I rode in a Waymo recently and was amazed at how good it was. It was easily in the top 5 best drivers I have ridden with. Really really impressive.
It wasn’t nearly as cautious or timid as I expected it to be. But it also wasn’t reckless. I would describe it as assertive, but I’m sure some people would call it aggressive.
It knows the rules and what’s going on around and acts accordingly. This really should be the norm.
For the first time in a decade I’m excited about an emerging technology. The future is bright.
I was just in CA for a few weeks and drove in SF quite a bit. There are Waymo cars everywhere and from the perspective of a driver I was more than a little cautious driving behind or next to one of them.
I would rate their driving as quite good, but their signaling of intent was terrible. More than once did I see them use turn signals, only to never actually turn. Turn the signals on, then off again, then on again, eventually turn perhaps but not always.
Humans are terrible at this too, but we also use more subtle and inadvertent signals like the slight movement to the line telling me they're likely about to switch lanes, only to see the signal go on halfway into it. Those subtle signals are easier to pick up on than a robot driving more or less perfectly straight while signaling, turning of signals, turning them back on again, and so on. Also super creepy with empty driver seats, and no one responsible to make eye contact with.
But the actual driving wasn't bad, as it turns out. Never had or saw a near miss, anything reckless or otherwise remarkable. I can see myself getting used to this tech and I would love for it to succeed, because man driving for anything but fun really sucks.
>More than once did I see them use turn signals, only to never actually turn. Turn the signals on, then off again, then on again, eventually turn perhaps but not always.
>Humans are terrible at this too, but we also use more subtle and inadvertent signals like the slight movement to the line telling me they're likely about to switch lanes, only to see the signal go on halfway into it.
Isn't waymo's behavior preferred? I'd rather see turn signals well in advance of any possible lane changes, rather than the moment before the cars actually start changing lanes. Sure, it'd be ideal if the turn signals were well in advance and there were no false positives, but road conditions are changing constantly, so by necessity you're going to be trading off between advance notice and false positives.
He said he can’t trust the turn signals. I’m case of a human driver, there would be the behavioral cues to fall back on. But with Waymo, you can’t predict its next move at all.
Not explicitly, I think. But if it uses the turn signals without turning, I wouldn’t feel to confident that an actual turn would be preceded by a signal, much less one with enough time for me to react (“I will turn” vs “I am turning now”).
The Waymo failure mode of false positives (signals but doesn't turn) seems much better than the human failure mode of false negatives (doesn't signal but turns)!
There is a signal light mounted on the top of the vehicle. When it detects a pedestrian in a crosswalk or waiting to cross, it lights up with a little ‘human walking’ picture. It makes it very clear to everyone it sees you.
Didn't know about this. Only Apple would think of taking live video of your eyes and projecting them on the exterior of the visor... and call it Eye Sight...
Potential users will read "external display" and immediately think of many uses, and I'd bet the last of them would be "stream my eyes". It's perfectly bland and corporate, I suppose Apple is aiming for normalization, "VR googles you could wear at a party without feeling dorky!".
The iFixIt teardown was very interesting, nevertheless.
I get what you are saying. However where I live in Florida, drivers are so pedestrian blind and oblivious that they’ll run you down right in the middle of the crosswalk while playing on their phone. Friend of a friend was just killed as a pedestrian a week ago. Bring on the robodrivers!
Having just rewatched the original Total Recall, I can only think of one perfect answer for this. Need to add a Johnny Cab guy to the driver seat and have his head look the direction he is going to go.
> use more subtle and inadvertent signals like the slight movement to the line telling me they're likely about to switch lanes
For what it's worth, this was explicitly taught in Northern European driving classes, especially in the context of turns. It would literally translate as "grouping" or "making a formation". You "group right" to make a right turn (think aligning text left/center/right). You can look at a line of cars and immediately know which ones are preparing to turn, even from an angle where turn signals aren't noticeable.
Good point about intent. As a passenger I appreciated how smooth it was but I can see how the lack of “body language” could be disorienting for human drivers.
The Waymo driving style is almost passive-aggressive. It has superhuman situational awareness and knowledge. I wonder if it isn’t using blinkers effectively but it’s everyone else that’s technically wrong.
It’s also not perfect so who knows. There’s probably still bugs to work out.
That sounds great for Waymo: I'd love to experience a ride in one and I'll make sure to catch a ride next time I am in CA.
I wonder what it does in common situations, like stop-and-go traffic with two streets joining at an angle, where one has to yield? In that case, if everyone simply followed the rules, the cars in the yielding street would never get through. While I am passive-aggressive myself as a driver, I religiously follow alternating one-by-one from either street "rule", even if I am in a street that has right of way, to ensure traffic moves for both of those and reduce chances of someone overly aggressively moving in from the yielding street.
If it is still passive aggressive in that case, it will work ok while there are other passive human drivers in the road, but would quickly fall apart when everyone drives the same.
I was talking about a case where there are cars behind me. And it's "unnecessarily" only legally, but most drivers understand this as a driving etiquette in congested traffic.
Picture two streets joining together, both with long lines of vehicles in them. One of them has right of way, the other one doesn't. When everyone drives "passively aggressively" (like Waymo is being described to), an opening never shows up, and "yielding" cars wait until the rush hour clears up, causing people in that street to go bonkers.
This happens quite frequently in many old European cities I've been to in the rush hour or when anything unusual happens (because there are no alternative routes and streets are tight: like a crash or broken-down car ahead might cause this with lots of traffic). Not sure how much it applies to Waymo and SFO, but I can imagine it certainly happens in some areas.
Ah I see, like merging on to a freeway. In Washington the law is actually to zipper merge (take turns) so I would expect Waymo to do that.
It’s a good question and I’m not sure what Waymo would do there. Based on my limited experience I expect it would navigate that situation correctly. The bias is to go (safely), not wait for perfect conditions.
Thanks for the term as well ("zipper merge") — that's what I was referring to.
Yeah, I really wonder how it'd fare in that case while differentiating between "there are just 3 cars there, no need to yield" and "it's an infinite inflow of traffic, better yield to one car"?
In England we call them "Indicators" as opposed to turn signals. My driving instructor said, they are call that because they are only indicating they might go that way. Never assume they will. It was some defensive driving advice.
I can understand that it's annoying that inconsistent use of turn signals is an annoyance, it isn't as bad a people that don't use them at all, but my understand is that the American road system doesn't have roundabouts like we do, which kind of require good indicator use for smooth driving.
I've driven in England and New Zealand, and they really had them everywhere.
Yes, we have them in the US, but we don't have enough of them that you can expect that all drivers know what to do with them. It's entirely too common for approaching drivers to be unaware that they must yield to vehicles already in the circle.
The whole thing with self-driving cars is that it feels like the story of NASA's space pen. All this money being poured into a technology that would not be needed if the richest country in the world had the common sense of simply getting rid of its car dependency and developed proper public transit infrastructure.
The NASA space pen story as you call it is a nice quip. It's also a complete lie. The Soviets using pencils wasn't a clever scrappy solution, it was an absolute nightmare for the Soviets. Pencil cores are made of graphite, when you write with it, small bits of graphite break off. On land it just falls down, but in space it floats around and attaches to things. Graphite conducts electricity and it caused countless shorts in electronics, as well as fires, all while in space. If the Soviets had space pens, they would've switched to those in a heartbeat.
I live in Switzerland which has a reputation for having good public transport. Still own a car because in many cases you can't beat convenience. Can't wait for Waymo to come here.
Self driving vehicles, with current technology, are unrealistic outside major cities. Major cities are the places where you least need a personal car.
Google Maps still thinks the speed limit on a nearby highway is 35 mph. We spent several months without lane markers on the highway. Any vehicle driving around here based on the current self-driving architecture, requiring central mapping/planning and good infrastructure, would be ridiculously unsafe.
Not to mention how lost they would be the moment they turned off the highway.
> Self driving vehicles, with current technology, are unrealistic outside major cities
Why? My Subaru can practically drive itself on most Wyoming roads, including in whiteout conditions. It’s easier to drive with fewer cars around. It’s just not as profitable to field a fleet in suburban and rural America.
My Waymo ride went through a construction zone without trouble. I don’t know how much mapping it needs or how they keep it current but it’s not constrained to ideal conditions.
As an example of an urban legend/facile pop science take being treated as fact [1]?
> if the richest country in the world had the common sense of simply getting rid of its car dependency and developed proper public transit infrastructure
Is there a single developed economy that doesn't make significant use of trucks and cars?
Depends on your population density. Tokyo does pretty well with purely public transport. There’s still cars of course, but you don’t use them for travel into the city proper if you don’t hate yourself.
> Tokyo does pretty well with purely public transport. There’s still cars of course, but you don’t use them for travel into the city proper if you don’t hate yourself
Japan has .67 vehicles per person, almost 80% America's .85 [1] and above many European countries. They travel about a fifth as many miles per capita as America [2]. But again, four fifths and a fifth. Nowhere close to zero.
There isn't a developed economy that wouldn't see a significant quality-of-life improvement from self-driving cars.
Hence why I’m talking about Tokyo. Not Japan. Living in the countryside basically requires a car by definition. People also often have a car, but don’t use it that much. Maybe you’ll get more interesting numbers if you compare miles traveled per person?
> Is there a single developed economy that doesn't make significant use of trucks and cars?
You do not even need to restrict it to developed economies. Every city (even not very large ones) I have been to around the world, developed or not, has streets lots of cars: London, Manchester, Paris, Nairobi, Colombo, Madras, Bangalore, Sydney, Singapore..... and smaller towns and rural areas in those countries too. Those cover a pretty wide range of quality of public transport too.
Way to miss the point. Yes, it is known that the "NASA space pen" was not really funded by NASA, but but its development still took millions of dollars in R&D.
> Is there a single developed economy that doesn't make significant use of trucks and cars?
For the 127th time: car dependency is not the same as car usage.
> its development still took millions of dollars in R&D
$1mm at the time, $10mm today. They sold for the equivalent of $30 each in today's money and saved manned spaceflight from having to deal with graphite shards in the air and electrical.
> car dependency is not the same as car usage
"If the richest country in the world" got "rid of its car dependency," we'd still want safer cars.
With the technology controlled by a handful of corporations? At an development cost that will force them to create any and every opportunity to seek rent on your mobility?
Many cities in Europe still have their Roman era street plans. The street layout of a city is incredible sticky. Once you have a city built around cars and all the private property and infrastructure that goes with it, you are basically stuck with it for 1000 years. And American city layouts are horrible for public transit and walking.
We can look at international cities with longing and jealousy, but LA, Houston, Miami, and so on will never have useable public transit. You’d have to bulldoze the whole city and start again. So great for Paris and London and yes I would rather live there, but we are stuck finding solutions that work inside the mess we made.
I agree with you in part - we do have 100 years of development patterns baked into concrete and steel, but they were not originally built just for cars.
LA was built around interurbans, the idea that I should drive my car all the way from San Bernardino (or frankly even from Pasadena) to Los Angeles would be foreign to the original planners and builders in Southern California. Consider that every major freeway in LA is basically in whole or in part paralleling a former PE line, and suddenly my assertion doesnt seem so odd.
LA is not the only place where the sprawled layout was created by interurbans, Arguably to a greater or lesser extent - Detroit, Dallas, Cleveland, Cincinatti, and there are others whose early development was deeply influenced by the interurban exist.
Wait, what? Paris streets are a mess, even after Haussmann. That does not prevent outstanding transit. Even before the latest obsessive aggressive agenda-driven nonsense. The two things that Paris has are (1) inhabitant and workplace density and (2) the will to pursue public transit decade after decade.
After that, there are lots of streets with a single lane of traffic - and buses go through them just fine. And there are lots of too narrow streets but they are isolated here and there - in between streets that are okay for buses. Equally there are lots of very wide avenues for several lanes of traffic and that's fine for pedestrians and transit even though it makes subway stations sometimes very wide. And there are lots of randomly angled streets. There are steep hills. There are rivers and canals and giant mining voids which make building subway tunnels interesting. There are lots of train track rights of way which sometimes block pedestrian traffic but are covered and crossed reasonably often (sometimes still not optimally so). There is snow, ice, floods, hot weather. You name it it's there.
Both of these are essential. "Decade after decade" is essential. Both density and building that infrastructure are impossible to build at the stroke of a pen. So an advantage of Paris compared to San Francisco is that it had decent inhabitant and workplace density from the infancy of public transit when it was private animal pulled carriages.
So what's missing in US cities: decade after decade of density followed by decade after decade of public intent. SF could do it - but for sure you wouldn't see the result tomorrow - but NIMBY so no, SF can't do it. Not before getting rid of NIMBY. LA is much more willing to build anything anywhere so it would have much more of a chance. But it's huge and people often live far from their place of work. Smaller areas with LA would have a better chance. Perhaps Glendale plus the entertainment industry north-south corridor? That might be manageable.
Yet still more "even then", ride Paris transit at rush hours in summer and you won't feel that it's all that good: it's sweaty, packed, and there are lots of timetable incidents. You will regret having chosen rush hour. You car would have been nicely air conditionned and comfortable (but depending on your trajectory would have been stuck in traffic for a while.)
For trips shorter than 2 miles, I can bet you that it's faster for me to hop on my bike (or to find an electric scooter) and get to my destination than it is to drive. And let's not even talk about the time to find parking.
The grocery store is 1.3 miles from here but it’s a 300’ climb and it’s 42f and raining right now. Even with my e-bike that’s not pleasant. On a conventional bike that’s unrealistic for all but the most hardcore cyclists.
There’s a bus that serves that route but the stop is uncovered and service is unreliable because it’s immediately after a rest stop where the human drivers take care of their biological needs.
A robo-bus would be more predictable and enable more frequent service.
And it doesn't even occur to you that the problem is the poor infrastructure? Do you honestly think that the optimal solution here is just to keep waiting for some trillion dollar corporation to be able to sell some new product that will bring you more "convenience" (at the cost of dependency and yet-more environmental costs related to cars and its required parking infrastructure) when you instead could be doing something like, I don't know, getting your city council to improve the fucking bus service?
> And it doesn't even occur to you that the problem is the poor infrastructure?
It has occurred to me.
> Do you honestly think that the optimal solution here is just to keep waiting for some trillion dollar corporation to be able to sell some new product that will bring you more "convenience"
Not sure where you got this idea. I’m not waiting for anything. I’m still excited for self driving vehicles to compliment the existing system.
> at the cost of dependency and yet-more environmental costs related to cars and its required parking infrastructure
Why would a self driving car need to park? The ability to always be in use is one of the advantages.
As I said there’s no requirement that the self driving vehicle be a car form factor. A self-driving bus could improve access and reliability.
> when you instead could be doing something like, I don't know, getting your city council to improve the fucking bus service?
How?
I choose to live in a city with good (for America) transit. I vote for every transit measure available to me. I ride the bus.
I commute by bike when I can. I don’t own a car.
> Why would a self driving car need to park? The ability to always be in use is one of the advantages.
They are not going to be always in use. Unless you got a perfectly uniform distribution of demand over time, eventually most of these cars will have to be parked somewhere.
> They don’t need the same parking infrastructure we have now.
The existing parking infrastructure is not going away because of self-driving cars. The exact same argument could've been applied to the car sharing programs that already exist. People that own cars are not going to leave their cars for a robo-taxi, in the same way that they are not getting rid of their cars in favor of a ZipCar membership.
> Buses also need a place to park and get maintenance
Do you know that picture showing people standing on the street to show the density of the different methods of transportation? What do you think needs a larger parking footprint? I'd guess that even if autonomous cars got to a remarkable 10% idle rate, they would still require more space than buses.
> Self-driving allows locating the parking in more distant locations.
Guess what companies will do when they realize that their customers do not want to wait 10 minutes for their ride to come pick them up at home? That's right, they will get lots of parking spots close to the urban/suburban city centers.
> The existing parking infrastructure is not going away because of self-driving cars. The exact same argument could've been applied to the car sharing programs that already exist.
Why would the existing infrastructure not continue to change? We are redeveloping parking lots today. Car and ride shares did reduce parking needs.
> People that own cars are not going to leave their cars for a robo-taxi, in the same way that they are not getting rid of their cars in favor of a ZipCar membership.
Building codes for parking spaces are constantly changing here and consider transit availability and car ownership. If self-driving vehicles (including expanded public transit) are available that will further reduce the requirements.
Self-driving vehicles (even cars) put downward pressure on parking needs.
> Do you know that picture showing people standing on the street to show the density of the different methods of transportation? What do you think needs a larger parking footprint? I'd guess that even if autonomous cars got to a remarkable 10% idle rate, they would still require more space than buses.
Why are you so obsessed with the car form factor? I don’t see how a robotaxi needs any more parking than a taxi. A self-driving bus needs as much space as a conventional bus today.
> Guess what companies will do when they realize that their customers do not want to wait 10 minutes for their ride to come pick them up at home? That's right, they will get lots of parking spots close to the urban/suburban city centers.
That’s not how busses, taxis or Ubers work. Ubers are not dispatched from a hub on every call. Demand is predicted and capacity is dispatched accordingly. As demand goes up dispatched capacity utilization goes up. More vehicles can be dispatched when demand approaches dispatched capacity.
This problem is the same no matter what form the driver takes. Taxis, ride shares, and busses already do capacity planning. Robot drivers just remove the human driver complexity from the equation.
Robotaxis (like busses and taxis) can use denser parking structures in their depots because they don’t have to deal with the limitations of the public, or even humans.
I routinely walk considerably further than that for groceries, in less time. On a route including a steep hill.
Short trips made any way but walking have considerable overhead. Bad urban planning makes it worse. How many times do you have to stop a car on that route?
> I routinely walk considerably further than that for groceries, in less time. On a route including a steep hill.
Not everyone has or wants to spend that kind of time. Especially when it’s not necessary and the weather is bad.
I’m hosting a party this evening so I need to run up there and get snacks and beer. The hour of time the e-bike saves lets me do that between work and guests arriving. I don’t have time to walk that distance and still eat dinner.
> Short trips made any way but walking have considerable overhead. Bad urban planning makes it worse. How many times do you have to stop a car on that route?
I don’t own a car. I ride my e-bike. There’s two lights and a stop sign on that route. It has a bike lane which is nice. Takes 10 minutes including time to put on my helmet and gloves and lock/unlock the bike.
If we both left my couch at the same time I’d be home on my bike before you even arrived at the store on foot.
That would be a more reasonable trajectory for US cities: more frequent, faster main line transit plus a flood of local inexpensive automated electric vehicles. That would be compatible with the current low density.
Yes, the last mile is walked. That’s the problem. It’s not practical to put a train station or bus stop at every corner. This is why cars exist everywhere. If those cars can be automated they can be shared even more efficiently than Ubers and allow more efficient bus and train routes.
> It’s not practical to put a train station or bus stop at every corner.
Tell that to Dublin Bus. Some routes are more or less unusable due to ridiculously high numbers of stops; there’s one in particular near me which has 37 stops over about 4km.
There are large parts of the suburbs here where you’re never more than a couple hundred meters from a bus stop. The last mile is, for much of the city, just not a problem, though when taken to extremes it does make the routes very slow.
> Tell that to Dublin Bus. Some routes are more or less unusable due to ridiculously high numbers of stops; there’s one in particular near me which has 37 stops over about 4km.
Sounds impractical.
> There are large parts of the suburbs here where you’re never more than a couple hundred meters from a bus stop. The last mile is, for much of the city, just not a problem, though when taken to extremes it does make the routes very slow.
> if the richest country in the world had the common sense of simply
- There is nothing simple about it.
- There is nothing inevitable to this appearance of "not enough money". For example very little of self-driving cars development has been on public funds. For example, there is plenty of money in California govt. And plenty of money in US federal govt.
- One doesn't prevent the other. You can't seriously argue that better transit would remove all needs for automated driving.
With self-driving tech, buses could be made a lot smaller and more frequent. They are large and infrequent because of fixed per-vehicle costs (mostly the driver).
EDIT: Small size then enables personalized routing: a multi-passenger taxi taking a set of people to similar destinations.
That makes a lot of sense, I never thought of the labor as the limiting factor on bus route (clearly I just never thought very hard about it) but boy would it be nice if there was a 9 passenger van every 5 minutes to replace a full size bus every 30.
Really the lack of stops and parking places is, and would be even worse for self driving vehicles deployed in numbers. Unless it's self driving motor bikes or such small personal transportation.
How much are you willing to bet that the price of a Waymo ride will converge to be within 10% of the price of an Uber ride today?
(Also, followup question: how are the prices of Uber rides compared with regular taxi service, now that Uber has stopped subsidizing most of their trips?)
> regardless of situation, the cost of labor has nothing to do with it
Sort of. The cost of labor is the cost floor of the competition. So long as Waymo et al are smaller than the competition, i.e. so long as they don’t need to compete on price, they should tack under their direction. They aren’t setting the market price.
The moment they need to compete, they can tank the price or increase service quality. (They could also stay put and milk it. But that invites new-entrant competition.)
Internet companies compete on price out the gate because we have low marginal costs. A company like Waymo has hardware costs that make competing on price out the gate prohibitive. So in a sense, their costs don’t have to do with labour. But that’s sort of true for any automation tech.
> They could also stay put and milk it. But that invites new-entrant competition.
And believing that this is likely to happen is the problem I have with the whole proposition. I don't see any real player with capital, data and tech know-how to get into this in a way to turn it into a proper competition.
I'm saying that we shouldn't be so enthusiastic about the idea of giving more power to one of the largest corporations in the world, and that the idea is only worth pursuing if we have better ways to put them under check. To call this "advocating for blowing everything up" is just weasel wording and really disgusting rhetoric.
And just like the NASA space pen story, the private sector is developing them on their own at their own risk, with maybe some indirect grants or subsidies, but not a direct contract specifically for the government to be a beneficiary.
Way more money is being wasted on the grifts surrounding developing public transit infrastructure. LA County for example is 88 different municipalities and it will never get resolved.
No amount of public transport can accomodate the personal whims and demands of anyone let alone everyone. Even Japan, famous for its public transport infrastructure, still has a healthy population of drivers both in metro and rural areas.
I am not saying to get rid of cars altogether. I am saying that it would be better (and cheaper) for everyone if we took all those resources being out into "autonomous vehicles" and developed public transit first. Reduce the amount of yearly trips that are done by car. Provide alternatives.
The important thing about Japan or Europe is not "they don't have cars", but "people make 3-4x less trips by car compared with the alternative modes of transportation". If you want to have safer roads (in a way that doesn't give even more power to tech companies) the best way to do it is by simply reducing the amount of trips taken.
>That is at best 4 miles of New York subway [2]. It's a third of a train between Bakersfield and Merced [3].
Bakersfield to Merced is 171 miles of intercity high-speed rail. It's unclear to me why Californians need to make such trips with any significant frequency, and it's certainly not what I'd normally think of as "public transit".
22 years ago in Toronto, the Sheppard line (3.4 mi, underground) was built for less than $1b CAD (https://en.wikipedia.org/wiki/Line_4_Sheppard). From what I can tell, costs in NYC are mainly a NYC problem.
Chicago is extending an existing above ground train line along an existing freeway (no right of way issues, not a subway, etc) 5.6 miles. It’s four stops and a new rail yard. 3.6B.
Something may have changed more recently. The first stage of the Eglinton Crosstown LRT line in Toronto looks like it will come in at around (converting currency and distances) 9B for 12 miles. Locals generally don't seem to think this represents a success. (Despite being light rail, about 2/3 of this first stage will run underground.)
>Because they’re both between Los Angeles and San Francisco.
This explains exactly nothing to me. There are many cities between Toronto and Montreal (the two largest CMAs in Canada, and about 88% of that driving distance) and I can't imagine why anyone would regularly travel between those two, either. Nor can I imagine why someone would make regular trips between Belleville and Ottawa (very close to the Bakersfield-Merced distance, and two of the larger cities in between).
>If not trains between cities what do you consider public transit?
Subways, light rail, streetcars and buses, all operating within a city. And, yes, trains operating within a CMA. (GO Transit trains from Toronto can get you as far as Kitchener, but as far as I know it's a vastly less popular route than the main Lakeshore lines. From Hamilton to Oshawa is definitely not solid built-up area, but it's pretty dense.)
I can't understand why this would be non-obvious. Maybe there's a cultural difference. Are Americans really so dedicated to urban sprawl that residents of a metropolis with an 7- or even 8-digit population might still require regular intercity travel to go about their lives?
> can't understand why this would be non-obvious. Maybe there's a cultural difference
Yes. Many Americans take it for granted that we have quick access to a variety of cities and landscapes. (Particularly on the coasts.)
Keep in mind that America is about a third richer, on a GDP per capita basis, than Canada. Even with stark income inequality, that produces a large disposable-income gap. Add to that the population gap and you have about the population of Canada travelling intercity many times a year.
Try using a unit per person. It starts to look much less rosy for Waymo and also do not forget roads are still public infrastructure that has to be paid for and maintained.
What, you think Waymo will be able to weasel out of road tax or drive without roads?
>OP argued we should devote self-driving resources to public transit.
Yes, because it would allow moving people more cost-effectively (among other things) - which is measured per person per unit distance traveled. (I'm not OP, but nothing in this discussion is even remotely new for transit advocates like myself.)
> because it would allow moving people more cost-effectively (among other things)
Add last-mile and model for the real American population distribution. Not an imagined America where everyone lives on Manhattan.
Not all miles travelled are equal. In value, urgency or desirability and thus price willing to be paid. Complaining about self driving cars is asinine. The scales are wrong. The problems are wrong. The places are wrong.
> nothing in this discussion is even remotely new for transit advocates like myself
I mean, exactly. We’re urbanising without commensurate increases in public transit. Deployment or use. Perhaps that hints at a change being in order.
Self-driving cars are public transit depending on the terms of use; but I assume we're really talking about mass transit here.
When I was in Germany 25 years ago, a district of about 25k people (less than 1k per square mile) had multiple bus routes (4 IIRC) connecting the main town to other settlements, running through farmland. It was not unduly expensive and the locals certainly seemed to think it made sense. Children would take it to school in lieu of a dedicated school-owned service.
You are starting to seem dishonest when you resort to outliers and present them as the baseline.
> To the extent we can compare them on capital efficiency
We can not. This logic is stupid. This logic will take you to the idea that the best thing to do is to get everyone a helicopter because helipads are cheap to build and it costs zero dollars to "build roads in the sky".
Yeah, you are right. These things are not competing. Yes, you are also right that proper infrastructure (re)development in North America will take a lot more money than what Big Tech has invested in self-driving cars.
The upsetting part is to see even supposedly smart people buying into the idea that American Exceptionalism is a real thing, and that you can keep holding on to the hope to find a shortcut away of your problems. Self-driving cars or not, the US is still going to be an expensive, inefficient country that can boast about its amazing economy, yet most people living there are at third-world country levels of development.
> Self-driving cars or not, the US is still going to be an expensive, inefficient country that can boast about its amazing economy, yet most people living there are at third-world country levels of development
For fuck’s sake, the world’s largest rail system and navigable waterways say hello.
I’ve advocated for public transit. But it’s turning into zealotry when an $11bn project showing actual gains is turned into a soapbox for decrying a pet project.
If Waymo/Google/Alphabet really cares about people's safety and this is nothing but a pet project where the amount raised ($11bn) in 15 years is ~3% of their whole revenue in 2023, they wouldn't mind open sourcing their whole system, including for other commercial ventures.
Until then, the whole thing is nothing but a Trojan horse to let encroach themselves even more into another aspect of our lives.
> they wouldn't mind open sourcing their whole system, including for other commercial ventures
You seem to just not like that this is done by Google. (Or by any private entity.) That is fine. I, like, hate this one restaurant in New York. But be honest about your motivations and check your biases.
This isn’t self-driving cars vs public transit debate because there isn’t one. If Google were funding a leg of passenger rail I suspect you’d be similarly incensed.
> world's largest freight rail system
The world’s largest rail system, period. It’s also the largest freight rail system. But passenger and freight are types of rail systems.
I despise this way of arguing. Laying down x length of rail track is not a difficult problem whatsoever. Such a crude metric is entirely insufficient to back up an argument about how great American railways supposedly are.
What's the ticket cost for the passengers compared to other countries? Average train delay? Cancellation rate? Speed? Death rate? The argument is not looking so hot anymore, right? Then stop it with the irrelevant statistics of rail track length.
>Laying down x length of rail track is not a difficult problem whatsoever.
It actually is. Japan's new maglev shinkansen line is stalling because they can't get land rights to lay rail through.[1]
An even more blasphemous example is California's high speed rail project which never accomplishes anything, but I'm pretty sure that also suffers from a severe case of legalized money laundering.
This is not a matter of who is doing it, but (a) why and (b) second-order effects.
> If Google were funding a leg of passenger rail I suspect you’d be similarly incensed.
It depends on (a) why and (b) second-order effects.
Are they doing it, e.g, because they want to build a new campus in a lower-cost-of-living area, and they want to make the idea of living in Tuscaloosa, AL more palatable by having it connected to Atlanta, GA with train service that does not take 6(!!!!) hours as Amtrak currently does? Amazing, go Google!
Oh, they want to make it so that everyone can use it at reasonable prices, but Google employees can do it for free and get priority boarding? Fine, if that's what it takes to get private enterprises investing in infrastructure, I'm okay with it.
Oh, they want to do it because they are going to use it as a test-bed for some revolutionary transportation technology that they pinky-promise will eventually work as some Futurama-style tube network where anyone can go door-to-door as fast as possible? And they are offering the whole thing for free (or heavily subsidized) for everyone that enrolls in their beta program? Then please Google go fuck itself, because we all know how this is just bait.
> if the richest country in the world had the common sense of simply getting rid of its car dependency and developed proper public transit infrastructure.
Then have the dignitaries and environmental policymakers give their uparmored gas guzzling cars up first, leading by example.
What are you talking about? Many politicians catch public transport around the world. I remember a few years back a mini nothingburger controversy about a politician 'caught' drinking a canned cocktail on the train in London https://www.theguardian.com/politics/2019/apr/20/diane-abbot...
>technology that would not be needed if... developed proper public transit infrastructure
NYC and London have had 'proper' public transit for over a century. It's handy but not that great. I was on a 52 year old Bakerloo line tube train the other day and they are much like modern tube trains, if grubbier. It's not going to suddenly turn wonderful and solve everything. (typed on a London bus)
Also while I'm not sure any technology is really needed, as in we could get by without it, there are about 1 million road deaths a year globally. A 90% reduction would save 900,000 lives which is a nice to have. And more meaningful than a space pen.
Public transport cannot do everything cars can do, especially outside big cities. Better public transport can reduce the need for car ownership - I did not own a car for much of my life for that reason. However, it does not eliminate it.
> Sure cars are helpful for transporting stuff. But most car rides is one dude sitting alone in a 1.5 ton machine moving his lazy ass around town.
I do not know about lazy. I find it a lot easier to use public transport myself. I would like to see numbers for "most" especially outside the US. Very few of my car trips are by myself without something to carry.
> Public transport, if done correctly, can replace that, also outside of big cities. There are good examples across Europe for this.
Where has public transport eliminated cars? Not anywhere I have been in Europe. You can significantly reduce car usage, but not eliminate it, so any safety improvements are still very important.
Another problem outside big cities is that some buses and trains, especially at quieter times, also run pretty empty. There were about four or five people in the bus my daughter got into this morning (on a weekday morning!) and its close to the end of a journey between two reasonably large towns. The next bus gets very full though.
One of the features of public transit--certainly the commuter rail relatively near my house (although I still need to drive to the station)--is that it's often very empty in off-hours but if you don't run (an insufficient number of) those near-empty trains, a lot of people are going to say screw it and drive. I won't take commuter rail into the city for an evening event in part because because I really need to time the return journey--and it takes quite a bit longer.
> Where has public transport eliminated cars? Not anywhere I have been in Europe. You can significantly reduce car usage, but not eliminate it, so any safety improvements are still very important.
Sure, not eliminated. There will always be cars. But reduced, as you said. I guess my main point was that one should not falsely believe that the existence of driverless buses and trains will somehow magically and significantly improve the public transportation situation in a given country.
Right now I have to walk 12 minutes to a bus stop because the one literally in front of my house was closed due to a lack of drivers. Then I wait 3-10 minutes for a bus that takes 20 minutes to get to the bus hub in my neighborhood. It takes 8 minutes to drive there. If my neighborhood bus route was replaced or supplemented with self driving shuttle busses feeding that hub then a 1 hour door-to-door trip could be 30 minutes.
Sure. I'm not saying that there no use cases. But I don't see the US turning into a public transportation utopia just because there are self-driving buses available. Because the reason public transportation sucks in countries where it sucks is not because drivers are expensive.
But it does suck because drivers are expensive. Seattle is struggling to find drivers. It's a constant battle to identify the most used routes and keep them staffed but that's why there's not a critical mass of riders to pay enough fares to hire drivers.
Doing the rare things that actually require a car, does not entail a need to own one. (And it's perfectly possible to have a fulfilling life without those rare things.)
Self-driving cars help get us past the path dependence and move toward better public transit.
I tried using the train for my commute. However, I had to maintain a car at each end and I could only use the 7AM train as the parking at my origin would be filled at any later time.
Self-driving cars break my dependence on a car at destination and allow me to pick an origin train independent of parking.
You're conflating two tangential things, automatic driving is automatic driving whether it's a car or a bus.
The biggest issue that automatic driving needs to overcome is that it's sharing the road with manually driven cars. We've already had the technology for a long time to have perfect automatic driving if the environment was fully automatic; computers are unparalleled at accurately sharing and processing data with each other.
This isn't to say that the solution is to get rid of human drivers, because driving a car has been one of the most empowering paradigm shifts for the commons. Being able to travel yourself timely anywhere anytime for any reason is a level of power that pre-automobile commons simply did not have. Subjugating your power to travel to a computer is surrendering that incredible power.
If (and likely when) we can figure out how to better share human data with computers and vice versa computer data with humans, everyone on the road will be better off.
GoA 4 (no trained personnel onboard) is only practical for entirely grade separated (ie it's almost impossible for anything else to be there, no cars, no bicycles, no pedestrians) routes.
Even GoA 3 is fraught without this constraint, and for GoA 2 you're still paying for a driver because although the machine does most of the work the human has to handle the inevitable deviations from the model.
GoA 4 railways exist, if you're putting a metro in tunnels or elevating it obviously this is grade separated, and in principle you can do it for long distance rail if you're willing to eliminate at grade crossings (expensive). But they are nothing like a city bus route for example.
>driving a car has been one of the most empowering paradigm shifts for the commons. Being able to travel yourself timely anywhere anytime for any reason
Everywhere that the roads go, sure.
The more other commoners are trying to do this, and the more useful places that the roads will allow it, the slower it gets. But not only that, it's very unevenly paced - and therefore inefficient and less safe.
The problems are compounded by North American road and intersection design. Many urban areas in Europe manage to get by with far fewer streetlights and stop signs (often by using roundabouts). But cars in North American cities routinely get to speeds that would be dangerous in a collision (which would be much less of a problem if this could be confined to routes where they're isolated from pedestrians - but this often doesn't happen), only to have to stop and wait for a minute or more at a time, an embarrassingly short distance thereafter. And then buses have to deal with the car traffic and the lights, make additional stops, and take extra safety precautions due to their size.
Everyone thinks of taking the bus as slow. If there were no cars, the roads could be designed for buses, and streets for pedestrians. But we don't even properly distinguish streets from roads around here.
There's a future, not in the short-term, where 99% of folks don't need to or want to own a car. You hail a ride, and a vehicle shows up, perhaps with 3 of people in the vehicle in their own bubble internally designed into the car. You get in, it drops offs others in the vehicle, and then finally drops you off. The vehicle exhausts its entire electric charge, recharges, and then gets back on the road. The number of vehicles on the road is down 50%.
Same for trains. I wouldn’t take the LIRR to JFK or Caltrain to SFO if I had to bike with my luggage or whatever. And I’m not adding an hour to my trip for the bus.
You'll be able to pay more to take a single dedicated vehicle the whole way, or less that includes a self-driving bus or two for most of the middle segment of your journey.
But the bus "routes" will be totally dynamic based on demand, and will only stop occasionally, and you won't have to wait long.
It's going to be a complete revolution in public transit.
I doubt that the reason public transport is not as prevalent e.g. in the US as in some European countries is the cost of the driver. The fixed costs of operating a train or a bus must be many many times more.
Ask the 2.3 million employed drivers in the United States how the future looks.
Because pretty soon, they'll all need to be doing something else. And I don't see a technology that will open up 2.3 million jobs for them to move into, or training initiatives to support them while they shift professions.
The future _could_ be bright, if we approach AI right, and build the social safety net to soften the massive amount of transition that will be necessary. But we're not. We're looking at cuts everywhere in the near future. And long term, it seems we have no strategy. Which makes it more likely that a lot of people will be harmed.
But you might decide to build in enough additional taxes on the self-driving vehicles to cover at least the state's increased costs in terms of unemployment, medical, and services.
This way all those people riding the fancy new cars are the ones subsidizing the new industry's hidden costs instead of every tax payer.
> But you might decide to build in enough additional taxes on the self-driving vehicles to cover at least the state's increased costs in terms of unemployment, medical, and services.
A better idea would be to tax capital returns more rather than selectively adopting taxes individually on newer, more efficient but also capital (vs. labor) intensive industries.
Make one change and solve the problem globally rather than creating a new legislative fight over each automating industry.
> This way all those people riding the fancy new cars are the ones subsidizing the new industry's hidden costs instead of every tax payer.
Reduced labor costs aren't a hidden cost (they are the opposite of a cost, in fact.)
If there is a hidden cost it is capitalist society hiding basic social support costs in payrolls, artificially inflating the cost of labor-intensive goods and services and subsidizing automation.
What is the normal annual turnover in those 2.3 million drivers? 20% perhaps? I doubt self driving cars are going to replace 20% a year - they'll take many years to roll out. So it'll just be less people entering the field rather than mass layoffs.
This presupposes that the move won't be forced. It'll show up as competition in the gig economy. And there will be slower adoption in the long-haul or commercial receivables market since someone will need to be responsible for the cargos.
In the end it will be forced, even if it's not a layoff. Fewer gig rides. Cheaper competition for hauling contracts as the driverless fleets grow and customers adapt. Forcing people to do more for less, or leave.
But where do those people go? The slow decline that forces people out by starving them out does not suit the situation well for retraining in a different career path.
The truckers will join all the tech support and customer service people queued up to interview for a smaller allotment of jobs. Then the programmers. Eventually even the business people who take up space following random business workflows that don't require an ounce of creativity will follow.
When automation starts to be really good, it will be the restaurant cooks. Servers will still be around, for dine-in restaurants, and personal services like massage therapists. But how long will it take for competition in the labor market to drive those wages down? Demand certainly won't be increasing since the pool of potential customers with disposable income will decrease.
The shrinking of the middle class that we've seen up until now is nothing compared to what we'll see in the next decade or two. This should not be handwaved away with an "Oh, those people? They'll just go do something else."
If we need to turn a 15-year long haul trucker into a solar power technician, what's the time frame and cost to do retrain, how does that trucker afford and avoid starvation while he does so?
Hopefully whatever minimally viable level of support we decide on as a society will be higher than "living in one of the growing tent cities while eating at the local soup kitchen". We're already struggling with those in the current labor market, which is a sign that the existing support system is insufficient. We're totally unprepared for the future labor market.
Technology will open up new jobs.
But I don't expect employed drivers to suddenly become data scientists.
Its a fallacy we constantly use as an argument when we talk about technology taking people's jobs.
Low specialty professions have been in my view a great way to have people employed, people who may be less fortunate, unspecialized or even students. Now that such professions start getting absorbed by tech, I don't know how those people are supposed to get by.
I can see how somebody would argue 6hay this has occurred before, but for example replacing horses with cars just made postpone jump to learning how to drive. They didn't have to "specialize for 4 years" and become computer scientists. Now this is a bolder issue imo. But my perspective can be argued
At some point my country started borrowing lots of money and spending it on social programs, which eventually led to its bankruptcy. Some people still think that the guy in charge back then was a great leader.
This is the type of comment that makes you look either like a privileged elite that is part of Big Tech, or just a clueless moron whose day at the barrel is yet to come.
Do you really need a list of ways how Big Tech has harmed society? If things like "Surveillance Capitalism" is not enough, maybe:
- the collapse of quality journalism, because SEO and content farms made impossible to maintain a financially sustainable team of independent reporters?
- The absolute destruction of people's ability to read a piece of text for longer than 15 seconds?
- The mental health crisis in teens and even young adults caused by social media?
- Planned obsolescence and the removal of our right to repair things on our own?
- The normalization of the "gig economy" not just for young kids who are looking for some extra cash, but as the main source of income for grown up adults, with actual families to support?
- (Google and Facebook mostly) adopting the absolutely anti-competitive practice of "when we get a competitor, we buy them out as early as possible or just keep offering our stuff for free until they eventually run out of money"?
- Apple outright refusing to let people install alternative app stores, because they know that without this artificial scarcity they would not be worth 1/10th of what they are today?
I'm not in SV and was trying to think of ways it had harmed me personally.
I haven't really been affected by any of the above. Maybe the app store thing a bit as I wrote an app and dealing with Apple was hard work. But then I wouldn't have been able to make an app without some company making the device for me to do it on.
Re society, I grew up pre internet and I'd say there are pros and cons. On the plus side you have access to the world's information and video and text communication for free across the planet which you didn't have before. On the minus side that's led to people spending a lot of time staring at their devices and less interacting with each other. Nothing's perfect I guess.
> I wouldn't have been able to make an app without some company making the device for me to do it on.
If you are not asking yourself why should you need to ask permission to develop an app, I'm sorry but you are lost beyond repair. Stockholm's Syndrome is real.
> If you are not asking yourself why should you need to ask permission to develop an app, I'm sorry but you are lost beyond repair
We should get out of our bubbles…and make transportation decisions…based on our preferences as developers. Mobile app developers. We’re in the SV bubble…
I just heard someone mention in a podcast how as a cyclist they need to make eye contact with drivers to be safe, and there is no one to make eye contact with.
Waymo would say "don't worry, it sees you" and maybe the numbers even show they do some percentage more than the average human driver, but that still does not reproduce the self-help aspect of you looking at a driver until you see that that driver sees you. I think that is a pretty significant thing. It needs some sort of better answer than "don't worry about it"
Maybe they will prove to be so reliable that you don't need to worry about them. You can just reliably assume and predict their behavior as animate objects about the same as you do for inanimate objects. IE, you can't make eye contact with jersey barrier, but you don't need to, you know what it will do (nothing). And you can't make eye contact with say, a motorized railroad crossing barrier, but you don't need to. It moves, but only in an absolutely predictable way. You can be a cyclist around the railroad crossing barrier no problem.
Maybe Waymo and others will be that consistent that you can treat them like "moving inanimate objects" safely.
I don't know if they are right now. They may be better than humans on avaerage, but that is a long way from predictable and safe.
Maybe They can build in some way for the car to signal back to the surrounding people what all it has taken note of, or not. Some way for a cyclist or pedestrian to look at a Waymo and see that it sees you. That should be possible.
Maybe it can go both ways, maybe we can develop some kind of standard where if you need to get a cars attention there is some gesture you can do that it will be especially watching for. So even if it doesn't see you, it will see you if you wave your hand a certain way or something. Only problem with that is you have to be able to do the thing, and some of the times you need a car to see you the most will be the same times when you are incapacitated, ie lying on the road after some accident or something.
I ride my bike several times a day, and I have very mixed views about Waymo, etc. I care immensely about traffic safety (so much I moved to the Netherlands).
If you took the world as it is now and left everything the same except that every car was a waymo and there were no human-driven cars left, I would be ecstatic. Humans are horrible drivers and some of them are downright murderous. If every car were a waymo I could even imagine letting my kids bike to school in the bay (where I used to live), which I wouldn't dream of now.
But I think the second-order effects of self driving cars could be terrible. It removes any incentive not to have an incredibly long commute (exacerbating sprawl), and so far every time there has been a situation where the needs of walkers and cyclists were pitted against the needs of drivers, drivers won. I think the same will happen with self-driving cars, and people will be made to wear beacons just to walk across the street.
But if we have proper regulation (pretty unlikely I would say) and use things like waymo to stop humans from driving (remember, drivers are the leading killers of children in the US!) that would be great.
>removes any incentive not to have an incredibly long commute
The roads only have so much capacity which is not going to change much whether the cars are self driven or not.
In London they have deliberately reduced road capacity to reduce traffic by blocking off lanes, side streets and the like. I could see a future where the normal way to get around is a self driving cab to the station at each end with a train for the main journey. I know you can do that now but the human cabs driven are kind of expensive.
Maybe something a bit like Zermatt where there is a train station and then it's pedestrianized apart from electric golf cart like taxis. It all works quite well really apart from being expensive. But property being like £1m+ is kind of a symptom of people wanting to be there.
Something like that would be nice - self-driving cabs for last-mile journeys make sense for areas that are hard to serve otherwise, though funny enough you just described my exact commute, except that I use a rental bike (ov-fiets) between the station and work.
It seems contradictory to care immensely about traffic safety but lament the idea of wearing a vest to cross the street. It seems like something that logically we should probably already be doing. Self driving cars are just making it more obvious.
>It seems contradictory to care immensely about [school] safety but lament the idea of wearing a [bulletproof] vest to [enter] the [school].
(Or be subjected to a metal detector screening, etc.)
No, it's perfectly consistent. The argument is that the underlying problem shouldn't exist to any significant extent in the first place, and doesn't need to (and doesn't in many other parts of the world).
A lot of traffic safety issues in North America derive from road design, which has historically prioritized the needs and wants of car drivers above everyone else. It is not "contradictory" to reject a solution that entails doing even more to accommodate cars.
As a cyclist I avoid eye contact like the plague: it's only a small minority of drivers, but that small minority is dangerous enough to outweigh all benefits eye contact might have with others. Some drivers will inevitably read eye contact as "the cyclist has seen me and understood that my big and dangerous car will crush them without taking harm so they will yield no matter what traffic rules might say, so I will go". Eye contact is like a green light to them.
If I do want to yield despite having priority I can do that just as safely without eye contact and if I want to assert priority it's objectively less dangerous (because of that minority) if I do it without eye contact. Nothing to gain, so much to lose. The reality on the ground is what the car does, and I will focus on that. Taking guesses from looking at the driver is counterproductive.
I don't cross in front of a car without knowing that that cars driver sees me. And sometimes you have to cross in front of a car. Most of those times everyone is just stopped at a light anyway, but not always, and some of those lights still allow the driver to proceed anyway if turning right, which they might do at any time.
I also ride a recumbent trike which is low to the ground. A diamond frame puts you up at eye level with most drivers and that makes everything easier.
Not just as a cyclist. As a pedestrian in one of those huge Asian metropolises with absolutely bonkers traffic, you should never ever make eye contact either. It’s the only way drivers will stop for you.
Really? I've seen a number of drivers in Shanghai stop for me while I'm just standing in the middle of the road, waiting for a clear spot in the traffic.
Actually, I got a fairly strong feeling that Shanghai traffic has gotten noticeably less crazy over time, but I can't rule out that I was just observing the local traffic of different regions of Shanghai at different times.
Haven’t experienced Shanghai personally but ten years ago in Xi’an and Chengdu, yes, this sometimes was the only way to cross roads. In Jakarta and Bangkok, however, this always is the only way. Compared to those cities, however, the traffic in Chinese cities I’ve seen was always very organized.
I already responded but really I got tricked into missing the actual point and arguing about an implimentation detail.
So you don't make eye contact. That's nice. But the issue isn't actually about eye contact. It's about the human having less control over their own well being.
Whether you personally ever look at a driver, it doesn't change the fact that it's a thing that a person can normally do to determine if it is a good idea to proceed or not. It is never wrong to be aware of your surroundings and look ahead into whatever you are about to do next.
If you look at a driver and that driver takes that as some sort of challenge like a monkey, well A: I've never seen that but whatever, anything is possible. B: It doesn't change anything. You still attained the goal of determining what to do. You now know to stay clear of that particular car.
With no driver, and no other form of feedback driven by you, the humans outside of the driverless car are more powerless than they already were. They are reduced to trusting and hoping.
A few weeks ago I had the nice surprise that my usual strategy of starting to walk on the pedestrian crossing to test if a car driver has the intention of stopping, resulted in an emergency stop from a Tesla coming for the right (it was a two-way street and I was testing the car coming from my left). The Tesla correctly stopped, in accordance with the law. Its driver wasn't happy.
The waymos in SF seem to indicate that they see a pedestrian (when they are waiting to turn) by showing a pedestrian symbol on their spinning lidar hat. It's pretty obvious, but I don't know how consistently it is shown.
Oh that's interesting thank you. Will look for it.
I do believe that computers are much more likely to have attention span and sensor capacity to reliably spot pedestrians and other random wheeled traffic. The problem is "reliably", which needs to be near 100% for the pedestrian to decide to ignore that it's still in full motion. So it will be interesting to see if such notification helps or not. If we were all wearing AR glasses, a specific car could signal a specific person that it noticed and is tracking them - that THEY are safe from IT. But we are a few years from that.
I'm always surprised how few people know as pedestrians to look at me, the driver and not my car. Some don't even look at the car. I developed a habit of looking for their gaze and if they don't look back, assume they're not fully aware and just am more cautious.
This works because I, as a human, know this and can compensate when they just rush the crosswalk without being fully aware of their surroundings.
Just because a pedestrian is looking in your direction, it doesn't mean they can actually meet your gaze. I don't drive and always try to make eye contact with drivers, but tinted windows and windshield glare often make it impossible in reality.
So I will stare at where I know the driver's face ought to be, but I can't actually tell whether they have seen me. Tinted windows darken the inside of the car and that makes windshield glare all I can actually see.
It's probably equally simple for a machine (if not simpler) to figure out if a human has looked at the oncoming vehicle with sufficiently advanced cameras and computer vision. However, we use a lot more hints (change of pace, facial expressions that indicate presence and focus, nodding etc), which, while not outside the realm of AI CV, would surely need lots of training, yet it comes pretty naturally to humans.
The GP was talking about the driver knowing if the "human has seen the machine".
For the inverse problem, we could simply start adding screens (instead of windshields?) to self-driving cars that acknowledge the pedestrians in a particular way (when there's only a few people, in the Black Mirror realm, they'd actually greet them by name using facial recognition and universal DB of everyone :).
What? I was absolutely talking about human cyclists and pedestrians needing to know that the machine sees the human, and not by blind faith, but through some active explicit demonstrable indication.
Of course there is, the behavior of the machine should be the same as of a driver - the car shaped object starts to slow down in a way that will make it stop before hitting me at the pedestrian crossing.
It is not possible for me to see you in your car with any sort of reliability; even if I could the benefits are dubious and again, unreliable.
Looking at your car is all I need anyway - I can tell if you've seen me by your behavior, you're either slowing down to yield to me or you're not. If you're not, the only possible outcome of knowing you're seeing me is being misled into stepping into your path of travel.
It's not all or nothing. A hundred different things all add up and none of them do the entire job, nor are entirely unnecessary simply because they don't do the entire job.
Hey I'm always surprised at how many pedestrians walk around all in black (or dark) with no light, no reflecting stripe, no nothing and no care in the world. Even with the best of intention of drivers, it's a death wish. And it's everywhere.
Or maybe it could have a signal that says it's safe to cross.
Humans tended to use hand gestures for that around here. Or flashing front or turn lights to disambiguate it.
And the person crossing would often nod or just start moving.
AI could be even more explicit about it. It can check if you're noticing the signal too, much like the driver does.
People are conditioned to respond to crossing signals, why not give the self driving car one of those.
> I just heard someone mention in a podcast how as a cyclist they need to make eye contact with drivers to be safe, and there is no one to make eye contact with.
Cyclist here.
Are you worried Waymos are going to blow through stop signs (or red lights) AND run you over?
Or are there other situations where you're making eye contact with drivers?
For example if a car is waiting at a stop sign, or getting out of a driveway, and I have priority, I always check if the driver is looking at me. If not, he possibly didn’t see me and might go while I pass.
It happens that drivers only check for any large moving object (because they expect cars) and don’t notice a cyclist.
Basically any time a car has to wait for me to pass, and I don’t have much distance to react if it doesn’t, I make sure the driver has seen me.
I feel like I saw a self driving concept in the last few years where the headlights had eyes that could appear to be looking at you to signal that it was aware of your presence?
This makes a lot of sense, also for pedestrians and other drivers. Having seen these things live in action, it creeped me out to see an empty drivers seat.
> Having seen these things live in action, it creeped me out to see an empty drivers seat.
I wonder if the safety gains could be just due to people being creeped out around them, and behaving more cautiously. Perhaps results would be different if they carried mannequins simulating a real driver in the wheel.
> Maybe they will prove to be so reliable that you don't need to worry about them. You can just reliably assume and predict their behavior as animate objects about the same as you do for inanimate objects.
This seems to be what is happening. There was a comment here by some user not too long ago that said they deliberately drive in front of the Waymo’s because they trust them to slow down.
You can also do that with human drivers and it will work for a while. The question is how often it doesn't work. A single person driving in front of a Waymo a couple of times can't determine that with any statistical significance.
I think insurance prices will drive adoption of self driving very quickly.
Consider: if a non-self driving car is in an accident with a self driving car, it'll almost always be the non-self driving car at fault. And with the telemetry from the self-driving car, they can prove it too, so accidents that would have been no-fault or shared fault become fully the non-self driving cars fault. And so I think insurance for non-self driving cars gets expensive fast as there are more and more self driving cars on the road.
Insurance for human drivers stays the same. If anything, it goes down, because the more Waymo-level autonomous vehicles on the road, the safer it is overall.
Insurance isn't a zero sum game, where "less insurance spent on autonomous vehicles" means the insurance companies have to make up for it somewhere else.
In a way, "car driving getting safer overall" isn't great for car insurers, because they make money financing auto risk, and if there is 50% less auto risk then they have less addressable market.
Insurance is, however, a for profit corporation, out to maximize profits, so they'll charge both groups a premium, and then add a service fee on top of that.
And when they post the obscene margins as a result of doing this, other insures can come in and undercut them. That's a key ingredient to making insurance work well. Its also why "insurance" doesn't work for healthcare.
Nothing about that changes with the introduction of autonomous vehicles though, and it's a very open and highly competitive market. It is extremely easy to switch your auto insurance provider, and anyone who can pay some actuaries can start up a new business. (Though few do, because there isn't a huge margin to take away from the industry as it is.)
Why would insurance need to get more expensive for manual drivers as self-driving adoption increases?
Ideally, insurance cost for self-driving cars would just be lower than insurance costs now (proportional to risk), and even insurance costs for manual drivers might go down because their risk decreases as well.
I think it's more about when the market is still majority manual. 20% self driving might see premiums for manual cars go up significantly because manual drivers would have more at fault accidents per mile than before (assuming the same rate of accidents, but most accidents with a self driving car are the manual driver's fault).
There will also be things like not having DWIs and even cheap parking (since the car can drive away and park) that'll net out for self driving. And feedback loops there- the same size police force only pulling over manual cars from a smaller and smaller pool.
> assuming the same rate of accidents, but most accidents with a self driving car are the manual driver's fault
This is a very shaky assumption in my view; more realistic would be (to me):
1) Overall accident rate reduction even for manual drivers
2) Reduced damage in accidents even for manual drivers
3) Slight increase in "at-fault" allocation toward manual drivers
It seems very likely to me that those would result in net favorable effects even for manual drivers.
I believe a lot of driving will move to full auto once viable, but not all of it because lots of people like to drive recreationally, and I simply don't see insurance rates for manual drivers spiking out of control-- just compare insurance pricing for less-safe vintage cars, which is also perfectly reasonable.
Insurance premiums are higher for higher risk individuals/circumstances and lower for lower risk ones.
If automatic driving lowers the floor of risk due to even lower risk than even the lowest risk human driver, insurance premiums will adjust to compensate and accomodate the now higher risk human driver compared to the now lower risk computer driver.
And here's the kicker: Insurance premiums don't even have to increase for the human drivers. Once owners of automatic driving cars pay even cheaper premiums, that becomes the new baseline for "cheap" car insurance and the rest should be plainly obvious. "Want lower rates? Get an automatic driving car."
I'm not disputing that insurance for full-auto cars is likely to be cheaper in the future, but I don't believe that manual driving insurance will ever get "cripplingly" expensive-- it might even go down compared to now thanks to reduced accident rates (contributed by full-auto cars).
Also compare vintage-car insurance (which are in many cases drastically less safe): Pricing is perfectly viable there and has been for decades.
This is not really how accident liability is sorted. Most countries have a clause in their traffic law that states that even if you're in the right, you must still do everything feasible to prevent an accident. If anything, I would imagine that self-driving cars would be held to a higher standard here, not humans.
The number of drivers that i see get into aggressive non-prevention habits just so they can "be right and angry" is kinda mind blowing. Ie someone goes into a roundabout when they perhaps shouldn't have, and so the other person who likely had rightofway guns the gas to make it even more obvious and risks and accident just to be "right".
If you watch those dashcam communities you see similar behavior in what feels like half of the accidents. Someone else in the wrong but boy oh boy did the camera driver do nearly nothing to prevent the accident.
Believe it or not, a lot of driving takes place in heavily driven territory :)
And I'm sure the autonomous vehicle will have a go-very-slowly mode for navigating people's driveways and similar places the mapping cars haven't been yet.
The world is pretty well mapped at this point I don't think this will be a showstopper.
And the nice thing about driver less cars is that they can drive wherever and whenever pretty cheaply. There's no driver to pay. Just the electricity bill for charging the vehicle and some servicing/vehicle depreciation and other fixed cost. That's a race to the bottom in terms of cost.
There's no good economical reason to limit this to just small areas. You might charge passengers a bit extra if they are further out or even for the distance the car has to drive to pick them up. But there's no good reason for that to be very expensive as it would be with a paid driver.
This. The boomer generation can not be outvoted of rights, even when they will crawl the streets being traffic hazzards with dementia. So they will be pushed howling and screaming towards automated driving.
These results are meaningless, Waymo only drives in controlled, good weather, urban streets, while the human control group does not. This is not science, this is PR. We need independent investigations.
We learned that the hard way with Boeing, you cannot trust a company to self-regulate their safety standards. But with Musk in power, the legislation might encourage them to do so.
"controlled areas" is relative - for example it includes mayhem zone SOMA and the hills. Also they just got started, relatively speaking, and have been expanding. Give it a minute.
Really, computers seem more naturally suited to driving in SF than humans - from the point of view of paying attention to enough things, not having to pay attention to map navigation and having more modes of sensory input than humans.
I'll believe such companies with my life and life of my family when they can handle usual South east Asia capital city for example also during monsoon, say Bombay. Because this is in stake, nothing less.
This aint some web app deployment where flaws are annoying but thats about it.
I've just finished 2 day 1500km drive back home across Europe in various storms, heavy rain, a bit of snow, a lot of it in the dark, tons of properly dangerous idiots, weird non standard marked repairs (we talk about germany), some serious accidents along the way, few nearly misses. Our F11 BMW 5 series took it on effortlessly, complementing my and wife's skills but not interfering. I am not stellar but definitely above average driver. I've done this drive maybe 30x over past 15 years, without kids normally in one 16-18 hour push. Nr of clueless idiots is definitely rising and roads are definitely more full.
No way in hell I'll trust something fully with most precious stuff in my life until its properly battle proven by billions of miles in harsh complex situations, which what you describe it isnt. Till then there is no self driving, just steps towards it. I don't need this for driving around some short distances that much. Good luck with beta testing to ya all.
I have taken a Waymo in the rain before, if they have stopped supporting that as part of the service that’s new, buts it’s definitely within the systems capabilities. It worked great
They can handle almost all weather except snow. Freeways are coming soon. From what I heard, San Francisco is way above average in terms of driving difficulty.
Did the paper really prove its claim? It set aside more than half of Waymo's claims history and compared only the most recent portion of its own history to the all drivers benchmark. All drivers do things that Waymo doesn't like operate outside specific city centers, drive in snow or other severe weather, real people drive more during the day in traffic but Waymo is biased toward nighttime (less traffic), and real people drive freeways and high speed roads.
If you rigorously enforced those limits on regular drivers and additionally imposed an interlock device to prevent DUI and remote insurance company monitoring of driving habits-- that's the comparison population I want to see. Otherwise, this is a press release disguised as an academic paper.
It seems like you want them to explore a different question. Im confident that the best human drivers are better than Waymo. They are also better than all human drivers, by definition.
If you are worried about other drivers hitting you, the "all driver" comparison seems relevant. If you want to know if Waymo is a safer driver than you are, that will depend on where you are on the spectrum.
Maybe DUI's should be included, but a paper drawing the conclusion that "Waymo is safer because its computer drivers aren't drunk" suggests very different policy actions.
Unless we really can finally stop drunk driving overnight after all these years of trying I don't agree the comparison should result in significantly different policy. I want to know how Waymo statistically compares to all of all the actual drivers when setting policy, not how Waymo compares to what we wish real drivers would have been like. The only mistake would be to compare solely to drunk drivers, which is what you comment makes it sound like.
I would like to see the data sets filtered to match driving areas though. I.e. Waymo drives almost exclusively in urban environments so per mile data should be compared almost exclusively to urban driver data.
> ..not how Waymo compares to what we wish real drivers would have been like.
I want to see "all the things", but in specific reply to this point, i want to see this because i have no intention to commit a DUI myself. So the bar for me personally is how well it'll do compared to safe drivers.
Lots of stats are useful, but i don't think we should undersell the various "ai vs me" comparisons that individuals could use. Since those are the ones that will help people decide if they themselves can/should/shouldn't take a Waymo or w/e.
But in terms of policy decisions, whether they should be allowed on the road, these are the correct numbers. Convincing people like you to ride is not the goal of this paper or its claims.
Results demonstrate that the Waymo ADS significantly outperformed both the overall driving population (88% reduction in property damage claims, 92% in bodily injury claims), and outperformed the more stringent latest-generation HDV benchmark (86% reduction in property damage claims and 90% in bodily injury claims).
I can't trust that number if it comes from Waymo.com. I'd need to see it from an independant third party who has no interest having their finger on the scale
So not independent. And published as a white paper on Waymo's website. And only addresses the collaboration between Waymo and the insurance company, no mention of who funded the academics. Pretty sure if you dig deep enough, Swiss Re is somehow a Waymo investor.
I'd trust this more if Waymo had a direct pipeline to an anonymized public repository of their complete data set for independent analysis
I mean.. I get that for the Waymo guys. But the SwissRE ones? Seriously? Who _should_ do that evaluation? You will necessarily have Waymo co-authors. They have to provide the data. They are the only ones who have that..
Most of this data they publish to the Government for legal reasons. Sure they could lie to the Government, but ask Cruise how well that went.
Anecdotally, I see Waymos around town, and they are very clearly driving in a slower, safer, more cautious manner. a 92% reduction seems very plausible.
Good to see studies of incidents in comparison to human drivers. Historically what I have seen is 'self-driving car crashes into X, which seems to be concentrating exclusively on outliers to prove a point.
This sounds quite a bit like massaging statistics until you get the numbers you like.
A few things that pop to mind: did we exclude drivers who were not eligible to drive (under the influence, no valid driving permit, cars with issues [braking broken]...)? (This mostly matters for evaluating the average performance compared to average non-impaired human — a drunk person might mess up with self-driving vehicle sensors as well just for kicks once they are in suitable numbers and widely available)
Did Waymo drive across all the same areas and time periods as the humans did with an equal "proportion"?
Could we compare "time behind wheel" instead of miles travelled? (Eg. lots of small crashes happen when traffic is at a standstill)
There are so many ways to slice and dice this and come up with favourable results for self-driving systems that I can't but be dubious of any claim.
But notably, if we can liberate ourselves from driving through congested traffic, it would certainly reduce the incentive to speed and aim for more efficient driving as we'd be free to do other stuff while we get to our destination — and simply slowing down would buy both human and software drivers time to react and avoid most crashes (nothing really saves you from someone recklessly driving straight into your car).
> did we exclude drivers who were not eligible to drive
One could argue here that it would be unfair to exclude the worst performing humans. Plenty of humans cause accidents for easily preventable reasons, but that's an honest part of our behavior. Or to rephrase, I think excluding the categories of people who drive terribly, but do still drive nonetheless, would be less representative of reality.
> Did Waymo drive across all the same areas and time periods as the humans did with an equal "proportion"?
No, they didn't. Their methodology didn't adjust for things like highway vs city street miles, although it did adjust for city and state.
> Could we compare "time behind wheel" instead of miles travelled?
We could, but that would incentivize slow driving rather than safe driving. You need to travel the same minimum distance from a hypothetical "Point A" to "Point B" no matter what, but the number of minutes could be inflated easily. Economic factors discourage inflating the number of miles.
> There are so many ways to slice and dice this and come up with favourable results for self-driving systems that I can't but be dubious of any claim.
I think it's fair to criticize the fact that they didn't adjust for geographic area with as much granularity as they should have (i.e. for highways vs city streets, or for excluded zones like airports). But other than that, their methodology seems solid to me.
I understand that viewpoint from an insurance perspective, but not necessarily from my own safety perspective: I wonder how well Waymo would fare in circumstances I see as a driver. Eg. humans can control whether they drive under the influence or not (eg. if you don't take your car to a bar, you ain't coming back driving it).
But most notably, aggregated statistics like these focus on comparisons similar to comparing averages vs better statistical representations (like average vs median or p75 or... eg. it's possible for one driver to cause 20 collisions, and in a set of 10 drivers with no other collisions, we'd be at 2 collisions per driver instead of 0 really).
It's also not only about highway/city street miles, it's also about time of day: if Waymo is proportionally less on the streets and areas where there is more risky driving (eg. around bars and around midnight to 3am when there are also drunk pedestrians around) compared to human drivers, that would obviously skew the numbers in SDV's favour. Again, potentially they did since they acknowledge a separate study focusing on those, but later on they only talk about aggregated insurance claims.
They also acknowledge they are not accounting for accident severity, not even by using the dollar amount: while they had no fatalities, it's obviously important to weigh accidents by their severity.
With all this said, this does demonstrate current insurance liability of an average human driver compared to a Waymo SDV in Phoenix and SFO, but only once we have more comparable data should we make a bigger claim.
I get where you are coming from and you are correct. This is, up to a point, a PR piece of Waymo and we should take that into account. But they are pretty transparent about what they are doing. You just have to read the report. And for their metric, which rounds to "all distance travelled in the interesting ZIP codes", they are succeeding. What I'd take away from this is: If you give them a different metric, they can optimise for that too and they will end up being better than humans.
That doesn't answer the questions: 1) Is it worth it? (I.e. what does it cost to achieve that) and 2) Is a Waymo better in any situation? I guess they don't have the answer to either yet. I'm pretty sure their cost-per-mile is still dropping like a rock (probably though not fast enough for their accountants) and they haven't reached all situations yet either. But I wouldn't bet on "because it's too hard", but rather "because we didn't get there yet" (and yes, they will have started with the easy part, as they should)
AFAIK they drive on highways for testing, but they don't offer commercial rides on highway routes to customers yet. If that's still the case, I'd expect the majority of their miles to be on city streets. As I already pointed out, that's a big flaw in their methodology.
Regarding drivers that are drunk or otherwise driving illegally-- why would you want to exclude them from such a study, if you can not exclude them from real traffic?
If our current system of checks and incentives is not able to keep them off the road, then you need to take them and the accidents they cause into account when comparing driving safety with an autonomous system.
Also note that the margins are too big, since only about a third of accidents involve intoxicated drivers, and the accident reduction rates from the paper are much higher than that.
I don't want to come across as harsh or accusing, but the "Methodology" section contains a lot of detail (and a more nuanced discussion could be had without disregarding it entirely :P).
Because they don't disappear even with half the non-illegal vehicles being autonomous — some crashes also happen due to avoidance manoeuvres.
So the fairer comparison would be to say that if existing SDVs were instead being driven by human drivers, we would have seen ~60 property claims more and ~40 bodily injury claims more (forgot the exact numbers from the paper), which is what, like 0.1% improvement in claims overall (certainly still a great result for the small comparative number of miles travelled, but doesn't jump off the page really).
But because the miles traveled are so far apart, nothing is really a "fair" comparison.
But.. they would disappear? Not all of course. But some drive under the influence because there is no other convenient way to get home after a night out. And Waymo solves that!
And they are highly aware of their "small number statistics", they literally provide uncertainty ranges. For both SDVs and HDVs. And their values are consistent with increasing data. I.e. they ~only get smaller uncertainties but no significant drift. But they are statistically significantly better (which wasn't the case two years ago)
Are you suggesting Waymo would become available where cabs aren't due to economics, or that Waymo-like technology in privately owned vehicles would solve that? Because I wouldn't buy the former, but I can see how the latter would, yet we are far away from that.
I am not referring only to "small number statistics", but to obvious miscomparisons. We don't want SDVs to beat "average human driver", because average is worsened by impaired humans — we need them to "beat" at least an average, non-impaired human. Going by the numbers alone, they would achieve that too, but not by the same amount.
But more importantly, we need them to compare both in identical driving conditions, which is where most of my complaint really is. Unless Waymo has driven in the same amount (proportionally) of rush-hour traffic, road-works, risky-drivers-around conditions, parking into tight spots and garages... of miles out of their 25M miles as their human counterparts, we could very well be looking into Waymo-in-simpler-situations vs humans-in-all-situations. And the numbers would be meaningless, which to me they seem they are.
And it is important that they have achieved similar success with 6x the miles, but they need to account for more challenging conditions. And while some tricks can work while they are in small numbers (eg. avoiding road works by re-routing), it will lead to congestion elsewhere when such technology dominates the traffic.
I would also note that it is impressive, regardless of the conditions they operated in, that they were involved in exactly 2 "bodily injury" claims over those 25M miles, and both of those appear not to have been the fault of Waymo.
> did we exclude drivers who were not eligible to drive
I don't think we can discount people who aren't fit/eligible to drive _if they still drove_. They might not have a _legal_ right to be on the road, but they are still on the road.
It depends on how you want to measure your safety when driven in one of these?
Do you want them to beat a drunk driver, or an average, non-impaired human? If you got into a cab and cab driver was obviously drunk, would you ride along or get out asap?
By "letting" them compare to an "average" (skewed by impaired drivers' incidents), we are letting them appear much better than an average, non-impaired driver, whereas they might be, or might not be (numbers clearly suggest they'd still beat them, but not by this much).
>>did we exclude drivers who were not eligible to drive (under the influence, no valid driving permit, cars with issues [braking broken]...)?
now THAT would be massaging statistics! Just take the best drivers, that had enough sleep, no drink, perfect car conditions... No. You have to compare with the whole world, "as it is"
What's the guarantee we'll have only fully up-to-date, safety-checked self-driving vehicles in the future with no messed up sensors or controls (there'll be impaired people doing the checks, right? :))?
This is simple with a small, experimental fleet — so we are looking at the best case scenario for these vehicles, but the question is what does it fall down to in a realistic commercial application?
That would also be comparing safety, because averages are always skewed by bad apples (i.e. 1 driver with 20 collisions gets an average to 2 for a group of 10 drivers with no other collisions). We at least need to start talking about medians, standard deviations groups and such.
And we need autonomous vehicles to beat or match good drivers, otherwise, good drivers are worse off in the streets (and due to how averages are used, this might be more than 50% of drivers). Not sure why that's so controversial?
If you've used Google Chromecast or Google Chrome, a Tesla, or an Xbox, keeping software up to date isn't the chore it once was. Or you have to update to go online/have the car drive. The "guarantee" is that the other players who were too aggressive and did bad had to exit the industry. Uber killed a woman and exited the industry. Cruise dragged a woman 20 feet and lied about having footage of it, and exited the industry.
So you ask what's the guarantee, and the guarantee is the same as when Google came out with Gmail, and everyone said that's stupid because how are you going to read your email when that website is down, so then they made it so that website doesn't go down.
How do you upgrade brake pads over the air? Or ensure tyre and wheel conditions? Or any of the other mechanical parts of the car (suspension, axles, transmission in-so-much-it's-still used...) Or ensure sensors are not interfered with? As soon as you get Waymo into private hands (or competitors need to cut corners), you'll start having issues like these.
I don't think hoping for self-driving cars is stupid, and I would certainly love to have one I can trust to take me and my family around safely. But to get to that trust, I want better reporting than what this study provides and proposes :)
I think you should open the PDF and look into that xD Really they discussed this. I'm still skeptical that they measure what we are interested in. But _your_ points they discuss
I went through the report, and they reference other studies that consider my points, but they end up only measuring incident numbers, don't specifically say how comparable were the driving conditions (iow whether they are representative). If anything, they could have removed incidents by human drivers which did not happen within their operating constraints, but I did not see that mentioned.
But that data isn’t available to them. They are (iiuc) simply looking at the full set of insurance data and their own set. There simply is no driver-level data which would allow statistically relevant conclusions.
Up to now their data for their own cars is still not able to strongly constrain the results. You wouldn’t be able to tell effects of sensor degradation.
But let me ask this instead: What kind of analysis would you like to see?
The report specifically mentions a number of other studies and claims it improves on their methodology, and seems to want to have their methodology become the "gold" standard of measuring SDV vs HDV performance:
"This methodology establishes a foundation for future research into the safety impact potential of ADS and offers a framework for assessment as these systems continue to scale and develop."
Or:
"By analyzing an unprecedented volume of autonomous miles, introducing a novel benchmark, and utilizing third-party liability insurance claims, we aim to set a new standard for ADS safety evaluation."
Which leads me to worry that they are pushing for this to be considered comparative, and they want policy to be influenced by their claims. And they do:
"Our findings have the potential to significantly impact policy decisions, insurance practices, and public acceptance of ADS technology, contributing to the broader societal dialogue on the future of autonomous transportation."
So I would like for them to compare to median non-impaired human driver with a car in good order (both in time-to-destination and safety), aggregate those over time of day, week and year (basically ensure identical driving conditions between comparison miles), adding claim dollar amount as weight for how severe the incident was. I believe that would be a great start.
With them being Google and already tracking most Android phones as they go through traffic (they use this for traffic conditions), they do have a very rich data set they could compare to as well, but excluding that, they could exclude insurance claims which have happened in conditions they don't operate in (claims will always have time of event and location).
The study is about today. I would bot ever think to extrapolate that to “forever in the future” also is only one vendor. That is what the statistic shows and I think is fair comparison against human drivers currently on the street.
The study is about today and certainly has plenty of factual data in there, and their safety record is amazing.
But they are purporting to make this the standard for comparing SDVs and HDVs, and explicitly want this report to influence policy. They say as much in the report.
Still worse than public transport, wastes more resources and endangers cyclists and pedestrians. I have yet to see anything that deserves to be called progress in self-driving. Yet another scam to sell more wasteful, inefficient vehicles.
Waymo is public transport. If they introduce automated vehicles that have fixed routes and larger carrying capacities then Waymo is indistinguishable from actual public transport.
You’re right that it’s not the answer to everything, but today car-only infrastructure is considered to be the answer to everything throughout America, so much so that we’ve reached the point where public transit is the only answer to how to move people more efficiently right now because it’s quite simply where all of our transit funding should be going.
I agree that the US needs better public transportation, but most of the US is not a city.
If the nearest grocery store is 30 miles away and the nearest bus stop is several miles away from your home, then using public transportation turns this into a half-day affair.
"More" is the critical word here. I don't see any reason why insurance premiums should increase. Insurance premiums aren't sin taxes or some governmental incentive, they are offered by private firms.
I would expect premium levels to continue to be driven by liability pool payouts plus a profit margin. Autonomous vehicles don't increase driver risk or liability pool payouts. IF anything, they reduce them due to safer driving conditions for the humans.
Insurance always balances between cost pooling and making people pay for their increased risk. Car insurance has always been more expensive for riskier drivers: young men pay more, driving “high risk” cars pay more.
If the overall liability pool goes down but the human drivers become an outsized part of it, the insurance company will likely shift that cost to them. If they don’t, someone could set up an autonomous-only insurance company (or google could just insure itself) to take advantage of the lower risk, making the human insurance more expensive (even more so that the liability since the overall pool will be much smaller).
Like the Borg, every Waymo is the same driver. So in column A, the insurance company has a meta-driver with many millions of miles of statistics. In column B, we have slow-reaction and non-lidar-equipped you, who had a fender-bender last year and a speeding ticket 3 years ago.
I apologize if I gave the impression that I had personal knowledge of your driving record and traffic violations spanning over the last 36 months. I assure you, my intent in using the word "you" was merely to illustrate a general situation in which an individual (though surely not yourself, palata, specifically), will find themselves being penalized for a prior incident. As my comment was surely read by thousands of HN readers, one imagines that as each of them read the word "you", they pondered their own driving history in their lived experience.
Now that we have established the fact that you, palata, did not have a speeding ticket 3 years ago, I would encourage you to think about the application of my statement, to your actual situation. That is, you may substitute the words "speeding ticket 3 years ago" with any traffic violation or insurance claim where you were determined to be at fault, over any time period which might not be 3 years. Then the question is raised of whether you, palata, would incur higher insurance costs than Waymo.
Let me just try to understand your point. Someone says "you are objectively higher risk than an autonomous car, so it's normal that your insurance makes you pay more if you drive".
To which I answer: "how objective is that, when the claim comes from the company selling the autonomous car?"
Then you tell me something that I understand as: "well, you are objectively higher risk because if you look at your record, you have made mistakes that show that your are".
And when I say "actually, I haven't", you patronize me? Is that how it works?
No. I haven't killed more people than autonomous cars have. Before you can prove that over my life as a driver, I will cause more damage than an autonomous car would have, then you don't get to say that I am objectively higher risk and therefore should pay more insurance.
Now you can try to look at studies that try to prove that. And when I ask "how objective is that?" when the study comes from those who benefit from those results, you can stop for a minute and consider that maybe, I am allowed to ask.
That's initially true, but as the car ages, and maintenance is done (or not done, or done improperly) the car's aspects will change.
And as new cars are developed, obviously there are more "drivers". Essentially every year there'll be new drivers for every company that makes autonomous cars, possibly multiple.
Yeah, it's still a lot fewer than the individual humans, but it's definitely not just "one driver".
They should also ask for more money from sick people when it comes to health insurance, as they’re more prone to getting sick again and again and hence consume resources, after all how many Luigis can there be?
Which is to say that naked capitalism works until it doesn’t, and when it stops working big chances are that it won’t do it gracefully, see how present-day health insurance CEOs are now scared shitless of Big City streets.
If self-driving cars are widely available, and they can drive everywhere a normal car can go, then you're not driving manually because you need to do so, you're driving manually because you like it, and the society isn't there to pay for your hobbies.
They won't be "widely available" because of the higher prices, just look at the EV market right now. You won't get a self-driving $2,000 clunker, and when you drive a $2,000 clunker you don't do it because you "like it", but because that's all that you can afford.
Did it happen for EVs in the here and now? Where can I find a SH EV clunker for $2,000? I'm talking about an EV with performances similar to those of a gasoline-based $2,000 clunker.
Otherwise, believing in "technological progress" as a potential solution is just just that, an irrational belief.
Despite the law, driving in America is a necessity for many. That means our bar for taking it away is high. Quantify the safety delta and have an available alternative, and we can start lowering the bar for taking dangerous drivers off the road. (Frankly either >1 DUI or >1 at-fault crash should result in license revocation. We should also delay license issuance until 18 years old and have restrictions on the license until 20 [1].)
> Quantify the safety delta and have an available alternative
I would love it if these comparisons included "bus stats / average passengers". There are ways to just reduce the miles driven and the chances of collisions at the same time, without new tech improvements.
I assume such a change would heavily impact you as well? 40% of all seriously injured in the traffic are single-bike accidents. Here in Sweden that is.
yea I don't particularly trust a for profit corporation to take my safety as a cyclist more than the minimum viable consideration. I used to be really excited about self driving cars until I realized that in US culture, it would basically be used to further prioritize car dependent infrastructure at the expense of proper walkable spaces and alternative transportation.
we could have amazing and SAFE cycling infrastructure NOW if american culture wasn't so objectively trash.
What year would the EU* ban human driven cars on public roads?
That's it. Million different opinions ranging from "never" to "within a couple decades for sure", weak, strong opinions, and of course many interesting topics to branch out into.
*: sometimes I ask with US, or particular countries, it doesn't really matter for the sake of conversation.
Fun fact: people working in automotive tend to say within this century, people who like and owned many cars say never :-)
It's not a particularly interesting question as it's just random guesswork and gutfeeling until self-driving cars see actual widespread availability and rollout.
If at some point all new cars have full self-driving capabilities with widely accepted performance, then it's not impossible that after a decade of that as status quo, cars end up shipping with manual driving only being an override for specific maneuvering, and eventually having regular licenses lose rights to manual driving on public roads.
This comment shows exactly why I like this question.
Lot of related topics to branch out to! I actually wasn't ever interested in the "cut off date", but all these side topics, and why some people think differently about what the trajectory looks like today.
They were a leading cause of death when they were a common mode of transport.
Riding a horse is very dangerous. Its just regarded as an acceptable risk (its something many people encourage kids to do) for cultural and social reasons. A British government scientific advisor was fired for pointing out that riding a horse carries about 30 times the risk of a "serious adverse effect" as taking ecstasy.
>A British government scientific advisor was fired for pointing out that riding a horse carries about 30 times the risk of a "serious adverse effect" as taking ecstasy.
A change like this is entirely unpredictable from here, because it would be driven (pun intended) far more by political processes than any objective safety consideration.
I neither own nor really like cars, but I think a ban within the next 50 years is super unlikely, and anyone that disagrees is delusional :P
Just consider vintage cars: These are vastly more dangerous for both occupants and pedestrians/cyclists than existing cars and also pollute more. They have not been banned in the last century and no ban seems to be coming anytime soon.
If you are not gonna ban vintage cars, why would you ever ban manual driving (which is much less harmful in multiple ways)?
Once adoption picks up, a ban becomes useless anyway because the fraction of problematic cars decreases (and no longer really matters), while any decisive legislative action towards bans gets tons of pushback from enthusiasts (and no real political gain, because the average voter does not give a shit about single digit traffic accident rate reduction anyway).
It's already borderline happening with the satellite supervision.
Here in Italy insurance is 20% more expensive at the very least and the gap is widening if you don't put a tracking device on your car which checks position, speed, etc.
I have a hard time justifying why one would not want it (besides privacy), all people that complain about it are people that regularly drive above the speed limit or pull dangerous overtaking maneuvers.
We have tracking like this for an app where you can rent electric cars in the city short term. Problem is it sometimes flags you, banning you for a month because you went under a bridge or the road has a smaller side road separated by a ~1 meter island limited at 30 but the road you were on is a 70. When it's a rental car app you don't really care to argue but I definitely wouldn't want to deal with this bs with an insurance company. They'd have 0 incentive to resolve this.
It's a question if you put people's individual "privacy" (quotes because you're driving in public, on public roads paid for by the public, under public laws, and under the public's view - there's not really privacy as to how you're driving, everyone out there can see it) over people's collective right to live?
Road deaths are among the leading causes of death in multiple developed countries.
Roads are public but unless you have someone following around you can't know where everyone is at all times. But with tracker now you have mass surveillance!
We criticize chinese government but we do much worse.
I did write "besides privacy" because it is a valid concern indeed.
But considering that most people don't give two damns about their privacy (or at least act like it, keeping 24/7 a tracking device on them and sharing all of their lives non stop) what would be their valid reason to not have a tracking device for insurance purposes on their cars?
Will the data collection and interpretation be perfect? What if the map with the speed limits is inaccurate and my commute goes through a road where the limit is 70, traffic drives at 70, but the system thinks the limit is 30?
My car displays the speed limit in the dash, as a helper, and sometimes the above happens. If it had automatic braking for crossing the speed limit, it would be a disaster.
Also, if I drive 70 on a 70 road completely covered in snow, will the system know I'm doing something dangerous?
Automatic judgement of people is a bad idea, and it surprises me that anyone who is working in software development would think otherwise.
Yes, and of course you must pay because the autopilot doesn't work when there's snow or rain or sun on the horizon or it's too dark, or the volcano near you is throwing ash and there's wind so you're now in a sandstorm, or wind is carrying sand from the sahara all over the mediterranean to you.
(all of these regularly happen all in the same place)
> More recently, light snowfalls occurred on 9 February 2015, 6 January 2017 and 5 January 2019, but the last heavy snowfall dates back to 17 December 1988.
Says wikipedia. Yeah, I don't know what I expected. Of course there's no actual snow in the meditteranean. It "lightly snowed" 3 times in the last 10 years.
Highways seem like they're safer than other roads but if we had fully autonomous highways I imagine speed limits could be effectively removed drastically reducing travel time.
Well really it would be "you get a discount for the percentage of your mileage driven with a autopilot enabled." You can pay less if you drive in a setting that saves the insurance company money (on average).
Is it possible to compare this to humans in another way? Could it be said that Waymo avoids 95% of the accidents that we regularly get in, and does have 3% of accidents that we manage to avoid. I think a major hurdle to mainstream acceptance is the errors that robots make, and a real mind doesn't.
Maybe Elon Musk will yield enough power in the US government to influence safety policies in such a way that is harmful to waymo. It doesn't seem like conflict of interest is bothering him much.
So Musk is effectively the most powerful man on the planet right now, he's a major player in 3 key industries, and the head of effective (de)regulation in the largest consumer market by dollar volume.
Isn't this a bit like ole' good Crassus in ancient Rome? Head of the government and richest buddy aroubnd? If history repeats, we'll see the American Imperial Era rising up very soon, were the interests of the state will fraction between the interests of the oligarchs and the lines blurry between law and personal convenience of the wealthist citizens
Crassus died pursuing an absurd extremely expensive boondoggle. So if history is to repeat itself maybe SpaceX is sending that rocket to Mars after all.
> the interests of the state will fraction between the interests of the oligarchs and the lines blurry between law and personal convenience of the wealthist citizens
That's basically cyberpunk. And yes, we're pretty much heading towards megacorps.
Have they solved their mapping problem? Because that, more so than the added Lidar hardware cost, seems like the lynchpin here: As soon as Tesla works, it works everywhere*.
Last I heard Waymo, they have to do every city, one at a time – and there are a lot of cities.
The mapping problem keeps getting brought up (often by Tesla enthusiasts).
It just doesn't seem like a huge problem? If you have cars equipped with high resolution Lidar and cameras, mapping new routes seems like it ought to scale as the fleet is expanded.
How did Waymo do this? As far as I know, replacing drivers with an AI was the goal of ever-money-losing Uber and Lyft, however, Waymo managed to get to this much sooner while having less funds.
Google/Waymo actually started way earlier, and had fistfuls of cash from the very start.
Google hired a bunch of people who'd done well in the 2005 DARPA Grand Challenge [1] - including Sebastian Thrun and Chris Urmson who lead the winning team.
Thrun is also behind Google Street View which in some regards [2] looks a lot like a self-driving-car sensor suite. So Google was having LIDAR-equipped, high-precision-GPS-equipped cars drive every street in every prosperous country, starting back in 2007. Uber wasn't even founded until 2009.
Other Google hires had a similar background - such as Anthony Levandowski who competed in the DARPA Grand Challenge with an autonomous motorcycle. He later gained fame after being caught stealing a bunch of LIDAR schematics and similar trade secrets while leaving Google for Uber.
We also know from court documents that Google was throwing around mountains of cash, even when the self-driving-car division had no revenue. Waymo was set up as an "internal startup" giving employees "equity" so Levandowski left not just with internal documents but also with over $100 million.
That's a stark contrast to a lot of other players who'd need to show investors a lot more to get a lot less. This endless money was undoubtedly helpful in giving them the confidence to design for L5 autonomy from the start, no need to design a lesser system to get the money coming in early. And of course if you can pay $100 million for one guy, you're not going to baulk at the cost of a few $10k LIDARs so long as the people making them claim the price will fall to $200 for automotive quantities.
The 2005 Grand Challenge simplified the driving problem a great deal - no pedestrians or moving vehicles to deal with, safe and driveable route guaranteed to exist - but it did a lot to focus development efforts.
I thought L4 was Tesla Autopilot style, someone in the driving seat, hands on the wheel, ready to take over with half a second's notice, and taking liability for insurance purposes? And the geofenced "freeway driving only" stuff some German carmakers are coming out with?
Whereas Waymo's taxis don't have anyone in the driving seat at all?
I was under the same impression, but the parent poster is right: the main difference between L4 and L5 is universal applicability. Waymo is L4 as it can drive without a driver, but only in certain areas.
> I thought L4 was Tesla Autopilot style, someone in the driving seat, hands on the wheel, ready to take over with half a second's notice, and taking liability for insurance purposes?
I've also not seen much drama coming out of Waymo in the past. I could be wrong but I guess this is the benefit of keeping your head down and focussing on the problem instead of growing too large.
I think they did something very smart, but I don't know if it was accidental or intentional.
Waymo existed for like a decade on the basis of "we have a pile of money and our founders think it's cool". They moved slow and broke nothing. They made slow, incremental progress exploring the self driving space for years with plenty of funding and no expectation of a product launch.
So when a bunch of other companies came along, years and years later, and were all "we're going to have self driving cars next year!!!", and Waymo had their "oh shit, we better actually make this into a product or why do we even exist " moment, they were already in a really good place. They weren't rushed, they probably could have pivoted to a product a year or three earlier. They'd already solved the problems everyone else was just handwaving.
If it was accidental, they sure got lucky in the amount of time they were given with relatively little pressure, and in the timing of that competitive pressure at the end.
Started 15-20 years ago, whereas Uber and Tesla hopped on the bandwagon late and tried to play catchup. I remember talking about the self-driving project at Google in 2008 or something like that. (IIRC they were trying to use Haskell for something.)
Google spent a lot of time and money attracting people who want to work on general hard problems rather than experts in specific domains. Maybe tasking a group of people like that is more likely to succeed in inventing something than building a team around the domain itself, especially in a space that has a lot of unknowns.
Waymo (January 2009) is older than Uber (March 2009) and has always been part of Google/Alphabet. So to answer your question they did it with enormous amounts of time, money, and talent.
* Time - Waymo started work on this in 2009 and are now getting a faster adoption in 2024. Uber started in 2015 and sold it off by 2020. Time has given Waymo the advantage of slowly ramping up and starting with toe-dips.
* Funding - Waymo has spent more money to get where they are now (probably 2-3x what Uber spent in total). Google also has deeper pockets than Uber, which means that there is less pressure on quickly ramping up Go To Market or immediately getting profitability.
* Culture - Waymo was much more cautious (likely because of funding structure) which is now paying dividends in terms of regulatory approval and consumer trust ('sometimes you have to go slow to go fast').
Not sure about the "less funds" bit, but Google started funding self driving cars in 2009. The continuous investment in tech is just now starting to make the dream feasible - as far as I know self driving cars were never more than a side project for Uber and Lyft.
With Cruise gone now there's basically just Waymo, Zoox, and Tesla (bit controversial) that are the names thrown around when talking about self driving market share, and out of those only Waymo has a functional service.
I adores wondered the same thing for any new breakthrough. I listened to the Eric Schmidt interview on Diary of a CEO regarding this and definitely agree with his reasoning. He basically says the reason new companies get ahead while we wonder why the existing companies who should’ve been doing this stuff are so far behind is because the old companies were also do everything else. He was especially talking in terms of AI and how Google seems behind while OpenAI came out so quickly.
Do you think the future is behind depth analysis from 2D pictures rather than in lidar scanning? I know lidars are quite pricey, but I haven’t heard of regular camera breakthroughs in the domain.
If you can't do reliable depth analysis from multi-angled pictures then you are not seeing the real world. That's somewhat of a problem for a task that is 100% visual and designed for human drivers.
Waymo's safety data, on carefully curated paths, with human assistance and in the most car-centric cities in the world, is worthless. When these things hit the real world, with black ice, open manhole covers potholes etc. etc. don't expect lidar to save you from disaster. Maybe the average safety case can be made, but that won't be acceptable if they regularly kill people by confusing cows with flying plastic bags.
No, it is not. I'm not saying stats that include DUI accidents are not good or useful, just that additional analysis (together with these results) would be interesting as it would give us better insight.
For example, let's say an analysis tells us that a disease X has 1.0% mortality rate. Would you consider additional analyses that tell us that 90% of these are kids under 10 and that mortality rate for kids under 10 is 10% an useful insight or not?
EDIT: Maybe a better example would be a results from real studies that show a slight U curve for relationship between alcohol intake and total mortality. People not drinking any alcohol have higher total mortality than people who drink in moderation. Without additional data it seems that a moderate amount of alcohol is good for your health. Data from a newer study however shows that, if you exclude people who don't drink because of health issues, there is no U curve anymore and people not drinking any alcohol don't have higher total mortality than people who drink in moderation.
I'm surprised that no one has yet commented on the inherent reporting bias here. Waymo commissioned a study, conducted by Waymo employees, for Waymo to study Waymo's own safety record, and found that Waymo is safer than human drivers. Has this finding ever been independently verified?
Looks like this is comparing to human drivers on the same miles and conditions:
only the claims associated with vehicles registered to addresses (i.e., where the insured resides) within Waymo's operating zip codes in San Francisco, the Phoenix metropolitan region, Los Angeles, and Austin were included
Given how bad most human drivers are in my neighborhood - and I see vast amounts of them on their phones and not looking at the road - I'm thinking that this is not very surprising.
I have never been responsible for an accident and was never implied in a serious one with body injuries. Maybe I'm already avoiding >95% of the accidents by not driving drunk, not texting, paying attention, not speeding & basically following the rules?
People might go sovietic saying Waymo is better overall, so let's force it on everyone. Maybe we could achieve an even better result by being stricter on applying the rules and leveraging far less complicated technologies to prevent car from going over limit, preventing it to start when driver is drunk...
I wonder if this will ever make it to Canada's winter wonderland where there is ice and snow on the road for 4 months of the year (more or less in other areas.)
They've done some testing in Buffalo NY, which gets more snow than pretty much anywhere in Canada.
The hard part about self-driving is dealing with / predicting the behaviour of humans. That's something that humans are good at and robots are bad at. And SF has more weird human behaviour than anywhere in Canada.
Predicting the behaviour of vehicles in snow is something computers are much better at than humans. Solving the sensor problem for driving in rain/snow also seems like a standard tech problem where vehicles can quickly become much better at than humans who are limited to mark 1 eyeball.
It seems like they compared apples to apples. Which make that an exciting result!
“The benchmark was calibrated using both mileage (driving exposure) and residence zip-code (geographic region). Specifically, only the claims associated with vehicles registered to addresses (i.e., where the insured resides) within Waymo’s operating zip codes in San Francisco, the Phoenix metropolitan region, Los Angeles, and Austin were included.” Page 11
“Waymo also almost exclusively operated on surface streets (non-access-controlled freeways) with a unique distribution of driving that is representative of a ride-hailing fleet. In contrast, the benchmark represents the privately insured driver population that resides in these geographic regions. The associated benchmark mileage has more freeway driving than the Waymo ADS.” Page 26
“Including freeway driving makes this benchmark crash rate artificially lower, so, by including freeways in this study’s benchmark, the benchmark crash rate underestimates the true driving risk of where the Waymo ADS operates.” Page 26
“Four claim frequencies were independently calculated for vehicles within Waymo’s operating zip codes in San Francisco, Phoenix, Los Angeles, and Austin. Thereafter, these frequencies were proportionally weighted according to Waymo’s mileage distribution across these four regions. This resulted in the definition of a benchmark driving population which had driven the same driving distribution across geographic regions as the Waymo Driver.” Page 12
“Waymo also almost exclusively operated on surface streets with a unique distribution of driving that is representative of a ride-hailing fleet… Due to all three of these limitations being expected to artificially suppress the benchmark crash rate (underestimation), the benchmarking results in this study are considered to be conservative.” Page 26
I don't think this is apples-to-apples. You need to somehow filter to people who are not speeding or otherwise breaking traffic laws. My problem with studies like these is "human-operated" is an enormous catch-all that blends together too many causes for accidents.
1) Not filtering out drunk drivers is downright misleading. I am less interested in "is mandatory self-driving safer than human drivers?" than "is mandatory self-driving safer than mandatory breathalyzer ignition?" There are some
uniquely human downsides which are fair comparisons to AI - distractions, slowness, anger -but drunk drivers should be removed from this.
2) The bigger problem is people rationally speeding or running red lights. Waymo is strictly legal, but Tesla self-driving historically encouraged speeding, rolling stop signs, etc. Ubers and Lyft are preferred if you're in a hurry precisely because human drivers bend the law. Human speeding is not caused by a lack of intelligence and needing an AI to figure it out that it's dangerous: humans do it because they are selfish and reckless, and someone like Tesla will make a driving AI that fulfills their selfish demands. There is absolutely no reason to think that a speeding Waymo would be safer than a speeding human.
Studies like this makes me worry about a medium-term where driving is more dangerous for everyone: lawful drivers get in more accidents because they're using sub-human AI, whereas nothing changes for unlawful drivers because they are using dangerous AI, turning off the AI and driving drunk, etc.
It depends what you want to know. "Are autonomous cars better than good drivers?", or "am I safer in Waymo than a random human driven taxi?". Both are very valid questions.
But "random human-driven taxi" is arbitrary since random humans are not driving taxis. This study doesn't at all show that Waymos are safer than Ubers driven by Uber drivers (though I suspect that is the case).
Why? They are measuring an accident rate per mile driven. Assuming that miles driven is a constant (which I'm actually not sure about), they show a reduction of accidents to be expected if all miles were driven by their cars. Which is the thing that society may care about?
> privately insured driver population that resides in these geographic regions
If I were to hide something fishy, I’d define the shared coverage to include zip codes with fewer human accidents:
* zip codes are relatively small, so you have a lot of variance per code due to randomness of rare events;
* Waymo mostly covers urban areas and nearby suburban areas, so you can claim suburban zones were only recently covered by Waymo or that the insurance had too few members in certain zip codes to include them.
I don’t think that study is wrong, but it is in Waymo’s interest to claim they have few accidents and in insurers’ interest to claim there have many, or rather more, accidents to raise their premiums.
>> “Including freeway driving makes this benchmark crash rate artificially lower, so, by including freeways in this study’s benchmark, the benchmark crash rate underestimates the true driving risk of where the Waymo ADS operates.” Page 26
Except of course if Waymo would crash a lot more in freeways, which we can't assume won't happen just because human drivers don't. Far from this making human drivers look better it makes the comparison unequal and for that reason, uninformative.
Half of those apples are orange and smell like citrus fruits.
Why? We aren't looking at introducing autonomous ambulances any time soon.
They should be compared to taxi/Uber/Lyft drivers that is what is being replaced by this and as such where the net harm increase or reduction will occur.
Why -> because if someone caused an accident I'd want to look them in the eyes and call them names; as opposed to calling a support number. If these vehicles were among the best drivers, at least I could live with it maybe. Emotions are real, you cannot always rationalize them away.
Average drivers sometimes end up in jail for their mistakes. If Waymo made those same mistakes, guess what will happen.
I'm not sure it's possible to solve for that with additional prowess.
If you are wronged by a machine you are always going to be emotional about it, i.e having your email account unceremoniously terminated by Google without recourse by an essentially automated process.
The thing to solve for is the lack of recourse. Very very difficult in American society given how skewed the relationship is between corporations and people (or the government for that matter).
If instead when a Waymo demonstratively caused an accident the payout was 10-100x higher than if a human was involved I suspect many peoples qualms would go away.
While I think just the "knowledge" (assuming Americans would accept facts...) that the driver is X% better than the best drivers on the roads would provide little comfort to most.
I think in societies outside of America these may be more successful actually. i.e most parts of Asia.
You just need to know the injury claims of those drivers vs the average to compare.
But I agree, if I were to trust such a system, I'd want it to be a better driver than me, not than average. (And like everyone, I believe I'm significantly above average...)
> Why? As long as it is better than the median driver the roads are safer.
The intent is to increase miles driven, and the drivers a commercial service can be expected to replace first are professionals, not "median drivers". In practice accident stats are also benchmarked against mean drivers in a world in which serious accidents disproportionately involve inadequate drivers (or conditions Waymo does not deploy in)
Ambulance drivers might be a step too far unless autonomous systems are expected to drive ambulances at high speed and run red lights. But my point is that they really do need to be a lot better than the median driver given their real world operational expectations, and especially significantly better than the mean driver they're being benchmarked against here, who probably loses their license for the bodily injuries they cause...
Well Uber drivers don’t have any special qualifications. That’s kinda the point of Uber. Is there any reason to expect them to be safer than the average driver?
I think we both agree Uber drivers aren't the worst drivers on the road but your assertion is that Uber drivers are professionals and so somehow better drivers than the average. But they aren't. They're actually average drivers. So a self-driving car actually has a very low bar to clear.
fake, fake, lying, no conection to the reality of
what is needed for 90% of short trips on roads.
millions of trips an hour require going into tight, poorly lit, uneven , and unmaped ,end destinations, that are where a huge amount of the included "bodily injury" is happening, and waymo
is never going into.
Add in other realitys like, heavy acqward baggage and cargo, tools, food, kids,dogs, etc,etc that waymo (wayno) aint carrying and there is no comparison to be made.
curb to curb, nicy nice areas, if you dont mind going slow inside a robot, by yourself, and pay through the nose, then sure, whatever
but WHEN, one of these things runs over someones kid, expect a backlash , right
Since you are completely removing at least 1 human from potential bodily injury claims, wouldn't you need a greater than 100% reduction in bodily injury claims in order for this to be better?
I mean when you are literally removing 1 person from possible injury, you literally can't have the same number of bodily injury claims. It can only be less.
are you? somewhere around 60-80% of miles driven are people alone in their vehicles. aside from the car driving itself somewhere, when there's a passenger in there that's the same as your average car.
This is covered in the paper, although, somewhat ironically, opposite the bias you're implying:
>The garaging zip code of the insured vehicle was used as a proxy for the city (Phoenix, San Francisco, Los Angeles, Austin) in which the vehicle drives. Waymo also almost exclusively operated on surface streets (non access-controlled freeways) with a unique distribution of driving that is representative of a ride-hailing fleet. In contrast, the benchmark represents the privately insured driver population that resides in these geographic regions. The associated benchmark mileage has more freeway driving than the Waymo ADS. There are several considerations when examining these results with respect to this limitation. First, freeway driving has a lower crash rate (Scanlon et al., 2024a). Including freeway driving makes this benchmark crash rate artificially lower, so, by including freeways in this study’s benchmark, the benchmark crash rate underestimates the true driving risk of where the Waymo ADS operates. Second, driving outside of these denser urban areas that the Waymo ADS operates would likely represent a reduction in overall relative crash risk. For example, commuters from the city would likely experience a reduced crash risk as they travel to less densely populated areas (Chen et al., 2024). Previous studies have shown that most injury collisions occur within a small radius from
residency, and that American drivers rarely travel far from their place of residence, with approximately 80% of one-way household trips being less than 10 miles (DOE, 2022). Third, the benchmark drivers garaged in the Waymo deployment area are not operating with the same distribution of mileage within the geographical limits as the Waymo ADS. Chen et al. (2024) explored the effect of Waymo’s driving distribution on benchmark crash risk and found that - should the benchmark driving distribution match Waymo’s in San Francisco, Phoenix, and
Los Angeles - the benchmark police-reported crash rates would have been between 14% and 38% higher. Due to all three of these limitations being expected to artificially suppress the benchmark crash rate (underestimation), the benchmarking results in this study are considered to be conservative. Surely, there is an opportunity in future work to leverage new data, such as insurance telematics, to more precisely define and leverage the benchmark driving exposure data to better account for this potential confounder.
If you had read the paper you would realize that it is not about the severity of the injuries, i.e. "how grave" they are, but the number of insurance claims, of which there have only been two and the second one involved a human driver running a red light.
It's not just a mere "reduction" as the percentages imply. The number of injuries is almost zero.
Remember kids: everything that can, eventually, will be turned against you!
Some fun predictions, gathered from other comments + my own:
- You ll have to dispute a dead family member against a faceless company that conveniently deleted crucial logs.
- They will make you wear wireless tags every time you cross a street, and fine & shame you, if you don't.
-Tags will contain your real identity, so that you ll get a bonus for your medical insurance.
- Insurance companies will refuse to insure your ild mechanical car, and will force you to buy a subscription for the new shiny car, that will have to be replaced every N years.
- The new car will refuse to drive you at X location because you overused your monthly C02 quota.
> - You ll have to dispute a dead family member against a faceless company that conveniently deleted crucial logs.
On the contrary, of all my negative experiences biking in NYC over 10 years, the _only_ time I was able to hold a driver accountable was when they were riding a Revel scooter in which case the "faceless company" took them off their platform. This comment seems to significantly overestimate how easy it to hold human drivers accountable.
>- You ll have to dispute a dead family member against a faceless company that conveniently deleted crucial logs.
What does this have to do with self-driving cars? What type of "crucial logs" do today's car even keep?
>- They will make you wear wireless tags every time you cross a street, and fine & shame you, if you don't.
>-Tags will contain your real identity, so that you ll get a bonus for your medical insurance.
1. people already carry "wireless tags" everywhere they go. It's called their phone. They don't seem to mind.
2. Why would they do this, when they spent billions into developing lidar? If anything mandating this is bad for them, because it weakens their moat against incumbents, who can develop self-driving systems that are cheaper because they don't need expensive lidar technology to avoid pedestrians.
>- Insurance companies will refuse to insure your ild mechanical car, and will force you to buy a subscription for the new shiny car, that will have to be replaced every N years.
1. Insurance companies can theoretically do this today. Why aren't they doing it? Why would they suddenly do it with self driving cars?
2. All of this seems... fine? If you're driving a deathtrap that's a hazard to you and others, and the insurance company is on the hook for the millions of damage you can inflict (if you severely injure someone), they should have the right to refuse coverage. I doubt they'll actually refuse insuring such cars though, it'll just be really expensive.
>- The new car will refuse to drive you at X location because you overused your monthly C02 quota.
It wasn’t nearly as cautious or timid as I expected it to be. But it also wasn’t reckless. I would describe it as assertive, but I’m sure some people would call it aggressive.
It knows the rules and what’s going on around and acts accordingly. This really should be the norm.
For the first time in a decade I’m excited about an emerging technology. The future is bright.
reply