I can’t wait for wheelchair accessible vehicles to be equipped with this technology. It looks like they’re using Chrysler Pacificas which already have wheelchair ramp conversions from VMI and Braun. Since the self driving equipment seems to be on the top of the vehicle, it seems trivial to mount it on a converted Pacifica. As a quadriplegic who doesn’t drive (yet), the possibilities with this excite me greatly.
I'm sure others are working on this as well but Panasonic recently had an exposition in Tokyo and one of the things they showed was a self driving vehicle platform. The vehicle was basically just a base. The motors, wheels, battery. It about as high as the wheels only. Then, depending on what the customer orders there were different tops that could be snapped on. A camper top, a cargo carrying top, a food stand top, a passenger top, a billboard top, ...
Sadly, one big barrier to this is that a lot of regulations prevent them from charging more money for this than for a regular ride in many jurisdictions. It’s obviously more expensive to provide so many companies don’t bother. An Uber is almost never accessible for a wheelchair van.
In Vancouver Canada, the government mandates taxi companies have x% accessible vehicles. I don't see why they can't do the same to Waymo, since they own the vehicles (note we still don't have Uber here.)
Rear entry is probably going to have physical fitment issues depending on the amount of space in the chassis to work with. Minivans are pretty homogeneous now (it's not the late 80s anymore) so this isn't really an issue but I bet there's a model or two out there that are simply a no-go because some component that can't practically be relocated is in the way.
Side entry is going to be able to fit a wider range of vehicles because the installation is less invasive so you can support more vehicles for cheaper (kind of like how a static webpage can be viewed on basically anything but a fancy web app has a narrower range of browsers).
You can get a rear entry kit for pretty much every minivan that's on the market. You can get a side entry kit for that and more. I didn't bother checking but I know I've seen side entry but never rear entry kits on full size vans (also it makes sense that you can't put a rear entry kit on a full size van because there's a frame in the way).
Fundamentally you're limited by head room. You can't have people smashing their faces into the top of the door opening.
Google's choice was probably not influenced by the availability wheelchair van conversion kits.
The electronics for the self-driving are apparently in the trunk (the cars have a sticker on the back that says rear storage is not available) so a rear entry is not likely to happen.
In terms of minivans, most American makes and models have conversions from at least one of the conversion manufacturers. Braun even has a converted Ford Explorer. The one issue I can see is that Waymo is using Pacifica Hybrids which can't be converted because the floor of converted vans is lowered ~12-14 inches which gets in the way of the batteries. But I can't imagine it would hurt to use a handful of gas-powered vehicles.
Could significantly impact the paratransit industry. I know those to be government subsidized here in the northeast. If those subsidies were offered to self-driving vehicles, it could help a startup not only establish itself but also begin to compete with Uber/Waymo giants.
Which got me recalling, I believe there is a self-driving startup dedicated to helping the elderly get around.
It would be huge for the wheelchair using community because while there's UberWAV and LyftAssist, those are contingent on someone owning a wheelchair accessible vehicle to start, which is rare. Uber is starting to partner with local paratransit companies but rollout is slow and doesn't offer great availability. A big advantage I can see with wheelchair accessible Waymos is that once they're comfortable running them without supervision, a wheelchair user could call them at all hours. Right now it's really hard to request paratransit service after-hours so my social life is pretty much restricted to daytime hours unless I have someone to drive my converted van.
Keeping the safety drivers in the beginning, scaling up, and getting a year or more worth of actual real world use, before slowly scaling back the drivers is what I'd call a very responsible choice.
IMHO, to do less would be reckless showboating that could do unimaginable damage to the AV industry, as well as end up endangering lives.
In the tech industry, we don't like to wait, we like moving fast, taking risks, shipping MVPs, learning by shipping, etc. And for the next great Camera filter app, perhaps that makes sense, but not for cars, planes, or medical equipment.
"After following the sedan for about seven miles trying to get him to respond to lights and sirens, the officers guessed that the Tesla’s semi-autonomous Autopilot driver assist system was engaged, so they pulled in front of it and started to slow down, according to a police report.
The sedan did the same, and the two vehicles safely came to a complete stop in the middle of the highway. With some difficulty, police roused the driver, who was identified as Los Altos Planning Commission chair ..."
>> Keeping the safety drivers in the beginning, scaling up, and getting a year or more worth of actual real world use, before slowly scaling back the drivers is what I'd call a very responsible choice.
I'm sorry but where does this "one year" expectation come from? What says that Waymo needs a year (or two years, or ten, or three months, etc) to start "scaling back the drivers"?
From what I can tell, the only way to judge the competence of their driverless systems is to compare where they are today with were they were 10 years ago when they started out. However, this is next to impossible- that data is not public.
So, where does that "one year" optimism come from? Autonomous driving won't magickally become easier in a year from now.
No, the only way to know how well it's doing is for it to get enough "interesting" miles, and then compare the success rate of those miles to similar human-driven miles. Right now Waymo is past 8 million miles driven. Miles driven is a better and normalized metric, with comparable real world statistics, unlike time.
Miles driven is a less valuable number if the miles are largely along "easy" paths. What interests me is challenging events, perhaps what you call "interesting miles" because the more of those that cars successfully navigate, the more real-world ready the autonomous vehicles will be. There are hundreds of different types of these situations that are common enough that these vehicles will have to react appropriately (construction, detours, officer using hand signals, missing or misleading road paint, inaccurate maps, etc.) I don't know where Google's cars have driven, so they may have encountered these things, but I don't know.
I did explicitly say "interesting" miles, but we shouldn't mix up Phoenix miles with New York miles or Canada miles. No one claims they are ready to run anywhere in the world, but they have enough miles in Phoenix that they're confident they've seen almost all interesting miles there.
Waymo is past 8 million miles total, how many miles on the latest version of the software? Certainly there are new features that are there in the last half than the first half? I think miles count is a bad metric and if it wasn't, Tesla would be the front runner.
All those miles are recorded, fuzzed and replayed in simulation, where they've driven over 5 billion miles. I assume they retest all the miles + fuzzed versions on every version release.
It’s past 10 million miles and it’s driving more than 500k mi per month.
If the last version of the software was in August that it’s more than 2 million miles by now.
The miles per month figure will only increase.
From what I can tell, the only way to judge the competence of their driverless systems is to compare where they are today with were they were 10 years ago
That’s hardly the only way to judge the competence of their driverless systems.
> the only way to judge the competence of their driverless systems is to compare where they are today with were they were 10 years ago when they started out.
How does a self-relative measure capture the absolute goodness of the vehicle?
I'm much better at skiing than I was two years ago, but I'm not ready for the Olympics. My history is irrelevant to my ability.
It's just my speculation, if after a year of a huge number of successfully completed rides without incident, they'll obviously be more confidence in the system.
My point is, the responsible thing for self driving companies to do is maintain safety drivers until confidence increases.
The real question about self driving cars though and safety isn't about how much practice these vehicles have, but when they get in an accident, are people inside them still safe. If the computer gets screwed up in an accident, could it end up accelerating randomly all over the place, pull itself to the side of the road and do a safety check of the people inside.
It's also very possible that there is a safer design for a car than having people sit right in front of huge glass panels.
Problems like these are easy to optimize out. For instance, you can put in fuses that disconnect the traction power or ignition in the event of a collision. If you’re worried about the computer getting scrambled, put in two CPU’s and compare the output - if they are different by a significant amount, then stop driving. Most of the prototypes I’ve seen also have a manual e-stop in the cabin.
As for getting rid of glass: safety glass is actually pretty darned safe. It pulverizes into shards in a collision. And, there are a lot of people, myself included, who will not be able to ride down the highway without a view out the front - otherwise, it will be barf-bag city.
"Keeping the safety drivers in the beginning, scaling up, and getting a year or more worth of actual real world use, before slowly scaling back the drivers is what I'd call a very responsible choice."
A whole year! Truly sage-like wisdom from the MoveFastBreakThings generation of technical development.
It is simultaneously comical, exasperating and sad to watch this tragedy unfold.
We will slowly build our own, modern Deuteronomy - instead of forbidding the consumption of shellfish or stipulating gifts for former slaves, ours will talk about agency and skin-in-the-game and aligned incentives ...
Meanwhile, there will be a multi year (decade ?) period of experimentation where we will all be terrified as we drive at highway speeds and another vehicle approaches that road at a right-angle. Currently, I know what that vehicle will do, because all other choices lead to death-of-driver.
But in this near (and short lived) future, that other vehicle could do anything.
That’s funny, a month ago my car was totaled by a guy who fell asleep and blew through a red light. That happened at a right angle but only at 30mph. Since then I’ve been concerned at every intersection about what the other cars will do.
I’m not sure how much of it can be blamed on cell phone use, but I swear that human drivers have been getting substantially worse over the last few years (at least in the Bay Area). It seems like every single week I avoid a major accident from somebody doing something stupid or not paying attention. And I don’t even drive that much!
So while Waymo continues improving, it appears that the bar for them to be better than human drivers also keeps getting lowered.
I think so much of it is just a scale issue. I grew up in a mid-sized suburb in the Northwest just a couple decades ago. Most road were only 2, or maximum 4 lanes. The idea of driving around without your headlights on in the dark would be unfathomable. And if you did, you'd get honked at, flashed, and presumably at some point within 5 or 10 minutes, a police officer would see you and stop you.
That area's gotten bigger, and I moved to the Bay Area, which is bigger still. 2 decades on, It's not even worth COUNTING the number of vehicles without headlights on in the dark or rain. I never thought I'd be one to complain about too few police, but I see at least 10-20x as many vehicles with no headlights as I see cops (for example). Much of my commuting is on the freeway, and CHP are few and far between but also police aren't so common on surface streets either. And when they are there, they don't bother pulling anyone over for "small potatoes" stuff like this. I don't know if they're waiting for the next violent crime, or what, but it would be nice if they did a little "broken windows" policing.
Why don't they bother? I don't know. Probably because everyone's busy, everyone has a sob story, their stupid new car with its light-up instruments on 24x7 makes it LOOK like the headlights are on, the daytime running lights make you THINK the headlights are on, are you really going to give a ticket to someone just trying to pick up their kid from school?
I wish the answer was YES, but it just feels like everyone's stopped knowing or caring. When you don't own the same vehicle for 15 years, you have no idea how the controls on this week's lease work, there are too many cars on the road, the problem has gotten too big, tech has taken over too much of it.. everything just feels out of control.
IIRC the #1 reason people get pulled over for drunk driving is because they don't have their headlights on, the police for sure are on the lookout for that.
In Arizona you pretty much have to make the cops do their job unless they think they can get you for a DUI then they'll pull you over for the slightest infraction. The things I've seen people do when there's a cop right next to me at an intersection and not get stopped is mind boggling.
Yeah, the problem is, in the winter, no headlights during the early evening hours, no big deal. 3am, they're out for blood if you don't have your lights on, for sure.
To add another anecdote, I thought I was the only one who thought drivers have gotten worse in the Bay Area over the past few years. The frequency of some driver not paying attention or not following the rules of the road and blowing through stop signs and red lights or abruptly turning into my path without a signal has been increasing, as a driver, cyclist, and pedestrian. Stay safe out there, especially during the winter months.
While that OP didn't put it eloquently we are still creating a system where it would be technically possible to centralise the control of almost every single car at once.
We're going to do it but as with so many things we might want to ask ourselves if its wise to automate so very much while we still haven't aced security.
I'm not sure where you live but in my country (USA, where these companies you are mocking operate) when a vehicle approaches the road and I am going highway speed, I get a bit nervous, cover the brake, and feel a nice sense of relaxing stress when I pass by. That other vehicle could do anything. 40,000 deaths a year and over a million world wide.
You have it exactly backwards. You will _know_ what the vehicle will do because it will be programmed to do it. Meanwhile, that other vehicle could do anything because the driver might be texting, might have fallen asleep, might be in a fit of rage.
Driving is an absurdly dangerous thing to do. We don't tolerate that level of risk in almost anything else in life. It just so happens that getting around at 60mph is so useful that we accept the risk. As soon as we have an alternative to human drivers, it will seem insane to keep risking as many lives as currently do on the highways.
> You will _know_ what the vehicle will do because it will be programmed to do it.
I don't share either the OP's confidence in the attentiveness of human drivers or yours in the existence of bug-free complex software
The biggest differences between the two is that (i) software bugs can be potentially be systematic across all vehicle types and sudden, whereas human failure modes are discrete (ii) we have pretty good data on human failure modes which set the mileage interval for catastrophic failure surprisingly high, even higher if excluding human failures linked to behaviours society has decided not to tolerate, such as DUI and road racing.
I don't have confidence in bug-free software, just in software that is bug-free enough to be better than the status quo. Systematic bugs can be fixed systematically. Humans will never not be too easily distracted, easily angered, easily bored. We're never going to fix ancient human characteristics so that driving stops being an unacceptable risk to life.
>But in this near (and short lived) future, that other vehicle could do anything.
It's not like we don't have this problem today: the other vehicle could be drunk driver, tired, on the phone, texting, fixing their GPS, tending the baby in the back, etc.
As a newer motorcycle rider in the past few years, I am even more ready for automated systems to augment human ability when it comes to driving to address problems of distraction, poor reflexes, among other things - and honestly would feel safer driving next to a self-driving Tesla than next to someone coming home from work on the highway, with only 4 hours of sleep, powered by their 5th cup of coffee, etc. etc.
Notably, "At the start, Waymo-trained drivers will be riding along to supervise our vehicles for riders’ comfort and convenience."
Edit: Re-reading my comment, it sounds negative, so I want to acknowledge that I think this is a big deal. No matter how "soft" this launch, it's still a historic moment.
The milestone to mark is the day when you or I can, at least theoretically, download the the fricken app and request a ride in an autonomous vehicle that has no safety driver.
I think that line is going to a bit more blurry than you think. Waymo has already done fully driverless testing during the early rider program, with plans to roll that out to Waymo One in the future[1]:
> We’re never done learning, and our early rider program will continue as a way for a select group to give us ongoing insights. They’ll help test early features before those new capabilities graduate to Waymo One. For example, we’ve already offered fully driverless rides to some of our early riders.
And the number of people with access to Waymo One is going to steadily increase over time[1]:
> We’ll first offer Waymo One to hundreds of early riders who have already been using our technology. Over time, we hope to make Waymo One available to even more members of the public as we add vehicles and drive in more places.
So with that in mind, let's imagine if, for example, at some point in the near future 10,000 people have access to Waymo One and 1% of Waymo One rides are being completed without a safety driver. Would that satisfy your milestone? What about 100,000 people with 10% driverless rides? If Waymo eventually reaches a point where 95% of rides are fully driverless, and 1 million people have access to the app, would Waymo opening the app up to everyone _really_ seem like that significant of a milestone anymore?
I've been following the industry so closely for so long that I have a bottle of champagne in the fridge to commemorate the day that a publicly accessible, fully driverless pilot project opens up somewhere in the world, and that bottle has been in the fridge for 3 years now.
However anticlimactic it may be when it arrives, that day is nonetheless the day when, in my mind, self driving cars have officially gone from zero to one. There will be no apprenticeship this year.
Assuming you are the same Fricken from reddit, it has been interesting watching your demeanor become more and more...realistic as time has gone on.
And it's still unclear to me if everybody is having their NDAs lifted, as they are making a distinction between early riders and Waymo One but not specifying if everybody is making the transition or just a handful of people (or hell, even any).
As always, time will tell - the rollout of this technology has been slow to those of us who have been following it over the years.
Right, the real milestone was when they set Waymo cars with no safety drivers on the streets a few months ago. This milestone realistically is just that the riders don't have an NDA anymore, unlike the previous ones. It's still a cool step forward, but it's still a "private beta". At least now we can start getting photos and videos from the users though which is cool.
That car would probably be called Windows Car (not to be confused with your car's windows), stop irregularly at weird places to allow for mandatory updates, and report your height and weight to Microsoft.
"Oh shit I'm late to the airport!" Get in car, car says, "Critical Updates required before start, please wait 5 minutes". Then 30 minutes goes by, update is at 99%, "Finalizing".
When its done, you say, "Ok car, get me to the airport ASAP!". Car says, "Sorry, why we optimize the download, speed is limited to 30mph, The ETA to your destination, the airport, is NaN"
Probably foolishly, I updated the firmware of my van (a Pacifica PHEV, same as Waymo is using) while driving a while back. Since the update was for the infotainment system, the vehicle continued operating just fine (other than temporarily losing access to settings such as climate control, automatic wipers, etc.).
Mark these words, "At the start", because we'll be hearing them for the next 10 years until Waymo closes up shop. That they are going to market with drivers in the car means that the business has pivoted from "can we do driverless taxi" to "can we make money as uber-but-with-employees".
The next announcement from Waymo will be that they are cutting staff "mostly from R&D". The announcement after that will be that they are opening up the Waymo app for anyone to be a driver. The announcement after that will be that they are closing up shop or being sold to Uber.
Mark these words. In 10 years time we'll still be hearing the same comments about Waymo & their competitors because there still will be certain combinations of passenger, route & conditions that Waymo will refuse to take without a safety driver along.
But since Waymo will handle the vast majority of rides we'll just tune out those comments.
Handling all rides & conditions is a very hard problem and it'll take decades to solve.
But scaling from "a few" to "most" is a much easier problem and one that I have confidence that Waymo will achieve. And "most" can be very useful & profitable.
I assume we're talking about city driving. I'm not sure about the USA, but being in Europe, the average speed my car says I go is 28.5 km/h. Do you think we could not solve traffic for 10 years if we slowed cars down to 30 km/h? With assumption being that a city-wide driverless system would easily 'swim' through the city without encountering red lights or slowdowns due to unpredictiveness of other drivers. We could even tolerate a little bit more errors because accidents would not be harmful too much.
I'm a big fan of Sweden's Project Zero, one leg of which is setting and enforcing a speed limit of 30 km/h on any road or street without fully separated pedestrian & cycle infrastructure. I don't know if self-driving will help accomplish this everywhere, but I'd be ecstatic if it did.
> We could even tolerate a little bit more errors because accidents would not be harmful too much.
instead of "harmful too much" you can simply use "fatal". 30km/h is a pretty magic value: below that speed pedestrian collisions are very rarely fatal. Serious perhaps, but not fatal. Above that speed they are often fatal.
The average human cannot sprint that fast. I'd peg the average person sprinting closer to half that (obviously there's a lot of variance, but I think only the top 1% of athletes could run 30 kph even for 5 seconds). For a frame of reference, try to run at the top speed of a treadmill (most max out at 15 mph, and most people can't go that fast - at least not at the gym).
Maybe if this was Microsoft, and Waymo was the Zune.
Alphabet has given it some 9 years just to get to this soft launch, I think they have the patience to see it through. There is enough money and interest in this space internationally that I'd give it some time.
The news would be different. This news item is focused around the driver and safety. They are slowly changing their messaging from being about self-driving to being about safe-driving.
They can't come out right away and say "guys we failed at self driving, it doesn't work", that would tank the personal careers of everyone important in the company. They have to slowly move the needle away from self driving, and then they'll give us the bomb drop, "we could only build better safety systems, not self-driving systems" and nobody will care because they are happy with safe-driving.
If they truly had self-driving tech, they'd be doubling down on that with their messaging. Not only to let everyone know there's a new sheriff in town (bye Uber!), but also to let their competitor's know to start their death clocks.
Maybe I haven't seen enough large-scale corporate failures (and the associated damage control, which is how it seems you are characterizing this post), but it feels like you're making this prediction on the basis of not enough data.
I can imagine there are other scenarios where Waymo PR makes this post, even though Waymo has not totally failed at producing self-driving cars, but perhaps I'm being insufficiently cynical.
Yes, I agree that I don't have enough data. But previous to this article, I was really supportive of Waymo. When I read the headline about "next steps", I thought we'd hear about some great new systems being rolled out or advancements towards the goal of self-driving, NOT about safety safety safety, drivers drivers drivers. This is how MBAs speak, not engineers.
Uber has poisoned the well with the type of reckless behaviour you are talking about. Self driving cars are no longer assumed to be safe by default. They kill people just like human driven cars, prrhaps they are even worse than humans at driving. So far Waymo has kept it's reputation as one of the few self driving companies that hasn't killed any of it's passengers.
Calling Wqymo an Uber with employees is a pretty big insult to the most advanced sdc company so far. Uber's self driving attempt has failed. Uber is nothing without it's drivers.
I don't buy this. Strategically, it makes way more sense to roll it out as a driver-assisted system. Instead of dealing with the negative PR associated with a mishap or accident (like Uber in Arizona), you can place blame on individual drivers while you gain adoption.
If Google believes its tech is that far ahead, it makes way more sense to minimize risks associated with the public perceiving Waymo's system is dangerous, and let people get comfortable with being in a self-driving car.
> They can't come out right away and say "guys we failed at self driving, it doesn't work", that would tank the personal careers of everyone important in the company.
That's exactly what would happen if a newly developed product killed some people. A human assisted rollout is a cautious strategy that allows them to learn in the real world without exposing their massive R&D investment to the fate of Uber's.
No, that's part of the strategy of shifting the messaging: Use both phrases together so people associate them, and then slowly drop the phrase that you want to go away.
Yeah, but the alternative bet is that they want the "safety driver" part to be the part that eventually goes away. If they want public acceptance, it will make sense to keep safety drivers.
Eventually they will go the way of the elevator operator.
As the article mentions they were always about safe driving and saving lives. I remember the founder talking about it ages ago. It's not really a change of messaging.
I agree strongly that the market for true self driving is a mirage. It's being sold as new free time in your vehicle, but it'll just end up more time for ads to saturate eyeballs.
Doesn't matter, it's not self-driving if you need a driver. And when that driver takes control of the vehicle away from the AI, it's going to be at a startling and dangerous moment when the AI makes a really bad decision, like directing the car towards a barrier or a stopped bus.
If the software driver is bad enough that the safety driver will feel compelled to pay attention at all times but not quite take control unless the car does something actually warranting it. It'll be like teaching a 16yo to drive but without them actually getting better (because the software will need to be tuned to keep the driver on edge). Doing it that way is kind of a backhanded approach but it would force the backup driver to pay attention.
One feature I like is the LCD displays for passengers, which (supposedly) show what the Waymo car is "thinking".
So if the car stops unexpectedly, the display might show, "waiting for a crossing pedestrian". Ideally with a simple map diagram of what's around the vehicle.
This is both really cool, and really smart strategically.
I'm sure the Waymo folks know that one of the largest barriers to success for automated vehicles is public perception and fear. Each time a rider rides in one of these, they get a beautifully rendered window into the car's brain, which helps build the trust needed to sell the technology.
> Waymo One is rolling out in the Metro Phoenix area first. Over time, we hope to expand to new places and more people.
You can sign up to express interest using the form at the bottom of this page[1] (asks for your gmail address, zip code and type of phone).
P.S. Why is the mobile experience on Medium so frustratingly bad? Between the permanent heading plus the "please download our app" heading I find it maddening. I also find the UX around reading comments to be unnecessarily confusing.
Don't worry, Medium on desktop is equally bad. Two persistent banners bottom and top plus a "pardon the interruption" popup. I really wish companies wouldn't use Medium as their blogging platform.
It's also started pushing entries into my browser history as I scroll past headings, stopping my back button from doing what I expect. Really gets my goat.
but do you know we have cookies? And now please, would you like some tracking before proceeding to marvel at our giant header and footer flanking 3KB of plaintext?
> Why is the mobile experience on Medium so frustratingly bad?
I believe they are trying to gain negative publicity, because you know, all publicity is good. you see such comments about every article (some of them by me)
I know this is a heated debate, but it's interesting to see the divergent approaches that are being taken to achieve self driving. In one corner, it's lidar-based, equipping vehicles produced by other companies, organically and incrementally growing the navigation technology.
In the other, it's video-based with now 100s of thousands of vehicles feeding data to a massive neural net.
Each has very well thought-out arguments for the approach they are taking, and I expect that both would ultimately 'work' by reducing the inherent risks of taking a car ride by a substantial amount (90%?). I think the winner will be the one who gets to that milestone soonest.
While I own a TM3 I have little faith in their full self driving solution and did not opt for that. I did take the enhanced auto pilot and recent additions show how much more work it will take.
An example would be that it now shows different vehicle types and even pedestrians near you yet when all surround objects are obviously not moving it still has them moving on the display. It still has issues with certain bridges and apparently it is hard coded disabled around certain bridges to the point there are videos of it not recognizing a merging semi.
with regards to Waymo's solution, I am concerned that it should have something in the driver's seat. I don't care if its a Total Recall "Jiffy Cab" type solution. not so much for the riders but for those who are also sharing the same road be that other vehicles or pedestrians. Something that clearly indicates it is under control.
It really doesn't seem like much of a heated debate. The entire industry aside from one company has decided to combine video+radar+LIDAR and the one company that isn't using LIDAR has shown no meaningful progress (in actual self-driving, not referring to driver assist tech) in years.
In your two camps, I'd be curious to see why someone would specifically choose just video based?
Ie, I'd expect (as a layman) neural nets to be able to consume many different data types. Why limit to just one? I wonder what is better about just video, rather than video+lidar+audio+x+y.
Hell, one of the most neat things to me was how Teslas can scan multiple cars ahead to know of non-visible traffic blockages/etc. Using just video seems like a step backwards. Though, if it can produce better AI, I suppose it's still forwards.
Well, the only person who appears to be stating that you can use video only is Tesla, and Elon Musk seems to be the only person stating that that's the right strategy. From what I've seen he basically says: If you focus on vision and solve it, you've got autonomous driving whereas if you use a kludge of different sensors none of them will be great and the final solution won't be great. As an engineer whose worked on similar projects this just seems down-right incorrect frankly- especially in safety critical systems. However, it's also worth noting that when he said that it was suggested an autonomous Tesla drive across the country would complete by EOY 2017[1]. It's ALSO worth noting that Tesla have really upsold this idea of 'Buy a Tesla now, and it'll magically become autonomous'[2]
Definitely entertaining to watch, I agree heartily. I don't agree that whoever gets there first will be the winner. As we see with Waymo one, it's quite probable that there will be a limited roll-out of features/locations, such that it'll be much like these electric scooters: available in some places, but not others. The roll-out should be much slower in winter months and in places with inclement weather, however I see no reason that warm/dry places couldn't expect to see a big roll-out in the next decade.
Phoenix is the beginning for Waymo, what about Tucson, Dallas, Houston, Miami, San Diego, etc? Whereas, with Tesla, we'll likely see features rolling out slowly over time, an incremental push towards FSD, with owners being responsible for safety monitoring their own vehicle.
In 2004-2007, DARPA had a few self driving contests in the desert. The documentary was an entertaining watch. IIRC, most of the cars didn't even make it to the end.
Would be fun to get another test with current technology to see how far we've come. Though self driving cars nowadays are more focused on safety, not speed.
Off road racing has come a long way since 2004. Go watch some old King of the Hammers races from the early 2000s and watch some from recent years.
Even if the AI goes off coarse it's going to be a lot harder to get into a in situations where "just point in the direction you want to go and pin the throttle" is not an acceptable solution in 2018 compared to 2004. The 2018 vehicle will be able to just ignore things that the 2004 vehicle had to actually navigate around.
We also have cars that are almost fully drive by wire in stock form which frees up a ton of resources for software and testing.
Basically in 2018 the vehicle side of things is a trivially solved problem and even with a "dumb" 2004 level AI at the wheel half the field won't DNF from mechanical failure
Curious; If you train a neural net to drive a car with X number of samples, and the car does something erratic, how do you determine which data item introduced the erratic behavior and how do you go about untraining that behavior?
Gradient descent is done in batches and over the aggregate collection of the training data so it's pretty lossy. Usually the error is more of the fault of the learned model itself, and then the culprit is attributed to neuron "mis"-activations post-hoc. Often, problems take the form of generalization error instead. And yes, all of these problems are still quite unsolved, which is why we don't have end-to-end neural network-based self-driving cars, we just have rule based systems with a perception framework that just uses convnets for vision.
I think there's this popular perception that you just stream in lots of data from road vehicles and the driving just gets better over time--rather than continuing to discover individual edge cases more or less manually and updating rules based on those.
It's simple, you retrain the model several times removing one sample at a time (leave-one-out cross-validation), until the erratic behavior disappears. You repeat until there are no more instances of erratic behavior. Sorry couldn't resist the attempt at humor, but also consider it as an argument by contradiction that what you are asking is not possible. It's like asking which lesson caused me to hit a forehand over the tramlines. The answer is the one I didn't have time for.
That is one reason they don't do that. (Another is that the net would be useless if you upgraded the cameras...)
Whatever the sensing technology, everyone uses it to build a "state of the world" model which is then fed into algorithms for planning movement. Nobody (to my knowledge) just wires up input pixel neurons to output pedal neurons.
"At first, Waymo-trained drivers will supervise our Waymo One vehicles."
That's about where they've been for years now - almost autonomous, with a safety driver. Back in November, Waymo started sending some cars out without a safety driver. But they backed off on that.[1]
I'm disappointed. I thought Waymo was ready to launch a real self-driving system. But no, not yet.
Arstechnica[1] asked about it and Waymo did not deny:
> In fact, last week the Information's Amir Efrati reported that Waymo may actually have moved backward recently.
> "Within the past month or so, due to concerns about safety, the Alphabet company put so-called safety drivers back behind the wheel of its most advanced prototypes, ending a year-long period in which those people generally sat in the passenger or back seat," Efrati wrote.
> We asked Waymo about this, and a spokesperson dismissed the report as nothing new, saying that the company regularly changes the mix of cars with and without safety drivers.
“Fares are based on time and distance, and customers can expect fares to be roughly on par with what you'd pay for an Uber or Lyft trip—perhaps even a bit lower. The above Waymo-provided screenshot shows a customer booking this trip, which is 4.6 miles long and takes about 12 minutes. Waymo charges $7.32 for the trip. I punched the same route into Lyft and Uber aps on Tuesday afternoon and got quotes of $8.29 and $9.38, respectively.”
I’m surprised to see, after what happened to Uber — the failure of their safety driver to stay engaged enough to stop the car before striking a pedestrian — that a safety driver who will be mostly disengaged from the driving experience is still thought to add an appropriate safety margin?
I'm surprised to see them use a location for the article's main photo that's right down the street from where Uber's robocar hit that lady. The bridge in the background is the actual street where it happened and the accident site is not very far from there, couple hundred meters maybe?
I'm confused as to what they're actually launching here. If i'm reading correctly, their existing service which was branded as an "early rider" program has been renamed to "waymo one", but the service hasn't actually changed at all and is still limited to the same region and to existing users of the early rider program?
They announced a change of brand a month ago. I guess adding “One” at the end of the existing name counts, and I like the implication that they want to communicate clearly on phases of the project but… Was it worth announcing a brand change? I feel like I’m missing something.
Yes, seems like this can be viewed as adding a sub-brand, not a brand change. Like Amazon Prime and Amazon AWS are both sub-brands of Amazon.
One could speculate they're planning ahead, making space for another brand, either a different variation of SDC as a service or some other type of product.
Maybe they'll have SDCs with multiple riders (like Uber Pool) and call it Waymo Share. Or maybe they'll have self-driving commuter buses (like Chariot) and call it Waymo Shuttle.
Once you've mastered SDCs, there are just so many applications that you almost expect to do multiple things with it. They could license to car manufacturers and call that WaymOEM, sell their own cars and call it Waymotors, do trucks and call it Waymo Freight, etc., etc.
It isn't public, randos in Chandler can't download the app and take a ride, they need to be approved as members of the easy rider program. Waymo has lifted their NDAs, so now they can talk openly about their experiences.
Absolutely — and given the potential global growth of the service, committing and delivering on a schedule is a promising sign. I was just puzzled by the added information that they would change the brand: they didn’t need to (and they haven’t, not really).
We still do not have much clarity on the price point. The screenshot (see gniv’s link) shows that a 24-min ride could cost around $8, with an apparent 20% rebate.
I suspect that they won’t try to under-cut Uber too dramatically but the long-term costs are bound to be much lower if you don’t need to pay a driver. That illustrative price seems to match this. They certainly have copied Uber’s focus on identifying the best pick-up and drop-off based on many factors, notably road setup. I’m less sure about child seats: their copy and photo seem to imply that there are infant raisers. However, it would be unusual to have those in every car. I’m not seeing a way to indicate that you want one in the interface, and presumably, that’s a request per journey, not an account permanent status.
People have suggested advertising but I’m not seeing any of it in the screenshot, nor do I expect to see it any time soon, until their model has imposed itself.
I suspect that in the long run they will have a combination of driver assisted and driverless car. If you set a pickup and destination point, Waymo (and other competitors) can calculate the difficulty of the drive. If it's an easy route...send a driverless car. Difficult? Driver assisted.
The driveless car will average out with a lower cost. As a consumer you're guaranteed a ride somewhere and know it will be cheaper than Uber.
I'm optimistic in driveless car, but we need to be realistic that it wont be solved 100% in the near term...and it doesn't need to be for the taxi service to be profitable.
I expect the funding model for self driving taxis will be like validated parking. You ride free (or reduced) if you spend $x at whatever business, shopping mall, etc.
Because if I'm Safeway and I can get elderly people to come to my business (for shopping, prescriptions, etc.) they won't be shopping somewhere else. The business gains/retains the clientele by providing the driving service.
Just fund a minibus from the retirement home. I think many people do that already?
Ok, so it’s an interesting idea to use personal transit, but it could be done already without self driving vehicles, it might be economical. But it would be pretty niche.
I think you're bang on - but it needn't be limited to driverless cars.
Honestly, I'm surprised it's not something that Google have already built heavily into Maps. People often search for what they want (e.g. burger, pizza) - this would be an obvious time to offer deals or discounts to drive customers to particular businesses. You then expand to compete with particular retailers - someone searches for Nike specifically, they get offers from the nearby Adidas store.
This is not really any different to what happens with Google searches, where advertisers can buy keywords related to their competitors.
I hope they roll this out in Miami soon. Miami has some of the laxest AV regulations in the country. Also would happily pay to not drive in this mess every day.
As excited as I am for self driving, it makes me sad that I get car/motion sick. I can't exploit self driving to do laptop work, reading, gaming, etc without getting sick - all I can do is sleep basically lol.
It will be bittersweet in a self driving car realizing that since I don't have to pay attention, I can do so much more with my time!.. and then promptly stare out the window for an hour.
Not sure how common this is, but I was able to improve my resistance to motion sickness by incrementally doing more reading (interspersed with breaks) over a period of a couple of years that I commuted by bus.
As someone who has experienced both, it is the sudden unexpected movements that cause motion sickness. And if I am tired, my mind cannot coordinate between the movement I see to the movement I feel.
And now I realize that in the first case, it is a personality issue where if I am stubborn/stiff/unyielding, I will get annoyed at unexpected movements. I find this to be true also with my wife :D
Possibly, but not entirely. I can't even spin around in a chair moderately fast for one full rotation without regretting it. Eyes open or closed don't seem to help.
In cars it's largely about how much I can see outside, if I'm facing the direction of movement, etc. Usually in common vehicles I do not suffer unless I read/phone/etc. However in a bus-limo thing on a work retreat we all sat sideways.. on windy roads.. I could still see out the big windows, and tried to face my body forward but it was a rough, rough ride regardless.
I've never investigated it much, as it's such an awful feeling I have a hard time debugging it.
I could be wrong, but a city with daily monsoon-like downpours part of the year and famously aggressive human drivers is probably not going to be an early adopter for AV tech.
Ford and Argo have a huge facility here already. Their theory if they can make it work in Miami with our terrible drivers, they can make it work anywhere. You can already get postmates and dominos delivered by Ford AV.
I have read a little bit about this topic at https://www.lemberglaw.com/self-driving-autonomous-car-accid.... I think it's interesting to talk about this future tech. However, I personally won't risk myself to ride inside one of these prototype. It needs to be 99% safe before I would by one of it.
