't need most of the cars in the first place—autonomous or otherwise—because the numbers say mass transit is the better option on all metrics: economic, environmental, safety, land use, etc. Cars are at best a backfill option.
Why not just make tons of public transit at that point using the level 5 tech instead and cut traffic?
Because there are places that public transit will never cover? I've been places in the American West that are 50+ miles from the nearest paved road, stop sign, or restroom. I don't think they're getting bus or train service any time soon.
Level 4+ cars may be quite expensive for some time. Operating costs of L4+ vehicles will likely be low compared to human-driven vehicles (counting the cost of the driver). So, it could lead to car sharing in situations where we don't accept it currently.
> that are 50+ miles from the nearest paved road, stop sign, or restroom.
US population density is lower than most of the developed world, but what you describe is a tiny share of miles driven and can be effectively ignored for now.
A lot of fatal accidents can be attributed to inexperience, distracted driving, running lights, drugs, alcohol, asleep at the wheel, medical events or extremely aggressive driving.
The distribution of unsafe driving is not even. Personally I had 2 accidents in my teens and 0 for the following 20 years.
Your typical taxi driver will be fired/fined/reported over time if they drive this way. Further, if you are in a taxi where you notice a driver making you uncomfortable, you can end the ride early (I have).
However, imagine a robodriver that is 4x safer, however every single vehicle on the road has the same probability in any instant of invoking the same "no human would make this mistake" driving error fatally.
An analogy would be that at any moment, your calm, courteous, focussed spouse behind the wheel suddenly transforms into a 17 year old teen in a Mustang.
The GP is arguing that there is a way to actively become safer than the average person by avoid getting into "dangerous" situations. This may or may not be 4x safer than average though. There's this element of control that is lost when it comes to AI driving.
The relevant point to you is, if you already have good driving skills, always drive responsibly, and generally avoid driving in areas/times that have more drunken drivers around, then you might not actually benefit 4x.
Next problem is everyone else's driving! It's like an arms race / tragedy of the commons issue as well. Your safe driving doesn't matter if other people are still running lights / driving drunk / etc. In absence of near-perfect AI driving, followed by government mandates to take everyone else off the street.. it doesn't matter. And this is never going to happen.
Unless the government is taking everyone else except the imaginary future perfect robots off the road, its all a moot point.
Anyway, the mechanisms I have listed can still drive adoption of self-driving cars and reduce the number of human-driven cars on the road which also reduces the number of fatalities if the self-driving ones outperform humans. And if we look at QALYs instead of fatalities there may be additional benefits.
almost all our tech wants to skirt to legal system.
I feel like the one in five Americans that live in rural areas is left out of the conversation though. You can't eliminate cars for those 60 million or so people.
Folks need to be transported at higher and higher densities as a city's population grows. Cars are the lowest density carrier available. Think of how many cars fit on a four-lane road on a mile stretch. How many people are in those cars? How many trains or busses would be needed to move that many people? Now visualize the space taken up by those cars versus the space taken up by busses.
That's how you solve traffic problems with a growing population, even for the folks who still need their cars because their destinations are sufficiently irregular. Mass transit helps those who need their cars too!
As a byproduct, you don't need so many and so large parking lots. Think of all the parking lots around you, which I'm sure there are many. Imagine 80% of them were replaced with housing, retail, office space, parks, meeting places, etc. Then convert the remaining 20% to multi-story parking.
Urban sprawl is a choice. Choose different.
You may not approve that such places exist but they do. And folks like to live in them.
But it's a choice. There are a lot of folks out there in the "good money after bad" camp who focus too much on what is and what was rather than what could be.
[rains] Radio traffic announcer: "It's slow out there because it's raining."
[cloudy] Radio traffic announcer: "It's slow out there because of low visibility."
[sunny] Radio traffic announcer: "It's slow out there because it's sunny."
I'm talking about segments of the other 80% that wants a dually truck because it makes them look "alpha". Folks buying huge SUVs to feel "safe" while being more prone to rollovers, less able to avoid collisions, and far more likely to kill others—especially pedestrians—in a crash in addition to monopolizing greater and greater proportions of limited land resources.
You live three miles from your nearest neighbor? Feel free to indulge in a raised pickup with 3 tons of bed capacity and 5 tons of towing with my blessing.
You live in one of the major metropolitan areas in the US? Don't buy a Hummer, Lexus SUV, or F-150, especially if safety is your goal. In fact, those large vehicles should require a new class of drivers license due to their size and performance characteristics just like school busses require a class B and motorcycles a class M due to their different structure and place within our highways. Buy a transit pass. Per capita, folks simply don't die in car accidents when they ride the bus or take a light rail. Don't have good/fast public transit infrastructure where you live? Time to vote for folks who will make it a priority.
Because widening stroads has been tried. It doesn't work. They have never worked. They don't make traffic better, they don't make us safer on the road, they don't get us to our destinations safer, and they certainly don't make the most efficient use of land. It's time to move on. Dump all the stupid, oversized, single-level, paved parking lots and replace them with mixed-use housing, retail, and office space with a public transit hub.
Make just enough parking so that the 20% folks who actually need their daily-use vehicles can visit easily. Preferably they can park in the park-n-rides at the outskirts and hop on a train to the city center so the parking fees are as cheap as possible. Let the 20% decide whether they want self-driving vehicles or not. The 80% should leave them alone and embrace the self-driving busses and trains for themselves.
It's a question when not if as it's clear that we haven't found fundamental issues.
The opposite happens: more people use it already than I thought and companies are stepping up with insurance to cover you.
Yes the Tesla car driving into a truck, horrible but there was no genuine global outcry.
I'd argue safety is not a concern. There are a lot of "safe" things we could do, but don't. A significant percentage of the country doesn't even vaccinate its children. Self-driving cars aren't suddenly going to make us aware of our own mortality in ways that life itself hasn't already.
The real fear is lack of accountability. If a drunk plows into a crowd of pedestrians, he will be dragged out of it and (metaphorically) lynched. Justice makes us feel better about circumstances beyond our control. If North Korea test-launched an ICBM that erroneously hit Japan, we'd declare war over their typo. But when a self-driving car erroneously mows down pedestrians, we're told to just accept it, nothing we could do, mumble-mumble-training-data, and tragedies like this happen so we can be safer.
Nothing is going to condition us to resent the idea of Safety more than having our personal agency and sense of justice taken away from us in its name.
Cars != Liberty
They never were. They have always been rapidly depreciating assets that are useful for one-off destinations and horrible externalities with regard to city planning.
To be clear: I do think, in my lifetime, AV will become normal in many places. Some horrific accidents will happen that result in massive fines to the companies. As a result, some of them will go bankrupt.
We're hosed until we can instruct our vehicles to act drunk.
Not having a track to follow on/in the road (magnets, sensors etc.) Not mandating all cars talk to each other, working together like a mesh/hive/colony.
I understand that has its own set of self starter issues, but it can be built in WHILE also doing what is currently happening. The fact that roads are being replaced TODAY and still nothing is going in them to help cars drive themselves, baffles me.
Everything you said made sense, except focusing on mass transit for FSD, for two reasons:
The intent of focusing on bus/train automation comes from an illusion of control (we can control the lane/track, thereby) – hence we tend to rudimentarily attribute easy outcomes to it (low risk, high value).
If we ignore the control part and properly think about it – mass transit actually higher risk for lower value.
1. Lower value: For something that involves 100+ people on dense economic centers, it's already running at an economy of scale where a human driver just makes sense. I live in Germany where the metro trains & trams are already crazy automated. There is a human driver there just in case, more as a supervisor for the people riding (controlling hooliganism, jammed doors, helping challenged people, dealing with emergencies, etc). I see German trains as already running on FSD4. FSD5 full automation is a waste of time here. Using buses for last mile coverage for few passengers, aka treating buses as "big taxis" is probably worse environmentally than actual taxis.
2. Higher risk: By the same logic you said for cars – "far fewer unintuitive errors" – at a much higher capacity of mass transit – is far more catastrophic. Imagine a self-driving train had just 1 accident in 10 years, but it affected 1000 people. It's sheer terror. Who is liable for it? Government. The problem with going down this mass-transit-first route is, one error means legislating away the entire sector.
Cars are actually lower risk (individual choice, individual liability, accidents don't deter others from adopting) and higher value (last mile, moving away from the dense urban city plans that come with high rents and chokepoints which are crippling even to my beloved, beautiful German cities where even with all the urban sprawl, last mile is still a problem outside A zones).
Your second point is completely wrong: we have trains, and have been running them for more than 100 years. We have real statistics to show in the real world they are much safer than cars. Sure you can imagine anything you want, but when real science has real data why would anyone look at your imagined data.
As for the first, it would be worthwhile taking the AI-tinted driverless-only FSD glasses off, and re-read the original comment and response.
I live in Germany where the metro trains & trams are already crazy automated.
- What happens when all of the infrastructure gets built up around self driving cars and peoples first hand knowledge of how to drive diminishes. Once a near monopoly/duopoly is attained by a select few SDC vendors, it will become a utility. Then what fallback does society have if the likely enshittification happens. Do we just have to live with it?
- Its all fine when the companies able to do this are part of the most elite, technology first companies - but what happens when companies known to take short cuts (like the ones who can barely get bluetooth working for their audio infotainment system) start try to enter the market by focusing on lobbying the SDC oversight board.
you'd be surprised what kind of mistakes humans make.
Anyway, snarky comment aside. The biggest reason for optimism is that a world full of AI cars will remove the reptile-brained jostling for position that's 90% the cause of all crashes today, and that it will overall _slow down_ traffic. Slower, calm, tepid moving traffic, a bunch of electric golfcarts puttering around the city. That's a future of AI-only traffic worth signing up for.
They will drive even faster thinking their AI will protect them.
I know this because Tesla has “ludicrous mode” and all other EV manufacturers are bragging about their insane 0-60 and 0-100 mph times, and owners love to show off.
Because the autonomous vehicle will be autonomous.
That's awful advice. It's something that feels right, but in reality only exaggerates the push-pull accordion effect of too fast, too heavy traffic.
It's not the speed. It's the speed difference as cars end up weaving between lanes in traffic because slower vehicles sit in the passing lane(s).
My favorite scenario is when someone super impatient pulls around (often suddenly without signaling) a car not noticing:
* The car in front of them is actually going the same speed as the car in front of them
* The lane they were all in is actually going faster than the lane they just pulled into
* Everyone is about to pass a slow person up ahead in the newly selected lane
Person predictably hits the gas to race ahead only to get stuck behind the slow car while the cars they thought they were passing proceed ahead in the lane they just left.
Sometimes frustration and increasingly eratic behavior ensues.
Never gets old :D
autonomous trains
This not only makes it more convenient to get around without worrying about buying transit tickets ahead of time, it means the busses can load from the front AND the back, making the time spent at each stop shorter and schedules easier to keep.
Much easier problem for autonomous setups when you remove fares and just have to monitor if anyone's in danger.
Public mass transit never breaks even from fares. What it does very well however is free up the money folks would have spent on gas, parking, car maintenance, etc., so they can move around and spend it more freely in the community.
You see this spectrum in various cities. Some cities make you pay depending on how far the light rail travels, so you naturally condition riders not to ride any more or further than they absolutely need to go. Places like New York City on the other hand just sell you a pass; taking an extra trip to the Met or to Central Park doesn't cost extra, so folks do it, socializing and stimulating the economy along the way.
New York's biggest transit mistake in my opinion is keeping the fares. They'll spend HUNDREDS OF MILLIONS of dollars on NYPD overtime to catch turnstile jumpers when they could spend less just letting people ride for free. https://gothamist.com/news/nypd-overtime-pay-in-the-subway-w...
It's like when Florida drug tested everyone trying to get public assistance and ended up spending far more for administering and processing the test to catch a relative small handful. https://www.aclu.org/news/smart-justice/just-we-suspected-fl...
Possible? Yes, but not for decades to come.
I'm still a bit torn on whether autonomous cars are a good thing once you consider all the second and third order effects (even more cars on the streets, less investment into better modes of transport, and traffic will get a lot worse once people are ok with sitting in bad traffic and watching Netflix). But I have to applaud Waymo for their great execution on a very difficult problem.
This one could go either way I think, traffic might actually improve once autonomous driving is the standard.
I also kinda-sorta hope that if autonomous driving takes over, that cars end up gaining the ability to switch onto and off of rails, I think this would be the ideal end-state... people still maintain the ability to move independently of each other but we have the improved safety of transport on rails.
In reality, self-driving cars would help by increasing following distance and leaving a genuine gap so people don't crash every goddamn day on the same arterial roads.
1. Will it mean fewer total cars? Probably. If I have to drive to work and then back home, and you do to, and we each own cars to do it, that's at least two cars. If I can take a Waymo to work, and then it can take you to work, that's only one car.
2. Will it mean fewer cars on the road? (Or, perhaps, let's say fewer car-miles driven.) Plausibly not. If I drive from home (A) to work (B), and you drive from home (C) to work (D), then if we own cars, we drive A-B and C-D. If we use Waymo, it may drive A-B-C-D, which is longer by the B-C leg. That takes up space on the road.
So we may have fewer total cars, but more car-miles driven, and therefore more traffic and congestion.
Owning a car is simply more economical. Now if I could buy into fractional ownership of a fleet of vehicles, that may make financial sense for me.
But if you own the car, it's just waiting wherever you left it. If you have an autonomous cab, it has to make an extra trip from wherever it dropped off the last driver to wherever it's picking you up. That alone increases the number of cars on the road. And that's before you consider the cab potentially driving a holding pattern when nobody is actively using or calling it.
But most of all roads are governed by induced demand. People would take a lot more and longer trips if there was the option to just teleport to the destination. The main downwards pressure on the number of trips is the time investment. That's why adding more lanes to roads often doesn't reduce traffic (outside of a short adjustment period): faster trips means more people willing to take it, which fills up that lane. But a trip people weren't willing to do for 40 minutes behind the wheel they might take if it's instead 60 minutes watching Netflix in a driverless car. Which makes the roads fuller and thus slower for everyone.
If autonomous cars drive down the costs of taking a taxi it’ll mean more people will do that versus public transit.
Anything that reduces public transit use or increases individual car use will be disasterous for traffic and transportation in our cities.
Imagine those unmutable video ads that are increasingly common at gas stations, but running constantly inside the car.
I guess it's always possible that they're lying about it, but given the depth of the regulatory and public relations fiasco that would cause I'd be very surprised.
That ship sailed a while ago and didn't need Google to push it.
I strongly believe it will go the other way, i.e. the 'robo-taxi' vision. Once cars can pick us up, take us where we want and then disappear, very few people will want to own their own car. I honestly think the vast majority of people already don't want to own one, but we don't have a better option. Why would a sane person want to deal with the maintenance, insurance, repairs, depreciation, etc.
Cars will just show up, take us places then go away to get someone else. We won't need nearly as many of them, and we won't need to dedicate so much of our cities to them, and especially not to parking them. We will be able to reclaim our cities.
NOTE: Old School automakers who can't/won't/don't adapt are going to push back on this HARD. But I still think it will happen.
For the record, I'm a car guy. I love cars. I will likely always have one for the weekends. If I was going into a city or commuting, I would take the robi-taxi every time.
Because it will probably still be cheaper if they use it regularly (as is owning in most cases). Because they want a specifically equipped vehicle for young kids/outdoor activities/etc. Because when they want a car, they want one right now.
I'm also skeptical that, if you own a vehicle, it would make any sense to then also rent robo-taxis locally. Certainly I can reserve a private car for an evening event today but it would be 10x or more the cost of parking/gas.
The biggest change would be the lack of a need for parking. This will allow us to build more densely.
Induced demand economics. Make something cheaper and people use more of it.
If Autonomous cars make ride hailing cheaper we’ll see an explosion in its use, and traffic will increase as cars are space inefficient and road size fixed.
If you are a single person working a flexible schedule (no mandatory meetings), with no other activities planned traffic and travel time are not a big deal. However if you have any life at all you will care about traffic and travel time because you have places to be. Watching netflix is not your goal it is how you kill time that you would prefer to do something else.
I'm as much of a transit advocate as the next guy, but I think a lot of people blind themselves to how annoying the last mile problem is for a lot of destinations anywhere outside of urban cores. There aren't going to be train stations built at every possible origin point and destination point, and even if there's a robust bus network, transfers, slow speed/frequent stopping, and the walk to/from your destination/origin are pretty damn annoying. They're not the end of the world by any means, to be clear, (I use buses too!) but it's just, if I have a car, why wouldn't I just drive?
Taxis have the potential to solve that in a great way. But I (...and probably most people?) don't currently use them for that purpose since they're way too expensive. As they should be, it's a whole human being tending to your transport personally for twenty minutes or more. If robotaxis can lower the price, it'd be great, but I don't know how confident I am on that happening. The equipment is presumably expensive, the car itself is expensive (though EVs do have much lower maintenance costs), the R&D is expensive. We'll see. Exciting times!
People will continue to all commute at similar times, resulting in the same traffic as before. That it is automated makes no difference.
And supposing a car can go off after and do some taxi work instead of being parked, well it still has to commute from here to there which is adding even more cars constantly onto the roads driving to where they’re needed whereas before they’d be parked.
The likely outcome is permanent rush hour as cars are constantly going back and forth on the highway.
EDIT: If you disagree, please link to the quantitative data that suggests Cruise isn't on track.
IMO stuff like this is going to lead the public to trust it less, since they're gaming numbers as hard as possible.
[0]: https://www.cnbc.com/2023/11/06/cruise-confirms-robotaxis-re...
The whole game plan is have a bank of human operators who prove remote assistance at initially high rates which is then driven lower over time as the edge cases are ironed out iteratively. The fact that Cruise is only using one human remote assistant to manage ~15 rides, as mentioned in the article you link, tells us that the rate of remote assistance is already so low that it will be a very modest expense. For more, see the comment from Cruise's CEO: https://news.ycombinator.com/item?id=38145997
California, at least, cites a disengagement as "whether because of technology failure or situations requiring the test driver/operator to take manual control of the vehicle to operate safely."[0]
Would a car being confused and not being able to proceed without input be a disengagement by that definition? I think so, based off of "technology failure", but it's not reported as that.
> It just means the car is confused. Conflating these things and accusing Cruise of deception is itself being dishonest.
When a car is confused what happens? It stops. That is a safety issue by itself, as it can lead to emergency services not being able to properly respond and killing someone[1].
The fact people are trying to downplay this as "nothing" is shocking imo. What happens when a fleet of vehicles get confused, they all stall and it results in gridlock and frustration.[2]
[0]: https://www.dmv.ca.gov/portal/vehicle-industry-services/auto...
[1]: https://sfstandard.com/2023/09/01/person-dies-cruise-robotax...
[2]: https://www.forbes.com/sites/bradtempleton/2022/07/08/cruise...
At your Ref. [0], I just opened up the CSV titled "2022 Autonomous Vehicle Disengagement Reports (CSV)" under the header "2022 Disengagement reports". Under the column "Driver present (yes or no)", every single entry said "yes".
> When a car is confused what happens? It stops. That is a safety issue by itself
No, the car pulls over, just as it and every other taxi does when picking people up or dropping them off. It does not just stop in the middle of an intersection. I had 3 of these events in 22 trips, which means the number of times the car pulled over was overwhelmingly dominated by normal pick-up and drop-off, not confusion.
> as it can lead to emergency services not being able to properly respond and killing someone[1]
This article is deceptive, and you're either being deceived or are furthering it. An ambulance being delayed for seconds or minutes by human-driven cars in the road happens all the time. It is a constant occurrence. "90 seconds elapsed between the patient being put on the stretcher and the ambulance leaving the scene" means that at worst the ambulance was delayed by 60 seconds because stretchers don't teleport instantly into ambulances. The article does not causally attribute the death to the delay because that is extremely unlikely. It's not how emergency medicine works. This is just a classic case of fear mongering. ("Ambulance has to take detour around construction. Patient died. Therefore, construction caused death." No.)
This of course doesn't mean that delaying ambulances unnecessarily by even a second should go without punishment/fine. It's avoidable and should be fixed. But it's wrong to think this doesn't happen with humans, and its slander to suggest the delay probably caused a death in this instance.
> The fact people are trying to downplay this as "nothing" is shocking imo. What happens when a fleet of vehicles get confused, they all stall and it results in gridlock and frustration.[2]
Be more quantitative.
Uh, that's kinda exactly my point? Any time a vehicle stalls, even when it has to be recovered by a person physically, somehow isn't a "disengagement."
> No, the car pulls over, just as it and every other taxi does when picking people up or dropping them off. It does not just stop in the middle of an intersection. I had 3 of these events in 22 trips, which means the number of times the car pulled over was overwhelmingly dominated by normal pick-up and drop-off, not confusion.
Easily refuted. [0], [1], [2], [3] should I continue?
> This article is deceptive, and you're either being deceived or are furthering it. An ambulance being delayed for seconds or minutes by human-driven cars in the road happens all the time. It is a constant occurrence.
"It already happens, so who cares if people die." Not even going to bother with the rest since it's clear you're pushing an angle from these two points alone.
[0]: https://www.youtube.com/watch?v=wWZGZWuUx-Y&t=59s
[1]: https://www.youtube.com/watch?v=h1k8raq83T4
[2]: https://www.cnn.com/2023/08/14/business/driverless-cars-san-...
[3]: https://www.sfchronicle.com/opinion/article/san-francisco-po...
Your second point is replying to a general quantitative statement with anecdotes. Of course there will be unusual situations. This does not support your original wrong claim that Cruise was being disingenuous.
TBH, I didn't even fully read your third point because it's clear you're pushing a certain perspective. I've provided evidence for all my claims, yet a single link hasn't shown up in yours.
10+ vehicles stall and require people to go out and retrieve the vehicles[0]...
Nope, not a disengagement. Clearly the cars didn't have "a technology failure" they just needed to be towed back...for reasons. I guess they all ran out of gas, right?
> Your second point is replying to a general quantitative statement with anecdotes. Of course there will be unusual situations. This does not support your original wrong claim that Cruise was being disingenuous.
He says, while doing the exact same thing.
The vast majority of recorded cases involve the cars "stalling out" in the middle of intersections, roads, or driveways. I have literally never seen evidence of a vehicle "pulling over" when confused.
Please present evidence for your anecdotes.
[0]: https://www.ktvu.com/news/driverless-cruise-cars-cause-traff...
Cruise is done.
Again, if you have data that shows Cruise is behind Waymo by a lot, or is behind any other company, please link it.
That's why it could be a good bet. Or not.
The existence of taxis is (obviously) not enough to curb car usage growth.
EDIT: Some specificity: How would robotaxis replace commuting for millions of people in a way that reduces car rides? The taxi has to move at least from the storage to the rider pickup to the rider dropoff. Without sharing, that's actually more miles and the same number of cars.
Instead, if it picks up two people per day, that's more miles, fewer cars in existence (since both riders dont need a car), but the same number of car trips (plus the to/from storage).
With taxis (robotic or otherwise) the number of miles driven is just going up unless people change their lifestyle. That doesn't do anything to curb care useage.
If it ended up being in the low hundreds, well, that's lower than a lot of people's car payments. Couples or roommates could share a car for non commuting purposes or trips.
You factor in intelligent ride sharing and you could halve the number of cars on the road most days.
So the leap here is based on "Autonomous taxi companies will charge less per ride than rideshare"?
perhaps.
And the cars and autonomous driving software itself is becoming more expensive and more subscription-based over time so those rents are going to have to be passed on to the consumer. Large autonomous taxi services may be able to strike better deals or even build their own software/vehicles if they're big enough, but you're not going to be able to compete with them effectively by purchasing a Tesla (and presumably consumer prices will rise as there's less individual-owned vehicles and companies go seeking after only the highest margins and abandon the toyota-corolla market to the robotaxi corporations).
In reality, I don't think it is useful to try to enumerate these small immediate changes that are distinct from the availability of taxis. The long term cultural shift of having autonomous vehicles may lead people to fundamentally share vehicles in a different way. This may lead to a situation where fewer vehicles are driving more miles.
Only if it is always picking the same people up. Otherwise this is a big negative. People often need to arrive someplace on time. If my car had decided to take a detour to pick someone else up and made me late for my early meeting I'd be mad. Car pools work - to the extent they do - because it is always the same people who need to arrive at the same time.
Cost. The cost of an Uber is way too much for daily travel (vs owning your own car or public transport).
A human-driven taxi needs to pay the driver's salary within an 8 hour shift. An autonomous taxi can run (almost) 24/7, 365 days a year. Which do you think will be the cheaper fare?
Another scenario is someone simply renting out their own car as an autonomous taxi whenever they aren't using it themselves (which is most of the time). Then it'll always be cheaper than current-day taxis because it's just a low-effort bonus source of income to the car owner.
And, for a car driven any reasonable amount, most of the cost is in the mileage.
Neither, any savings will trickle up to the investors. The price of robotaxis is going to be just below the limit where it would make sense to own a car.
I think most people will try the taxi, but if you already own a car (that is transit doesn't make sense for most trips) you will discover it isn't much cheaper than owning your own car, and your own car is waiting outside when you want to go (one big advantage of owning a car over transit is the car is ready when you want to go instead of having to call or hope one is waiting - if cars need to wait outside your office all day in case your kid gets sick that increases costs). Instead you can just buy a self driving car and then leave your things in the car if you go shopping over lunch - something you cannot do with a shared car.
That doesn't mean I agree with the GP's point about it lowering car usage overall. The reduced cost of auto taxies also pushes against your reluctance to take one, though perhaps not all the way to "use my car whenever I leave the house" levels. I also think that once people begin replacing their cars with autonomous taxis, they'll sign up for all kinds of taxi subscriptions that will further reduce that reluctance. After all, driving your car now isn't completely free: it still costs you gas money, plus the hassle of actually driving it. And other forms of transportation aren't free either. So the bar here isn't 0.
I see this all the time and I just do not believe it is true. Uber/Lyft/etc undercut taxis for users to take market share, and have drastically raised prices to become marginally profitable.
Autonomous cars are more expensive, and the labor in non-autonomous cars is not the majority of the costs. In NYC, a 1hr Uber could easily cost $100 against a minimum wage of $15.
The idea that a taxi trip becoming cheaper than a car owners marginal car trip would require dramatic dropping of taxi prices. Even halving is not really going to do it, and I don't think removing the driver even halves the costs.
The autonomous taxi boosters also seem to overlook what happens to unattended, unmonitored public infrastructure in urban areas of this country. The reason I stopped using Zipcar in NYC was because they were typically trashed inside by the previous drivers. Now imagine an autonomous taxi that gets turned over 10x as often. Good luck.
To the extent that self driving taxi services are cheaper than human driven taxi services, they will also increase use of taxi services.
There's no reason to assume that on the balance people will end up driving less as a result of a technology that makes driving significantly more convenient simply because it might make taxi services somewhat cheaper and therefore potentially might make it easier to not own a car and encourage people to use other modes of transportation for some trips.
While I’d certainly prefer to watch Netflix than actively drive, I’ve still got stuff I need/want to do that I can’t in a car even as a passenger. And it’s just not comfortable for long periods of time. A lot of people get motion sickness staring at a screen in a moving car. Etc.
A lot of people own pickups just because they occasionally want to tow something or move something large. A lot of people own second cars for occasional use. These might become rentals instead when it can affordably just show up at my door in a half hour.
There’s no way to tell how this plays out. There will be some amount of induced demand, there will be some amount of reduction in use. One never knows which will be bigger.
What I do know is traffic deaths kill over 40,000 Americans a year, and driverless cars could potentially get that to 0 or near it, whereas human drivers cannot. I do know we can electrify cars and power them all with renewable energy, not immediately of course, and remove many of the environmental concerns. We can enhance mobility for the elderly and children and mentally disabled who can’t drive.
There’s a strange amount of anti-car propaganda that has gotten people worried about this, but I look forward to a driverless future in which cars are cheap, clean, safe, and available to all.
It’s not propaganda but jumbled concerns which are often poorly expressed. I think the strongest arguments are:
1. Self-driving cars don’t change pollution - even EVs are better for local air quality but still cause massive carbon emissions and unchanged or worse tire particulates, etc. – and may even make it worse locally with the extra mileage from taxi fleets.
2. Self-driving cars only lightly improve congestion, and then only to the extent that they can coordinate and you can ban non-AI drivers from certain chokepoints at certain times. The form factor unavoidably needs far more space per passenger than anything else.
3. Self-driving cars don’t really help with affordability – even if the current prices come closer to parity, that’s a financial stress for many people (e.g. in the region where I live, the average family spends as much on vehicles as they do food).
4. Self-driving safety needs a different relationship with the manufacturer. There are many areas where they can be safer but failures can also be correlated so we really need companies to share liability and have rigorous safety oversight.
As a pedestrian, I’m fairly bullish on the concept given how dangerous the average driver is now compared to 20 years ago but I worry that a lot of politicians are going to ignore the other issues because those require hard choices whereas it’s so compatible with American culture to say you can solve major problems by making an expensive purchase. These shouldn’t be opposing issues, of course, and I’d really like to combine them because autonomous vehicles should soon, if not already, be much better about following speed limits, staying out of bus lanes, etc. Making advanced automatic braking a requirement to enter a city could save thousands of lives every year.
I had half a mind to write a long treatise on why I think we'll only see significant EV adoption if/when cars become driverless, but I'll save it and just go with this. Someone I know was killed last week in a hit and run. She got in a minor car accident, got out to check on it, and a third driver hit her and took off.
When it comes to affordability, economists generally set the economic value of an average American life at ~$10 million, and 40,000 people die from traffic deaths every year. Even if we just look at the numbers, Americans buy about 3 million cars a year. So 40,000 * $10 million divided by 3 million is a savings of over $133k per car, which is far in excess of the average car's lifetime cost. Even a 50% reduction in deaths, which for all I know currently existing driverless cars could achieve, would be the same as making all cars free in terms of average cost.
And even if driverless cars are a total push in every other respect (and I think they'll be much better) 40,000 families a year (and I assume globally, at least 5x that) not losing a wife and mother that way is more than worth whatever we have to do to make it happen.
Stay safe.
That’s not the argument being made. Everyone knows they pollute less per mile – but unfortunately the manufacturing is roughly half of the lifetime pollution from a vehicle.
https://cleantechnica.com/2022/07/27/ucs-study-shows-lifetim...
https://www.iea.org/data-and-statistics/charts/comparative-l...
This matters especially because consumers have been getting heavily marketed into getting massive trucks and SUVs, where the sheer size of the vehicle means the lifetime emissions are greater than a small ICE because the lack of tailpipe emissions can’t make up for that even if it’s powered entirely off of renewables.
That doesn’t mean that we shouldn’t be electrifying the vehicle fleet quickly but it’s buying time on the trip to zero emissions, not a solution. Buses and e-bikes get us much further because they don’t suffer from emissions the inherent inefficiency of automobiles.
It's a free country: people are free to choose to use autonomous cars over ebikes and buses and why wouldn't they? The emissions profile of a personal electric car being unaffordable[0] doesn't pass the sniff test.
[0]Fair economic taxation of externalities - considering current status quo.
I don’t particularly love cars or anything, and would be really happy to not have to have one, but there’s no way I’m going to try to rely on buses or bikes. I value my time, too much for buses and my life, and not being either frozen or covered in sweat too much for any sort of bike.
A car gets you from point A to point B quickly, reliably, comfortably, and with cargo. Nothing else does that, and we are willing to spend a significant portion of our income for it.
I think taxing carbon would be a great way to encourage people to reconsider how they travel, and would expect many people to pick things like those small EVs for urban usage if that became common.
Studies are all over the map but the ones that put it anywhere near 50% are all from China.
If all you pay is the marginal cost then those that live an hour away will pay six times those that live ten minutes away
No way I can do that and be functional the next day, but if the car could drive itself, I’d probably be going.
I prefer Caltrain.
This would also reduce the cost on doordash type services so if instead of paying an extra $10 for your food/groceries/everyhing to be delivered you paid orders of magnitude less.
This might reduce the traffic on the road.
The pessimist in me makes me think once they got sufficient market share price would go back up and wed be worse off than before lol.
> even more cars on the streets
You don’t know that. I could make a prediction that it would lead to fewer cars on the street. Fewer parked cars especially.
> less investment into better modes of transport
I assume you mean subways, buses, and trams here. But I don’t think it’s fair to call them “better”. They’re hugely expensive and can be disruptive in many ways, are much less accessible.
> traffic will get a lot worse once people are ok with sitting in bad traffic
You also don’t know that traffic will get worse. Traffic could potentially get much better with better drivers. But also, if people are ok with it, then who cares?
Every single time scientists and city planners are called to answer "how we make the city more livable and reduce traffic" the answer is always better public transport (more trains especially).
The only part I could resonate with you is that we don't know whether SDC could lead to less cars. That's true if people will use more self driving taxis over personal cars.
Now this is a bad take. Public transit is _always_ better than individual vehicles when we are talking about a metropolitan area. The amount of resources, land, and pedestrian freedom that is eaten up for roads is insane. Imagine how many people can fit in a subway, and then expand that to each of them individually being in a car on the road.
Public transit is expensive, but so are new highways, highway maintenance, road accidents, speed enforcement... Etc. The worst thing is that many times people who don't own cars pay for those services they won't use. All the while public transit is getting it's funding cut.
I think the original comment is a little off in that more autonomous drivers does not directly lead to less public transit. But it is a concern that these profit/investor driven companies will be competing with public transportation and this has a lot of implications.
It's far cheaper to live in an autonomous motorhome that drives around all day and happens to arrive at work just as you need to be there each morning than to rent an apartment in San Francisco. Driving about is probably cheaper than paying for parking too, especially if you deliberately head for the busiest traffic.
You could already commute with cheap taxis (eg in the developing world). The more important thing is that people want to live in San Francisco, they wouldn’t be happy commuting from some far off place in the first place. And as stealth campers have already figured out, not a lot of places available to camp in your car even 40-60 miles out, so might as well be where you want.
But the idea that your car could just involve itself in traffic jams all day rather than pay for parking is interesting, it could also look for time limited but free parking and move on to somewhere else when that expired, which is more common outside the city. Heck, it could park at a shopping mall that doesn’t allow walk offs…because no one is walking off.
Already, though, you've got a budget of >$4000/mo for a more comfortable studio apartment.
Must as well have the cars in the US drive themselves so we can all get a nap at least.
You're being overly pessimistic. I can see the opposite occuring on each of your points.
- less traffic due to more efficient driving: once automated driving is pervasive it's natural that cars and traffic as a whole will coordinate and optimise use of the road. You should be able to predict traffic accurately and choose the optimal time to travel. Car speeds will coordinate to maximise flow through roads. Improved public transport will increase the number of passengers per vehicle and reduce personal vehicles.
- more investment into better modes of transport due to lowered costs: the cost structure of buses (and trains) lends itself to larger vehicles with less stops. Without having to pay someone to drive you can remake public transport into something that takes less people at a time to more places, without requiring expensive infrastructure. Think small automated busses that serve a web of points instead of routes, so people can request to get from A to B and the system delivers from as close to A and to as close to B as possible as soon as possible at the lowest cost.
- less car ownership: most people don't want to own cars, so it's very likely that car ownership will drop significantly. With new privately and publicly owned forms of public transport, the need to own a car will disappear in many cases.
I feel that almost all technology is positive (not sure about social media), since it generally gives people more choices and abilities. Automated cars have very few downsides.
- Increase cycling and walking because the roads are much safer
- Less noise from cars revving their engines, or being poorly maintained (holes in mufflers, underinflated tires, etc.)
- No carjacking
They have a ton of literature at https://waymo.com/research/ and tech talks on YouTube (search talks by Drago Anguelov). They make heavy use of simulators [1] where they simulate weather events and create their own weather maps [2]. It's a very sophisticated stack.
[1] https://waymo.com/blog/2021/06/SimulationCity.html
[2] https://waymo.com/blog/2022/11/using-cutting-edge-weather-re...
Its impressive how the lidar resolution evolved as per the youtube video. The color added, i wonder if its post-processing.
https://wondery.com/shows/how-i-built-this/episode/10386-the...
few examples : https://www.youtube.com/watch?v=S4aBNYcBoLI ; https://www.youtube.com/watch?v=B8TGFA6SfAo
Looking forward to expand its coverage to SFO, that will be a game-changer.
Still not sure of it economics though. Its current price is on-par with Uber Comfort / a little bit over Uber X. How that can support the R&D or future capital-heavy expansion?
I guess if it's showing enough promise to be profitable on its own (workout R&D and expansion costs), Google can probably spare a few more billions.
There you go - I would have said their operating cost is much lower. Paying the wages of the drivers for a year costs more than the car - even a car plus all those fancy LIDAR's.
Their development costs are a different story. I suspect only a company like Google could sustain it. But presumably it's one off, and if they spread it over 1 million taxi's in the USA it would only be a fraction of the revenue.
Those development costs have an upside too. It's a moat. If they pull if off they will have a monopoly. The will get away with being able to change just under the cost of a real driver for years. We may well be bitching here in a decades time at the obscene profits alphabet is making off us, and yet we have no obvious way out.
No, Cruise shared a few months ago that operating costs per mile is still higher than a car with a paid driver.
https://gmauthority.com/blog/2023/07/gms-cruise-operating-co...
If operating costs were below a human driver, they would be scaling as fast as possible to recoup the developmnent costs for the current level of tech, which have already been expended.
The page you link to says a) it (the entire operation?) is bleeding money, and b) it's operating costs are reducing, and c) quote efficiency improvements helping reduce the costs of simulation and machine learning used to test and improve AV performance.
I can't see where it explicitly says it's operating costs alone that exceed they call the magic threshold of $1/mile (which is presume is what a car + driver costs). And given they mention (c) above it seems unlikely they meant it was just operating costs.
That aside, they go onto say quote Operating costs per mile for Cruise autonomous vehicles or AVs has fallen by an average 15 percent monthly over the first half of 2023. That's an extraordinary rate for costs to fall. If indeed operating costs exceed that of a human driver it wouldn't have remained that way for long, had they continued.
Basically straight lines and 365 days of sun, now send them to Europe small towns/mountain roads/&c.
https://www.youtube.com/watch?v=RIyEg35Stbo
What do you want me to do? Pretend?
Perhaps MDS is real ...
You don't get brownie points if you deliberately hamstring yourselves (no sensors, no maps) and make the problem harder than it already is. It's like trying to jump higher and higher to fly to moon instead of using rockets. It doesn't mean you're tackling a fundamentally harder problem.
He repeatedly claimed that having contradictory radar and camera data means the system won't know what data to trust. Anybody who has worked with data in any capacity can tell you that is complete nonsense. In the past, people would call combining this data "sensor fusion." These days, even high school kids know you just throw all your features into a model. You don't discard red channel data from your cameras because it might conflict with green channel data. Tesla Insurance doesn't throw away all features except driver age when calculating premiums.
For what it's worth, as a pedestrian walking around SF late at night I absolutely do trust Waymo more than human driven cars.
The bigger problem I would imagine is finding the car not responding to inputs in a predictable way. I would image at a certain point they are going to give up. Even a human driver might need a push in some conditions.
It never snows, but between the fog and strong winter storms directly off the Pacific, visibility can often be near zero in SF. I've lost sight of lanes and surroundings many times.
tbh waymo's ceo said what I said here: https://www.cnet.com/roadshow/news/alphabet-google-waymo-ceo...
Even the straight, flat, streets can be nuts - like Polk St: pedestrians can jump out anywhere, slalom of delivery trucks, delivery scooters, ride-hail cars, double parallel-parked, with bike lanes - its a major cyclist route also busses and those little tourist scooters - oh and the electric unicycles that are always blowing stop-signs! the type, shape, size, and behavior is all over the map. go two blocks over and you have to deal with cable-pulled trolleys having the right of way across the 12% grade intersection. last week there were hundreds of drunken santa-clauses jay-walking between bars. oh and the line of tourist cars (who knows what traffic behavior they might follow!) heading up to see the crooked street - as the high school is getting out.
LOL get your memory checked, memory checked, if you recall driving in SF as unproblematic.
All the CEO said was that he thinks autonomous driving won't work in every condition or ever be perfect, and that they're always finding and working through unique challenges. That's quite a bit different from "our cars only work on straight lines and with 365 days of sun".
The traffic fatalities are higher (18 > 15), the lack of lanes is more egregious (and people will dive across three lanes more aggressively), the amount of driving the wrong way down busy multi lane roads is more (which still happens in Beijing, but like not as often).
Beirut (or at least the outskirts) also has things Beijing doesn't: civilian enforcement of traffic etiquette via brandishing automatic rifles, traffic intersections primarily controlled by tanks instead of signals, and the annoying practice carried over from Saudis of doing that stupid angled drive-with-the-car-tilted-on-two-wheels thing in thd middle of traffic.
It's similar to Beijing in it always being slow though, so that's probably why the fatality rate isn't 30+ lol.
But ya, none of the drivers were armed, so we had that at least :)
Nevertheless I do believe it might be easier to deploy self driving cars in more “feral” places where anything goes as traffic rules go. In those places, what I’ve observed, is that you can at any point actually come to a complete standstill and everyone will just navigate around you (within cities that is). This actually can work in the favor of these cars to be honest.
(BTW, Boston's reputation as a place for aggressive driving is undeserved. Insurance companies pay a lot for fender repair for the reason above, making the city look bad, but for injuries against a person, insurers just pay the limit, which an injured pedestrian can run up in the first 24 hours in a hospital. Anything beyond that doesn't show up in car insurance statistics. Houston and Austin have far more aggressive drivers, for instance.)
https://youtu.be/5wXO05s-pLc?si=5W-SW5zGXIwgnQpG
It makes a mistake in that one, around 5 minutes in, due to not understanding a construction worker's gesture, but I presume it phones home for advice and someone gets it moving again. Everything else seemed to be handled rather impressively.
Disclaimer: Google employee. My job has nothing to do with cars. But I do love technology and hate driving, so I'd love to see this problem solved. I'm actually quite skeptical that I'll ever have a truly self-driving car, as I also live in a place with weather.
Understanding someone directing traffic is probably one of the harder problems.
To another point, to the degree that it's predictable ahead of time, I'd be perfectly happy with a car that only could self-drive in some conditions. Only highway for some subset of weather conditions? That would still be super-useful for some of us.
It's just like how iPhones are wildly popular in the US. I'm sure there's someone who's like "Yeah, try to sell a $1k device to a Romanian and see how it goes" and Apple only gets 25% market share there but they're one of the most successful companies in the world, without Romania.
Some markets can be irrelevant.
It also gets quantitative measures of how confident it should be in what it just classified. And if your confidence measures decline, for any reason, it can just slow down. Small European towns should be doable.
First, as all the other respondents have pointed out, your characterization of San Francisco as "straight lines and 365 days of sun" is way, way off. But more importantly, as a consumer, I'd be thrilled to own an autonomous vehicle even if it didn't work in bad weather. There's easily enough data to have a car say "there is a storm coming in your area, can't drive autonomously" long before it would become a safety issue.
And, of course, wouldn't one expect an autonomous vehicle to start in places with better conditions vs yolo'ing it in a blizzard whiteout?
My 2023 car with cameras and radar usually cannot activate lane centering - even in good weather - because it cannot tell where the lanes are.
Having driven in both San Francisco and throughout the midwest, I agree with the parent commenter that San Francisco is a much easier environment for self-driving cars.
More and more I believe in Louis CK's bit about "everything is amazing and nobody's happy": https://www.youtube.com/watch?v=PdFB7q89_3U
Then there's the sheisters trying to sell that technology and convince governments that "for safety" people should be required to buy their product. Poking holes in the technology is people's natural response to the very real fear that their ability to operate a vehicle that isn't literally controlled by a huge corporation will be taken away.
The technology is amazing, humans suck.
You see the same thing with EVs, rather than it being a purely positive thing that the market has more options for people with different driving patterns you have governments committing to banning the sale of new ICE cars when the deployment is literally in its infancy
That feels like a massive straw man. I'm not aware of anyone at this point, or with any plans, to require the use of autonomous vehicles.
> You see the same thing with EVs, rather than it being a purely positive thing that the market has more options for people with different driving patterns you have governments committing to banning the sale of new ICE cars when the deployment is literally in its infancy
That also doesn't make sense to me. Governments don't want to push to EVs because the tech is some nirvana or something, but continued use of ICE vehicles is a large contributor to climate change and we have very limited time to address that.
Wait.. what? That’s definitely a new one to me…
Tech is already inundated with corporations using technology to maintain an iron grip on their products against their users interest and I'm just sad to see that autonomous driving seems to be the excuse to extend this grip further with always-on connectivity that you can't turn off "for safety" that will also conveniently used for even more tracking. I don't think it will be used for population control or anything like that but I expect it will be expand the already existing police remote shutdown.
Autonomous cars that are truly autonomous would be amazing. But what I fear we'll get is DrivingAsAService and regulatory capture.
I am pleased to introduce you to: https://en.wikipedia.org/wiki/Horseshoe_theory
I was just like the last thing in the universe I want is for a company to exercise Apple like control over my car. And despite making no logical sense companies seem to be using EVs and self-driving as an excuse to create artificial lock-in and add even more data collection. New and "different" seems to mean you're allowed to break consumer expectations for the worse.
If we do reach a point where self driving systems consistently outperfom humans in term or safety, I would absolutely expect a push to atleast require it in new vehicles if not some form of new law to push people towards using that technology. Personally, I find that idea attractive but I understand why people would fear it.
My bet is you'll be amazed by how little changes. Sure you won't spend any time at gas stations/car dealerships and you'll be able to watch YouTube on your commute, but that's how millions of New Yorkers live and, well, it's fine. That is to say, we could have just built subways/mass transit and gotten all of the benefits.
There is a major downside though, and it's that we'll be doubling down on cars. That means all the mining we have to do for all the components, all the road/vehicle maintenance we have to do, all the waste we have to manage, all the pollution from tires, all the space taken up by roads, garages, parking spaces, all the batteries we have to build/maintain/recycle, all the sprawl and isolation we suffer will increase.
I'm a big fan of public transportation, and I think it's sad that so many cities have underinvested. But it doesn't help your point to pretend that commuting on a bus or subway is the same thing as commuting in an autonomous personal vehicle, and these kinds of false equivalencies only serve to discredit mass transit advocates.
I mean the point is pretty straight forward. A study like this is tautological. "self driving cars are safe on exactly the fraction of hand selected roads where self driving companies are willing to put them" doesn't tell us anything. It's a fluff piece.
If you want to make an actual assessment about the safety you need data from randomized, representative roads, not 0.x% of the streets in America. Which itself is a very generous country to pick. Put these things into peak traffic in Rome for an hour
Tautological puff pieces are what Tesla does with many of its cherry-picked stats, but this data is much more definitive.
No, you can't. You can say the drivers are worse only on the segments where Waymo was willing to test it, and that is worthless. It literally is cherry picking. What does it matter to people driving in Boston, Chicago, New York, Paris or Berlin if cars can drive safely on a suburban grid in Arizona?
If you get to pick the sample you get to pick the outcome. We could literally put these cars on the worst 1% of roads and get the opposite result. In fact we could put them on the worst 90% of roads and get the opposite result. Even the average American consumer gets no meaningful info out of this.
Waymo is doing a level 4 rollout to specific areas where they are confident they can operate them safely. This study shows they are doing a good job with that. People in these markets get access to safer transportation once they can scale. Indeed, the majority of the US population lives in areas that are roughly comparable in driving complexity to either SF for Pheonix.
The real question is if these safety rates can be maintained while reducing the operational costs to the point where this can be scaled.
> You can say the drivers are worse only on the segments where Waymo was willing to test it, and that is worthless.
That's not worthless at all. There are tons of places in the US, maybe even a majority, that are quite similar to something between SF and Phoenix. Yes, absolutely, there are tons of places it's not. Again, so what? I just don't understand why anyone would expect that Waymo would not start with a progressive rollout.
Also, frankly, humans drive in situations where they just shouldn't. There are times where the conditions are so bad that the only correct decision is to just say it is not possible to safely drive today. Personally, I don't drive in dense fog because your options are basically drive too fast and risk hitting something you can't see, or drive at a sensible slow speed and risk being hit from behind - which isn't your fault but might still kill you. Even the most advanced aeroplanes with all the latest and greatest sensors still have defined limits where they won't fly.
Also, that Arc de Triomphe example should just be replaced with a safe priority system. Priorité à droite is a silly rule, and it's especially silly on a roundabout. In my opinion, crashes there are the fault of the road, not the driver.
how about no, let's not send waymo cars into vastly more challenging conditions. i want them to keep being cautious and responsible.
Not straight lines, lots of traffic, cars parked on both sides of roads, plenty of human drivers seemingly unable to cope.
(Admittedly, mostly sunny weather.)
What I'm interested in knowing is how these cars drive compared to the average competent driver in the exact same environment.
This depends on what type of drivers waymo is displacing. If it's displacing mostly professional drivers, then that substitution would be reasonable, because the choices we have are either professional drivers or waymo. However, if it's displacing "normal" drivers, including "first-time drivers, drunk drivers", then the substitution wouldn't be reasonable because banning waymo would mean those rides would be replaced with "normal" drivers, not super-safe professional drivers.
Even otherwise though, if at some point in the future I have the option of trading my car in for a fully autonomous one, the only thing I'm really interested in knowing is how my personal accident risk in getting from point A to point B changes compared to if I was driving myself. These benchmarks are meaningless in that regard. I'm not driving drunk, I'm not driving in bad weather conditions, I'm not being needlessly arrogant or reckless on the road. Can this car drive better than me?
but thanks to the gig economy, "professional taxi drivers" basically means anyone with a drivers license.
But there seems to be some stats. https://www.nytimes.com/2006/04/28/nyregion/that-wild-taxi-r... got "crash rates one-third lower". Still great result from Waymo.
When you say professional taxi drivers, do you mean people who drive for Uber/Lyft? Because that's who these companies are looking to replace.
I don't see millions running out and buying a real self-driving car kitted with a spinning lidar "hat" and visible radar transmitters sticking out everywhere, even if doing so meant safer roads for everyone.
What people in the market actually want is what Tesla has been selling (however fraudulently): a car that looks/performs very nice and claims full-self-driving capability, not a goofy looking car that self-drives very well in particular locations and use cases. Cars are about personal identity and power at least as much as they are about functional transportation.
I'd like to be wrong about all that, and would like a future where swarms of electric self-driving buses that route-optimize based on demand pick up people very close to where they are. But I also realize that the reptilian brains of consumers tend to decide how these things eventually pan out, and not the solutions that are optimized for efficiency and safety.
Hyper-commuting isn't really an aspiration for most people. The aspiration is usually either having the big house where land is expensive, or having the even bigger house where land is cheap plus not having to commute. More abstractly, there are 2 location luxuries: 1) proximity to opportunity, and 2) not needing the proximity to opportunity because you are wealthy enough not to care.
Furthermore, time spent in a car isn't time spent with family - which is presumably most peoples' reason for wanting a big house. Being stuck in a self-driving car for hours commuting doesn't give you that kind of time back.
> I doubt there are fundamental reasons why sensors couldn't be made more appealing for people who value aesthetics
My take is that making them more appealing means making them nearly invisible. To the best of my knowledge, in the case of lasers and other visible-light spectrum sensors, making them more invisible inhibits their function. That's probably a big reason Tesla dismissed lidar.
Also, if there's one thing Elon has taught us, it's that you sell more people on new car ideas by appealing to their base instincts, not their higher selves.
It can. Most would probably prefer a 1 hour self-driving commute to a 30 minute manually-driven commute. That hour of self-driving time can be used for chores (paying taxes, calling the plumber, etc.) that might otherwise have to be done during family time.
Most likely, the existing Waymo cars have their LiDAR sensors equipped like that because they need to be maintained and swapped at a regular basis.
I think the value proposition is more than enough for mass adoption, once people realize they can work and commute at the same time.
Wouldn't Uber/Lyft have already sopped up whatever demand exists for that? If anything Waymo is going to take that business away.
Here is a video of Waymo going through the Broadway Tunnel in SF back in Oct 2023 to give you a sense of it. >> https://mer.gy/broadwaytunnelwaymo
Which is by the way, a good reason to be skeptical of it. I remember talking to someone who worked with BMW on their self driving a long time ago and their take on Tesla's self driving effort was (a) It's fine for Tesla to have a bad reputation for safey but BMW simply can't choose to get a bad reputation for safety they sell far too many non-autonomous cars and (b) It's actually not fine for Tesla (and others) to be rushing ahead with self-driving because they will kill people and they're just as likely to kill the whole self-driving industry at the same time.
I have no doubt that even if the data looked terrible, Waymo would find a way to spin it to look safe. I also have no doubt that even if the data is good, it's not indicative of self-driving being safer in the average situation.
Sure, I'm not so naive to think there's no corporate spin, and especially with Alphabet's marketing and media resources and the need for positive sentiment towards Waymo, but I think if Waymo's safety record were anywhere near Tesla's (or Cruise's) it would be very hard to spin this well and not be refuted, they would be better off not giving an update in that case.
I'm not saying this is happening - but it would be entirely unsurprising to me if it were. And I think actually there are some rumours of Tesla engaging in that kind of behaviour - classifying crashes as being unrelated to Autopilot because Autopilot disengages in the moments leading up to the crash. How would anyone refute that? And if they tried, do you think they'd face significant legal jeopardy from Tesla.
In some places, this is a real choice to be made, and developed world luxuries of “holding professionals to higher standards” aren’t available.
US is a beast where super majority of people have given up on public transport and infrastructure. Otherwise I would hold the bar higher and look into alternatives.
We are here to offer you a correction of this regrettable mistake. Could you please call your family and follow us to the flesh compactor? We assure you that it is operated by a real, drunk human.
Consider that the entire article starts with, “Safety leads everything we do at Waymo.” A clear and almost painfully direct PR to what took down Cruise last week. You could almost add “unlike our competitors” to the headlines haha
[1] https://www.nhtsa.gov/sites/nhtsa.gov/files/2023-04/Second-A...
(Over ~4 Cruise rides, I’d put them closer to median)
Humans die from driving only once every 100 million miles on average.
So until there is a comparison at this scale, to me it’s a very incomplete picture of safety to say that you are outperforming humans.
If driving my own car means colliding 100x as often but I expect to die 2x less often, I would consider self driving to be much much much more unsafe. There is really no way for me to understand the risk of fatalities from a sample of only 7 million miles, since either humans or a system with 2x the fatality rate of humans would both be expected to have 0 fatalities at this scale.
Given we are at 7m miles, hopefully this comparison is coming soon and I will be much more convinced.
The alternative requires a scenario where waymo is especially likely to get into fatal accidents while being very good at avoiding non-fatal ones, right? Seems far-fetched.
But regardless, the claim here isn’t “under some assumptions that some people (and not others) find reasonable, we can extrapolate and predict that self driving cars will be found to be safer”
It’s “self driving cars are safer”, which there isn’t enough evidence to claim yet.
- The major caveat is that Waymo is not being directly compared against sober humans driving lawfully.[1] The reason why this caveat is so important is that technology which makes it impossible for humans to exceed a posted speed limit might be overall much safer than replacing human drivers with autonomous drivers. Uber isn't more dangerous than Waymo because humans are incompetent, it's because humans obey orders from impatient drivers and Waymo currently does not. This is a UI choice, not an AI advancement.
- More specifically, lawful driving is an important caveat because Tesla Autopilot had two different settings for driving unlawfully, according to the users' own sense of personal risk. An AV manufacturer who advertises "AI-assisted speeding" will almost certainly find a lot of customers, even if it's under the table. People don't speed and run red lights because they're too stupid to understand why it's dangerous: they do it because they're reckless and selfish. AI won't stop that, only regulation will.
- Another caveat is that Waymo was trained on human-dominated streets. Waymo being safer in a sea of human vehicles does not actually translate to Waymo being safer in a sea of Waymos. I think this is a low-probability risk but it's hardly a simple question: I believe Waymo has had issues where several AVs occupied the same street after an event and blocked traffic because they couldn't decide what to do - they were waiting on each other to behave like a human. But again, the risk seems like gridlock, not property damage or injury.
- And a minor but still important caveat is that SF and Phoenix have modern linear grids which have been mapped to death by AV manufacturers. As a Boston resident I am still holding my breath about their performance here :)
[1] Not because of anything insidious, it's just a granularity that both the analysis and the data struggle to capture.
There is the risk that they will misbehave or deadlock around each other, but also a great opportunity for them to communicate intent to each other at a level human drivers just can't do from a sound-insulated cabin.
I am not saying that it's doomed to fail - simulations can probably do most of the heavy lifting, and of course the tech itself will advance. I am saying it's a mistake to assume stuff like this is somehow a free lunch, or even "the easy part." It's unknowns about unknowns.
The hardest thing about driving in Boston is that the streets are wildly non-rectilinear. Waymos obviously don't understand "the rules of the road," they are exhaustively trained on the rules of the road until they can imitate them. For rectilinear intersections they can transfer this training to a variety of different streets; not so in Boston, where you might have a weird three-way intersection involving oblique angles. I am also not sure if SF / Phoenix have many roundabouts. Or, near me, this horrific pair of four-way-intersections somehow combining into a five-way(?) intersection: https://www.google.com/maps/place/42%C2%B022'50.0%22N+71%C2%... I suspect Waymo would get badly confused here unless it was directly trained on this specific intersection.
You're thinking of some of the Cruise incidents like https://m.youtube.com/watch?v=kyAX28dapps .
> And a minor but still important caveat is that SF and Phoenix have modern linear grids
San Francisco has some amusing areas as well. If you look around the streets near Buena Vista park or even Twin Peaks, you'll see plenty of funky curves, six way "intersections" and so on. Lots of two-ways that are only lane wide.
Like the older parts of Boston and Cambridge, it's a holdover from the days of horses and the geography. That's where the paths were, people built houses, and now you've got to deal with it :).
Right, because Waymo has now tuned their cars to cut off each other even if it results in illegal and unsafe turns https://youtu.be/kN0MLclnWa0?si=5yM847zMqnH90d52
You can't claim better than human when humans are driving under different conditions.
I looked around to try to find the actual data but it's just marketing materials. Are these miles on the same roads under the same traffic and weather conditions?
Unlike, Uber , Tesla and Cruise.
But if you cannot see the issue with what has been promised, and what has been delivered, you're not living in vapour world, you're living in kool-aid world. You should be insulted, but instead you drink it up.
Again, no offense, you make your choices, everyone else makes theres, and the world moves on ... but the FSD saga has been an absolute sh1tshow by any reasonable definition.
vapourware /ˈveɪpəwɛː/ nouninformal•Computing noun: vapourware; noun: vaporware
software or hardware that has been advertised but is not yet available to buy, either because it is only a concept or because it is still being written or designed.Very much still being written and designed. The fact that people are paying for a beta product is problematic to me, but that goes back to the snake-oil argument - a phenomenon well understood over time.
Hence, let's shake hands on: condensed-vapourware.
Good day.
And that's not even the start of it. Is FSD feature complete but with bugs for just basic everyday use? No. And hasn't been for years. It's been revised and messed around since the start and is nowhere near ready for primetime, and possibly a decade or two away from what was promised near a decade ago.
I understand development processes, but not while you, me, or our families are on-board. If you're cool being the guinnea pig? Cool, be a guinnea pig. It's a free country, but that will not change my opinon about Tesla's.
Let's end this now. You love Tesla's and FSD. Cool, good for you.
I think their technology and engineers are impressive, but their quality as a carmaker, organisation, and their design language are absolutely sh1t ... (and I'm an EV fan, so don't go there).
Agree to disagree.
There's no point lying about how finicky it is, or the number of disengagements you encounter. There's too much evidence to the contrary.
Recently one of the biggest FSD proponents on Twitter put out, essentially, a promotional video for FSD, and it showed multiple disengagements and an incredibly dangerous blown stop sign.
Nobody cares that half your commute you can sit there pretending not to drive down a highway. That's not full self driving.
I have the software. I use it daily. I am happy with it.
Somebody in twitter did a dangerous demonstration. This doesn’t change much.
Let’s talk about a different technology and maybe that could help you understand this:
There was a time when consumer drone tech would be looked at as “vaporware”. Almost certainly somebody in the 90s advertised something looking functionally like a DJI phantom and never delivered on it. Pretty much ever you robot I got advertised as a child comes to mind.
If somebody took a DJI phantom and crashed it into a building, does that make the current, existing, very real technology into “vaporware” in your mind?
There was a time when self driving tech was also vaporware, but that time has passed. Somebody using existing, very real technology in a stupid way doesn’t mean that it suddenly doesn’t exist.
Don't forget, it was actually a comparison between Waymo and Tesla in SF. They tried so hard to show Tesla was better, but it was embarrassing how flat out dangerous it was.
Here's the compilation: https://twitter.com/elaifresh/status/1727889381309239731
Our economy is built on a regulatory framework that protects consumers from companies that cut corners.
I've heard people say that Tesla is in the right because in the long term the tech will save more lives.
That's nonsense. If Tesla didn't cut corners, Waymo would get there just as fast as Tesla will cutting corners.
They’ve also driven by far the most miles. I’m no diehard Tesla fan, but I hate it when people don’t properly scale their data.
Tesla killed people, it is up to them to demonstrate the rate is sufficiently low to be considered safe. However, if Tesla chooses, as they do, to not produce robust evidence of its death rate that competent third parties are allowed to audit, then it should be presumed to be unsafe. Anything less does not meet even the basic standards expected of safety-critical product development.
1. "We are leaders in FSD technology, but you have to hover your hands over the wheel and foot over the brake to take control at any second."
2. "Our FSD is far safer than manual driving, so says our data. But you have to take our word for it, because we won't release the data for research."
But it looks like Google/Waymo is the only one with reliable self driving technology.
If they don't give up like Google typically does. This could actually be a huge business for them.
Everyone I know who has used the Waymo service had good things to say about it and we are just at the beginning.
The FSD problem will become easily as you scale up due to the fact that these AI can easily communicate with all the cars around them unlike human drivers.
Disclaimer: Google LLC employee. No relations to Waymo, just a fan of self driving in general.
As a long time anti-abuse engineer, I cannot agree. The game theoretic aspect of trusting data acquired from outside parties cannot be overcome. Cooperative AVs will never exist.
He's been saying that for a long time. I think it was 2017. "By 2019, it will be financially irresponsible not to own a Tesla." and that the average owner would be making $30K a year with their Tesla out taxiing when they weren't using the vehicle.
Which should have been alarm bells for the bullshit meter for anyone with a room temperature IQ. "We've designed this money-printing machine. And rather than keeping all that money for ourselves, we're going to sell them all to you! (After we charge you $12,000 for it...)"
And you know that if Robotaxi Tesla ever does actually appear, it will be another charge atop the FSD for that software too.
All it does is give me vibes of "Here's how you can get rich quick, like me: step 1, give me money to hear me tell you how people can get rich quick."
I really like this because I think this happens a lot with human drivers. There are many instances where there was a potential crash that was avoided by me or the other drivers. I think that's crucial to autonomous driving. Not only should they have the ability to make mistake but to also compensate for other drivers' mistakes. It's all very good to say "we did everything we were supposed to" after an incident but it's even better to never have the incident at all. An AI that can react well to the unexpected would be a huge milestone.
I'm not actually sure why everyone is so concerned about driving in winter conditions. What is the primary concern, other than sensors freezing? As a human driver, I find the biggest problem to be other human drivers who insist on driving fast even in winter weather. It seems like almost all of the complications could be avoided by slowing down.
All of these are very hard problems that aren't quite addressed by a bus driving on cleared bare roads. It's good to see someone tried to work on them instead of just pretending all of earth is like SF. Too bad it seems to have failed.
https://www.forbes.com/sites/nargessbanks/2019/03/15/muji-ga...
However it appears the local company that made the autonomous driving system went bankrupt:
https://truckandbusbuilder.com/article/2023/07/18/sensible-4...
It makes you wonder how autonomous driving systems get supported after their maker's bankruptcy.
It is still pretty impressive, but we should compare average speed and such.
But a lot of that data also seems to come from San Francisco, so I have to admit I'm impressed.
(In constrast to say, Albuquerque. Drivers are a lot nicer and polite. They just have traffic signals that are confusing, and the regulations on signs and lines placement sets you up for failure).
But to me, I often wonder about once this tech comes out what happens once theoretically all of the self driving cars can communicate with each other? So if you need to break suddenly, instead of just your car than its a series of cars that can do it together or example?
I have been wondering if there has been any talk about a standard around doing this proposed since to me that is the true power of self driving cars, not just how do they operate with other unpredictable drivers.
Then of course there are the issues with car-to-car communication. It would open up tons of opportunities for bugs, hacks, and other forms of abuse.
That said, if Waymo _can_ drive as safe as my the typical defensive human driver, without the need for any human driver to take over, then I would applaud it. Maybe it does, I only follow this news from a distance. We all know what happened with the Tesla hype but I'm more willing to give Waymo the benefit of the doubt. Waiting patiently for self driving cars to become available in my city.
Yep, for instance: If the car gets confused and can't proceed, what is the outcome? Well it's not a "disengagement" or recorded as anything publicly. It's just "the car contacting support for directions."
Wanna see how this plays out? Here's Cruise, bragging about reaching 1 disengagement per 96,000 miles[0], sounds impressive right? Meanwhile, in reality Cruise cars needed remote assistance every 4-5 miles[1].
An even more fun fact? Cruise had at least 60 vehicles stall at once in 2022[2], but somehow they claim only "9 disengagements."
[0]: https://www.eetimes.com/waymo-cruise-dominate-av-testing/
[1]: https://www.cnbc.com/2023/11/06/cruise-confirms-robotaxis-re...
[2]: https://jalopnik.com/cruise-server-crash-causes-self-driving...
Waymo has the same problem of stalling and contacting support[0], we just don't know the numbers.
As an example, [0] is video of a Waymo stalling (starting at 0:14) for 50 seconds before even contacting rider support. Rider support then takes an additional 47 seconds to reroute before the car starts moving again. The entire time (1m, 37s) the Waymo is blocking the only lane through this area.
There is no safety driver. This study is discussing the fully driverless level 4 autonomous vehicles that Waymo operates in California and Arizona.
This particular study show that those vehicles had accident rates 3-9x lower than human drivers, even after controlling for the limitations on where they operate (such as not operating on highways).
If we made all cars self driving and they killed 30,000 per year instead of 40,000, the story becomes "Tesla and Google" murder 30,000 Americans every year. I don't think these companies could survive that amount of bad PR and legal liability.
And yes, I get that self driving cars have safety network effects, but they need to be at least 1 order of magnitude safer, if not more BEFORE those network effects start to compound.
If the cause of the crash is the program, it's the company's fault. If the cause of the crash is a person, it's the person's fault. It's not that complicated.
I'm not interested in a self-driving car being better than the average person at driving. I'm already better-than-average.
The cars need to be as good as the best drivers out there.
> The cars need to be as good as the best drivers out there.
They're getting there. Self-driving vehicles already have huge advantages in sensors, attention, and reaction time. Progress on reasoning is improving steadily.
But self driving cars can interpret them at the speed of light.
Given the life and death potential, I consider any signal I get from another car "trust but verify". Having a radio signal about a car a few cars up would maybe get me to slow down but not emergency brake until I had more information.
The hacking potential is just to great.
Is it enough to support a pretty extraordinary claim that a fully self-driving system that will then be sold in millions of vehicles is now perfectly safe?
This is an incredible volume of data, and very encouraging. What it will justify is the ramp from 10k trips a week to 100k trips a week.
As they go, they'll start rapidly collecting much more safety data, which will then be used to justify the ramp to 1M trips a week.
As they roll out to new environments (snow, harder and new urban settings, airports) they're taking it slowly enough to collect strong evidence that it's a justified decision.
And local regulators are scrutinizing this aggressively.
I’m imagining a fleet of cars driving in generally sunny weather cities of SF, LA, and Phoenix, driving every street thousands of times over. After all that, saying they perform better than a human just doesn’t wow me that much.
After 40M miles of testing, bring it to Chicago, Boston and NY and let us see how they do in the summer and then in winter. Are they going to do terribly at first because they need another 100M+ more miles of training?
Tyler Cowen: Uncertainty should not paralyse you. Try to do your best, pursue maximum expected value, and just avoid the moral nervousness. Be a little Straussian about it. Like here's a rule, on average it's a good rule, we're all gonna follow it. Bravo, go on to the next thing. Be a builder.
Joe Walker: Get on with it?
Tyler Cowen: Yes. Because ultimately the nervous Nellies, they're not philosophically sophisticated, they're overindulging in their own neuroticism when you get right down to it. So it's not like there's some brute let's be a builder' view and then in contrast there's some deeper wisdom that the real philosophers pursue. It's: you be a builder or you're a nervous Nelly. Take your pick. I say be a builder.
Also, I'd find the statistics more believable if they were published by an unbiased third party.
The other issue I see is that I believe Waymo doesn’t live map the terrain in SF and Arizona right? They are using intricately created maps and so my question is what’s the leap for Waymo to operate with this success in say Pittsburgh?
My theory is that as we face climate change more and more and as the economy goes out of whack owning a car will be a privilege for the rich and super rich. Sort of like it's already the case in parts of Europe.
At that point this self-driving crap is going to look ridiculous. Even without it it's ridiculous testing this technology on public roads and gambling with everyone's life.
Cruise's problem was hiding the footage from the regulators. As the saying goes, “It's not the crime, it's the cover-up.”
https://www.cbsnews.com/sanfrancisco/news/gms-cruise-robotax...
> In a Dec. 1 filing recounting how Cruise handled disclosures about the accident, the Public Utilities Commission asserted the company tried to conceal how its robotaxi reacted to the accident for more than two weeks. Cruise didn't provide the video footage until Oct. 19, according to the regulatory filing. The cover-up spanned 15 days, according to the PUC, exposing Cruise and GM to potential fines of $100,000 per day, or $1.5 million.
While a car "dragging" someone can be traumatic and horrible in it's own right... objectively, if we could get rid of thousands of deaths a year in exchange for fewer "mishaps"?
In 2021 2,116 people age 15-20 died in car crashes. 13,384 people for DUI related deaths in 2021. 42,939 car related deaths.
If we could exchange 42k deaths for 1,000? wouldn't that be worth the tradeoff?
Basically the trolly problem where you have to choose between letting a train go forward and kill 5 people or hitting a button to change tracks and kill 1.
If we can get into the "9's" in accuracy and trust bypassing humans... why wouldn't we?
If you’re drunk you’re going to have pay for a ride either by a person or possible by program or you can break the law. I don’t see anyone who was going to drive drunk not do it because they could pay for a robot taxi over a real taxi. Here’s my comparison. If your self-driving car kills or maims someone in a situation where a normal sober person driving wouldn’t have it shouldn’t be on the road.
What if it does this while also _not_ killing/harming in many other situations where a human would have?
"if your thing makes a mistake it shouldn't be on the road" yet the roads are full of people that make mistakes. (Hint: thousands of them die every year learning. See above about kids dying each yar).
So you'll focus on one of many reasons to sidestep the question (bias of some sort?) so I'll say it again:
At some point computers will be smart enough, fast enough and safe enough to do a better job at driving than people. Mistakes will be made but - like drunk drivers - humans make mistakes too. So if we can move to a safer way... why wouldn't we? Why are you willing to sacrifice thousands of kids lives if we can eventually progress to a better and safer option?
that's like asking "do you trust a spreadsheet where the math is done by a computer and not verified by a human".
At some point in history... the answer would be "no way I'd trust numbers not checked by a human".
Today? It's easy to say we can trust software produced numbers.
The question becomes what is that cross over point?
Are you saying you don't think we'll get to a point where computers drive better than humans? I think we're closer to that point than you think and with defensive programming we can do so safely.
Have you been paying attention to the status quo at all? https://www.iihs.org/topics/fatality-statistics/detail/state...
If not, you might want to. https://jakeseliger.com/2019/12/16/maybe-cars-are-just-reall...
I live in Arizona and have taken Waymo a bunch. It's great! https://twitter.com/seligerj/status/1707259311809073313/vide...
Humans make mistakes and car accidents happen. We don't and won't accept the same from machines. Why is that so hard to understand?
It seems like we should have a lot more small single occupant vehicles that are effectively caged motorcycles, but actually safe.
It’s not fair to compare it to drunk drivers, or to the small portion of people (likely causing an oversized portion of accidents) who shouldn’t have a license.
I’d like to see these comparisons made against a typical good driver, who never drivers impaired and obeys the law.
I’m simply saying that factors unrelated to driving skill should be excluded. Things like being inebriated, purposely breaking driving laws, looking at a cell phone.
The problem is that studies like these are used to imply that AVs are more skilled than humans at driving, but it’s not a conclusion that can be drawn, because the AVs are operating optimally and some percentage of humans are operating suboptimally.
What results like this actually show are that replacing all human drivers with AVs would likely result in less accidents (because by definition you’re excluding people who are malicious), which is not the same thing as saying that AVs are more skilled at driving.
What is driving skill if not knowing not to do these things?
Removing the mistakes from the human data makes the whole exercise futile. The whole idea is people fuck up. A lot.
Consider this: suppose that Waymo drives more safely than "US drivers overall" but slightly less safely than "US drivers who never drive impaired". Then… replacing a uniform slice of US drivers with Waymo vehicles still improves safety! It would only be problematic if you somehow adverse-selected by replacing "the safest" human drivers with Waymo but letting the most dangerous ones continue to drive by themselves. Which is interesting to consider, but I don't see why it would be the case.
I'm not sure that we've actually tried very hard. We could, say, mandate an ignition interlock device in every vehicle. Pretty draconian, sure, but outlawing human drivers would also be.
Point being, if the problem with human drivers lies outside of some fundamental flaw, there are lots of solutions that can be explored beyond just self driving cars.
For what it's worth, it's obvious that AVs will exceed human skill at some point. I just don't think we're there yet, and none of the data I've seen has swayed me.
Spend an hour watching this channel: https://www.youtube.com/@DashcamLessons
If you have a strong stomach, search youtube for "brutal and fatal".
Humans are shit drivers. Remember that when you hand the car keys over to your teenager.
Of course other objects such as pedestrians etc will always make things more complicated.
However, while differing vehicle types were mentioned as a source of variation, there was almost no indication that this factor was applied to the numbers. Also, my understanding is that this service does limit its coverage area, so I’m curious what sort of impact that has on the numbers.
One other interesting fact. They claim 7+ million miles on 700,000 trips. So the average trip is over 10 miles, which I found surprising, but perhaps I shouldn’t since most of the data is probably from the Phoenix area.
What is the crash rate for humans under good to light rain/fog conditions? This doesn’t seem to be comparing apples to apples.
[1] https://assets.ctfassets.net/e6t5diu0txbw/54ngcIlGK4EZnUapYv...
:)
Edit to add this link to some footage: https://twitter.com/dmitri_dolgov/status/1489318507342807041
[1] https://www.americanrallyassociation.org/
I've always said that self-crashing cars will win my stamp of approval when they can compete in stage rally. It'd likely push the Waymo team to evolve even further by getting involved in something like that, even with just the basics.
I would think a better scoring system should be based upon human interventions.
tell us who on the programming team will act as tribute and you can have your mostly autonomous wealth consumption device.
remember, it's driving itself, by itself, that's wasting resources when public transportation would eliminate orders of magnitude more issues.
I think a lot of that was done in a completely flat and mostly empty suburb of Phoenix where the accident and death rates may be much much lower than the national average.
OK they have a lower crash rate compared to humans. They also just stop in the middle of the street when they get confused and do nothing until someone remotes in to drive them.
I’m sure humans would do a lot better if every time they got unsure they just stopped and never moved again.
Waymo is clearly out in front by a few hundred miles, but touting this seems a little disingenuous to me.
Or even against humans... not as a claim that is unverified, but to work to verify the claim with a third party.
This is of course great news for people living in the desert, but let us know when they achieve this in Toronto or Boston year-round.
Try the same thing in a place with winter on a normal suburban city road. The road edges and evolving swarm-defined lanes have little to do with the absolute GPS position of the unobscured lanes and edges. An all the road makings are often obscured for weeks at at time (or longer). The road surface snow looks just like the road edge snow. And it's a semi-permanent slippery surface.
There are a lot of challenges left in non-cherry picked regions before autonomous driving can be said to "outperform comparable human benchmarks" without this qualifier.
I guess it's pointless to point out the WeWork-level of creative reporting bcs the "if we can just save one life" people don't want to hear it. The reality outside of the engineering bubble is that self-driving is an expensive dud with a smaller potential market than initially hoped (major cities like NYC will not see self-driving anytime soon). The expectation that there just needs to be a little more time for development has run its course and money for speculative bets are harder to come by (especially for something that's burned through so much cash already).
Compare this to Tesla's AI-first approach, aggregating training data from millions of drivers across the US. There's no overhead to introducing Autopilot in a new city it's never been in before--it just "works," with no expensive 3D mapping required, by extrapolating cheaply acquired vision data from other cities.
Elon really got this right from the beginning, but Waymo's aggressive PR and marketing have pitted public opinion against Tesla, as is very clearly evident even in these HN comments.
Yet it actually works. Going to level 4 and slowly adding areas prioritizes safety and reduces the risk of project failure. You can argue the Waymo approach is slower, but is the only one that has worked so far.
> There's no overhead to introducing Autopilot in a new city it's never been in before--it just "works,"
Yet it doesn't. For all their grandstanding, Tesla has remained stuck at level 3 for years. I wouldn't say that trading cheap deployment for public safety is something we should praise.
For certain values of "works"
> Elon really got this right from the beginning, but Waymo's aggressive PR and marketing have pitted public opinion against Tesla
Consider the opposite
Um, yes.
Watch videos from Waymo customers in San Francisco. There are plenty of them online. See Waymo cars navigate San Francisco. See them cope successfully with traffic cones, bicycles, cable car tracks tracks, double-parked delivery trucks, small animals, construction, and wandering druggies.[1][2] Totally automatic. Nobody behind the wheel.
Then watch videos of Tesla self-driving in a city, with the latest "beta self driving". Two manual interventions in the first minute.[3]
Waymo has real self-driving. Tesla has fake self-driving.
[1] https://www.youtube.com/watch?v=t7LXFjQ7hHs
I would encourage you to review the actual data. FSD is currently the industry leader in numbers of miles driven per incident. https://x.com/TashaARK/status/1727463951254863949
I expect this to improve even more with FSD 12.
All of this is easily proven wrong. Waymo works in rain and fog in hilly San Francisco flawlessly. There are YouTube videos anyone can look up. One look at https://waymo.com/research/ tells you they're using more AI than everyone else combined. They've used an ML-based planner for years now. Saying they have "hardcoded" to drive at the scale of entire cities is just dumb. Somehow only Tesla fans think it's possible.
Tesla has no overhead introducing in a new city because, obviously, it's a driver assistance system. It "works" only because there's a driver ready to prevent crashes. That's not a fully autonomous system by any stretch of imagination.
update: I see we have some Waymo engineers in the house!!!!!!! (which is why I'm getting downvoted)
Edit: I am wrong Ill take the L
But most airplanes still have pilots. We are talking about cars with no drivers.
https://arstechnica.com/cars/2019/03/arizona-prosecutor-wont...
https://en.wikipedia.org/wiki/Death_of_Elaine_Herzberg#Legal...
Is this true? My admittedly small data sample suggests it isn't.
Source: idk
If the process of this company's brakes failing reaches a sufficient degree of negligence, those people will go to jail.
However the process can take some time. For instance, only a couple of people were sent to jail for Volkswagen's fraud since Germany protects its own for the most part. One of them was jailed when the US grabbed him in Florida while he was trying to return to Germany from being on vacation.
And if a company is found to produce cars that kill an unexpected number of people we already have criminal laws to deal with it. These are hardly the first heavy machines we build. Even on normal cars you already have this issue with stuck pedals, engines that might not turn off and similar manufacturing defects.
If a car causes injury or death or property damage due to a manufacturing defect, then there is chain-of-commerce liability for everyone between the manufacturer and the buyer, most jurisdictions make the operator responsible for assuring it is maintained in safe condition as well. And none of these liabilities excluded the others.
But, no, usually, unless there is an unusual degree of intentionaloty and/or deception, no one will go to prison.
As someone who has lost close ones to car accidents, I don't give a shit about liability if we reduce casualties by an order of magnitude. I didn't downvote you but it doesn't surprise me that your take is unpopular.
This technology will increase the issues we have as a society around the use of a car based transportation system, as they are simultaneously expensive to maintain for low density environments while also conversely being almost impossible to maintain for high density environments.
Car transport systems have no true limit on capacity so as the efficiency of the transportation system increases, so will demand and density.
If everyone changes to another transportation system powered by Waymo and others, what’s to stop a large number of people moving to self-driving cars, increasing the number on the road, and causing grid lock again meaning we have to increase massively the amount of car-centric infrastructure all over again. Isn’t that what history has shown us will happen?
Why do we keep supporting the most inefficient and expensive form of transport ever?
Downvote if you must but it is a warranted and substitute argument and worth considering. Even if you think it’s not or not on the appropriate post of whatever criticisms you wish to levy (or just as likely work on self-driving systems). It’s always amazing to me how hard people will defend the status quo while calling themselves “technology innovators.”
Public mass transit infrastructure can't really be created by private companies.
So, if you want mass transit, driverless cars really has nothing to do with it. Petition the government to get off their asses and do their job.
Not sure that’s actually true.
Japan’s has one of the most effective models and contrary to what most people believe is privately owned and operated. About 70 percent of Japan's railway network is operated by the Japan Railways (JR), while the rest is served by dozens of other private railway companies, especially in and around metropolitan areas.
https://ftp.dot.state.tx.us/pub/txdot-info/rail/high_speed/s...
There is a lot more going on here with self-driving cars being pushed as the only answer that is deeply worthy of examination and discussion, but private companies have been extremely successful in the past before the oil and car companies started acting like cartels and systematically dismantled and destroyed public transportation across the western world.
Keeping “public transit” as a government only activity while convincing the public that government is impotent at its implementation was the core strategy.
https://www.quora.com/Are-there-any-paper-trails-showing-car...
https://medium.com/modern-city/how-the-united-states-ended-u...
Look, I drive. I just think we can do better and deserve to have better conversations about alternatives than what we’re having.
But I don’t agree with you on your other point. Look at the first link I offered. We have our first commercial high speed rail project going online in the US and they are already breaking ground on the second — LA to Vegas.
We also have a long history of successful private transit systems.
As has been written on hacker news before:
> The public transit system used to be privately owned and operated. Then the cities took over and kicked out the private sector. Now we have terrible public transit.
> Even the subways in NYC were built and operated by private companies (there were two, and they competed).
Truth is we need both.
Self-driving vehicles paired with congestion taxes could usher in a golden age of mass transit. The congestion tax is the important bit though.