From my discussions with a friend with intimate knowledge of the software side of Cruise, there is wild skepticism within the co that they even have the hardware needed to go L4 or L5. Big promises being made to GM by Kyle which is leading to smoke and mirrors to distract/buy time from more critical problems "under the hood."
I've been speaking to friends working at different companies across the self-driving industry, and almost all of them have their fair share of skepticism of L4 or L5, including those who are actively researching in the field.
The main reason for this is that current self-driving tech has plateaued with most companies going with the redundancy based approach (let's have 3 LIDARs instead of 1, and make that 4 IR sensors on each side instead of 2 etc.) simply because there is no revolutionary new tech that is a panacea to current L1 problems.
Sadly, explaining this to investors is next to impossible, so everyone keeps selling the same idea with a few more whistles as the next big thing while everyone is waiting for the big tech breakthrough that makes actual self-driving cars possible.
Until then, expect many more controlled environment (eg. airports or freeways where traffic is mostly smooth flowing, or roads with perfectly marked stop signs and pavement paint).
I recently read about Luminar (https://www.luminartech.com/) and their focus on making affordable, accurate LIDAR tech. Wonder if there are other components like this that need to be optimized and combined to get to L4 and L5
They say they are putting them on the roads of CA in a matter of weeks. I believe miles driven and disengagements must be reported around the end of 2017. So it should be easy enough to determine is four months or so if they're doing significant autonomous miles.
hmm I thought the bottleneck would be on the software side? I saw at least 2 lidars (possibly 3) on the car in the article. Of course we don't know if that's sufficient for L4, L5 but other companies are running with only 1 lidar.
People can get by with zero LIDARs. There's no reason why autonomous cars couldn't. The real problems are elsewhere (emergency vehicles, ambiguous situations, construction, …). We'll need operators on standby for a looong time still. Good luck driving in areas without decent fast cell networking.
Cruise Automation can produce a good demo. Shortcuts have been taken to produce good-looking results.
Don't be deceived. No car running the current software (or any software derived from it) will ever operate reliably.
Posted below, you will find a deleted comment from one of my fellow Cruise employees. That comment is accurate in its characterization of both Kyle and Cruise's technology.
Here is the deleted post from below, repeated verbatim. I'll put it here so it doesn't get lost.
QUOTE:
"Cruise employee here. throwaway for obvious reasons. It needs to be said that this is entirely marketing smoke and mirrors. The Cruise platform is significantly behind most other players in the market, and is having many many technical problems because of shortsighted leadership on Kyle's part, including high level people quitting (the head of planning and controls left last month because of Kyle, and more are threatening to quit) The reason I'm saying this is that this sort of public "everything is fine, we are the best" posturing leaks inwards. The opposite is very much true. Employees are very upset with Kyle because of things just like this. I make no exaggeration when I say Kyle is a mean-spirited, selfish person in private, and is very quick to publicly take credit for the work of others that he has emotionally and verbally abused. It's truly one of the most toxic environments I've ever been a part of."
I don't know what to believe. A few days ago I spent some time checking publicly available data and the result is that the current state of art is extremely far from an autonomous car that, could like used to drive a taxi in a busy city like Rome. Not far as "still some effort needed" but potentially "not clear when this could be available" far. Now there are those claims made by multiple companies... Go figure.
Based on my own observations, there's probably a big difference between operating as a taxi in Berlin and Rome, and another between operating in Rome and Mumbai. This only factors in other human drivers and infrastructure design/signage, not weather conditions (a different can of worms).
I suspect that we'll see a progression like:
1) Well-regulated urban and highway environments
2) Well-regulated rural environments
3) Moderately regulated urban and rural environments
4) Poorly regulated rural environments
5) Poorly regulated urban environments
And I think we'll probably be stuck on #1 for a while. I would also venture that most companies claiming to "be there" or "almost there" are referring to #1 as well.
I also think that it's possible that politicians may try to legislate a transition up that list as that first case starts taking off (greater traffic enforcement, better signage/infrastructure, more rigorous driver training, etc).
What are the special situations where 1 gives way to 5? The well-regulated part would imply to me that the possibility of slipping to another situation are eliminated or not allowed.
Blizzard. Heavy rainfall or wind. Sudden crowd / riot / chaotic event.
Pretty sure you can't legislate those out of existence. And sadly I think the fact that these vehicles are being developed and tested in Silicon Valley is either a cause or an symptom of their probably inability to deal with those kinds of conditions.
Erm, china has more than a dozen well funded self driving efforts, some of them by very big companies (baidu). Those are being tested in places like Beijing, you never hear of them because you aren't looking outside of the USA.
I wasn't suggesting legislating the conditions out of existence. I was thinking that kind of situation could be considered in the regulation. Much like VFR flight conditions a self-driving car would not operate in heavy rain/wind or low temperatures.
Road spills, random obstacles and animals all have to be dealt with for any of this to work at all. Construction can be considered in the regulations and self-driving cars can just be restricted from auto-drive mode in construction areas.
The thing is, for car driving you need combinatorial search combined with a good model (statistical) of the world. Research I've seen does not incorporate any kind of meaningful search.
Company that has all the roads and surroundings scanned and knows exactly where the car is at all times is at a huge advantage.
Partially observed world gets your head hurting. Not modelling other cars practically makes everything useless.
If you can't hallucinate everything around you as close as possible, the problem gets very very difficult.
I'm pretty sure there are companies close to solving the problem, as they have more data and some also know practically every road in the modern world.
Modelling and tracking other cars is a bigger issue, including pedestrians.
> Modelling and tracking other cars is a bigger issue
Aren't there very reliable ways to model other cars? Something as big as a car can't move very unpredictably in a millisecond-to-millisecond basis. And at the end of the day, all you need to do is be able to model the behavior of your car.
If car in front of me starts to swerve over on a collision course with my car, my car can note the probability of a collision a long (computer-long) time before it really needs to do anything about it. Knowing full-well its own braking, swerving and acceleration capabilities at any given speed, it should be able to know when it _must_ take action to avoid the actions of another, and it doesn't even have to rely on reaction time.
No, I think the real problem here is for cars to do things like understand _context_ of a situation, like, for instance, how can it tell that there's a bus in front of me turning around that's going to be clear of my way in a moment vs. a tree that's been felled? That's something that computers are and will probably continue to be terrible at until they can closely mimic the capabilities of the human brain.
> Modelling and tracking other cars is a bigger issue, including pedestrians.
Unscented Kalman filters and particle filters are used for that - they are pretty good if you sample environment with >10Hz rate. Most of what you said has been taken car of pretty well, those aren't really hot issues anymore. Also, the thinking these days is that humans (even mediocre) can do a good job driving using eyes only and more robust camera solutions based on deep learning are here. Their problem is inference speed as HW is not fast enough to allow FHD or 4K input feed to be processed realtime, so imagery has to be downscaled significantly, leading to errors. Similarly, tracking other cars only depends on how big supercomputer can you fit into your trunk.
That gets boring fast, if you go by the evidence the only player in the self-driving space is Google. Ironically they are the most restrained in public statements while there is a long list of pretenders announcing they'll ship in 2020 and already discussing far-fetched (entirely) philosophical scenarios, like the whole abortion of an discussion we had on "should the car hit pedestrians or save its driver".
The whole conversation is unfortunately poisoned by endless wishful thinking instead of discussing hard evidence. I mean there is still a large proportion of people on here who think Tesla will somehow leapfrog Google because they have all those cars out there streaming the raw video data from their umpteen cameras back to Tesla HQ, utilizing an imaginary mobile communications standard that leapfrogs LTE by two orders of magnitude that they invented alongside their LIDAR-less L5 autonomous car software in record time.
My personal skepticism on this subject has made me very unpopular at parties (well I don't actually go to parties, but you know) where lots of techy types go. Also earned me lots of downvotes here.
They're bulk uploads over home WiFi, and Tesla hired plenty of ML names like Andrej Karpathy. Having said that, AGI is the bottleneck for full self-driving.
OTOH it is easy to teach human drivers to behave carefully and differently when they encounter a self driving car. They could even require special marking or color on the cars. People learn to be careful with trains, trams etc so i don't see why they won't learn to behave more rationally when a SD car is passing by.
Trains and trams are fairly easy to predict, though. They can only really go in two directions, and change speeds very gradually. And for all that we've "learned to be careful" around them, they still rack up 2-3 thousand crashes every year, even with extensive signing and signaling.
There are a lot of big claims in this article. As someone who hasn't been paying much attention, is it credible that they have a near-ready-to-go self-driving car right now?
Evaluating technical capabilities in autonomy based on youtube demo videos is like evaluating the military potential by watching army parades on holiday.
The truth is, none of these companies is any close to full autonomy. Full autonomy _testing_ will begin once they deploy cars without backup drivers, to actually see how bad can things go without watchful human supervision.
Waymo had <100 situations where they required human involvement in tens of million miles they drove so far. They are probably already safer than human drivers and would likely cause fewer accidents. The issue there is a legal one - who is going to be responsible for a crash? The company that produced the driving software/hardware, or the owner of the car? Or even the passengers in case there will be city-wide car pools one could call when needed?
No. They had plenty more. Approximately 100 cases over 600k miles last year (their own data from California) which would have lead to a dangerous situation, and 9 cases which by their own admission would have resulted with a collision. Now it is hard to compare crashes of different magnitude, people often are engaged in fender benders, but only on average after 100 million miles driven there is a fatal accident, who knows how many of these 9 cases would have ended up in a fatality. But nevertheless, this is only 600k miles not 100m. Stop spreading misinformation.
Tesla is the leader leader in terms of released driver-assist technology. But they're years behind in the full-self-driving game. They're no way near where Waymo is now (and most likely Cruise, Drive.ai, maybe others).
Their in-house developed Autopilot v2 is still worse than the third-party (MobileEye) Autopilot v1 they got rid of last year.
Yep, they're pretty close. Late this year or 2018 they'll have cars driving around empty. GM has consistently been underestimated by the public, but they've played their cards very well. It's in no small part due to the drive and leadership of Kyle Vogt, who is on his way to the big leagues to join the ranks of Musk, Bezos, and Jobs and those guys.
Once they actually have cars driving empty we will actually learn how close or how far are they. I've seen a lot of PR muscle flexing from big companies before and no longer believe any such stuff. They do that to impress investors and the public, but this has nothing to do with reality.
Your comments sound like a PR statement, what goods? How do you know how close they are?
Looks to me that they designed a mass-producible car outfitted with self-driving systems (which is very cool!). But they are only just starting to do employee testing. Waymo has been doing that for years.
Waymo is ahead on the technology. They were ready to begin public trials with their Koala cars in 2015, but the California DMV shut them down, and many of their key people jumped ship.
Waymo has since been proceeding a bit more cautiously, and their efforts at this point are oriented more towards smoothing out ride feel, testing under adverse weather conditions, and their early rider program in Phoenix.
Whether it's Waymo or GM to officially be the first to have an empty car driving on public roads, the other won't be far behind, though the victory will be largely symbolic. GM's cars still need a lot of polish, and Waymo needs better manufacturing partnerships.
What's also interesting is that Cruise is hiring 'autonomous vehicle support advisors', which is for remote assistance. Not something you need to do if you plan to have a test driver in the front seat for year(s) to come:
> It's a prediction, I can't provide a source for events that haven't occurred.
Using the words "they will" makes it sound like an assertion. How are you so confident about "events that haven't occurred" will indeed occur? You listed out reports and demos, none of which necessarily suggest that "Late this year or 2018 they'll have cars driving around empty"? Do you have additional inside information that the rest of us don't have?
> Also, because Kyle Vogt said so, and he's never bullshitted me before, he's got credibility.
If I tell you the sun is going to rise tomorrow morning, I don't actually know it's going to do that until it does. That's kind of the thing about future events, it should go without saying. Nothing you've written is a counter argument. You just used way more words than you need to to say 'I don't believe you'. Yay.
My points are that GM is doing well, and that they are scaling up their operation to a degree that wouldn't make sense if they didn't intend to deliver the minimum viable commercial robotaxi service sometime over the next year or so.
The counterpoints are that 'this smells like PR bullshit', and that Kyle Vogt is an asshole (because no asshole has ever successfully led a technology company before).
You guys have nothing substantiative to argue with.
For comparison, it's about 28x higher[1] than Waymo reported for the same period. I'm not sure how similar the driving conditions were but I would guess pretty close.
That's their average for the year. They were way worse at the beginning, and way better by the end. They continue to improve, though I can't give you data on that until late December when this year's disengagement reports are filed. Like once you get to the point where you can go, say, 10 miles in urban traffic without a disengagement, the improvements from then on (as measured by disengagements) improve dramatically, up until you get to the point where it's really hard to discover new edge cases. Waymo in 2016 was at 1 disengagement every 5000 miles, and that's a 4x improvement over 2015.
Arizona. Their weather and regulatory climate has Waymo, Cruise, and Uber all testing there. I believe Michigan and Florida are also willing to permit fully autonomous cars on their roads, but most of the action is in and around Phoenix.
Volvos were supposed to be rolling in London and Gothenburg this year but it hasn't happened yet. They were supposed to have a member of the public who could snooze or read a book.
Wow. FYI - This is how you mess up content marketing. If the author genuinely wanted me to read the article and get to know the brand, he shouldn't have used such a scummy title. Yes, the article is /technically/ true. But they are only talking about the hardware, and acknowledge in the second paragraph that they don't have the software. So while the title implies self-driving cars are a solved problem, they are really only making claims about hardware. Cool I guess? But tesla probably already beat you to it?
Titles like these are what happen when marketers optimize everything towards clicks, and don't care about actual impact on the brand itself.
Saying that self driving car hardware problem is solved and all we need is software is like saying that one solved the problem of building interstellar space ship. All they need now is the warp drive. Otherwise it is all solved...
I didn't get that impression. Paragraph two says they have "the most advanced self-driving software ever demonstrated," which has been in use driving cars around in real city traffic on their prototypes or whatever (according to their youtube videos), but the article seems to be saying that they now have designed and built mass-producible cars that can run the software. (If I'm wrong, then yeah, not only the title but the whole article is misleading.)
Neither of those companies make cars.. The article is about actually putting a "self-driving" (however they define it) car through a proper manufacturing engineering environment to develop something that can scale to hundreds of thousands of cars per year.
Redesigning the software of a production car and actually integrating the tech at the factory-level gets you much closer to a mass-volume car than making some fancy software with a sensor package that you bolt to the top of a minivan.
Kinda seems like putting the cart before the horse though.
Despite what the mountain of "Autonomous vehicles are around the corner" PR articles say there's still a lot of problems to be worked out before we see any fully autonomous vehicles for sale.
And you need to work out your prototypes issues before you start mass producing something so that you mass produce something that you can actually sell.
Eh. I work at a digital marketing agency and plenty of brands are spending lots of money just so you hear the name and think that brand is good. Some never mention a product or a service, just pure branding.
So GM engineering got the hardware integrated with the vehicle and engineered it for assembly line production. That's what GM engineering does. Whether the Cruise software is good enough for self-driving without driver attention remains to be seen.
Volvo is doing well with their self-driving car, but admits the hardware cost is over $100K per car at present.
I don't really believe it is possible.
Lidars were the most expensive part and are much, much cheaper now, I can't really imagine how can they spend more than 100k$ of self driving hardware per car.
This article is filled with very carefully crafted language to leave the impression that this car 1) is being mass-produced, and is 2) capable of driving itself. But if you read it closely enough, it's clear that neither of these things is true.
I'm sure there's political pressure within GM for Cruise to announce these sorts of "wins" to keep management satisfied their money is being well-spent. But the breathless tone of this write-up conveys a sense of accomplishment that doesn't actually exist, and the failure to set reasonable expectations will just make the disillusionment that much stronger when it happens.
While I'm almost always bullish on new tech, self-driving cars are one where I frankly have a pretty hard time buying into the hype. I think when they really start to hit the streets, some issues are going to show up.
Here's what I mean. Imagine this situation:
You're pulling into your driveway in your self-driving car, and one of your kids left a lawn gnome there, blocking it. The car won't go forward because there's an obstacle.
Simultaneously, a driver looking at their phone and not paying attention comes careening down the road. They're clearly going to rear-end you, hard. You COULD just pull forward and out of the way.....but the car won't, because of the lawn gnome. It can't know the lawn gnome doesn't matter in comparison to the damage your car is about to suffer.
Imagine this flip scenario:
You're pulling into your driveway in your self-driving car, and one of your kids, a toddler, is in your way. The car won't go forward because there's an obstacle. Otherwise, same situation as above. Now, you ABSOLUTELY do NOT want to pull forward, because you'd rather get rear-ended than run over your child.
Even if you want to nitpick this exact scenario, surely you can think of a hundred similar contexts yourself, where driving involves understanding the value of different pieces of the car's immediate environment. A car cannot know the value of a child versus a lawn gnome, nor can it reliably detect the difference.
Even if such a situation only arises once in a thousand days of driving, with hundreds of millions of people on the road everyday, that's thousands of times per day. That's really bad.
Is it statistically better than the amount of harm done by human drivers? Possibly. But we all still get in cars believing we'll be better than other drivers. Who's going to get in a car when any given car has an equivalent chance of making a bad decision all on its own? It's just like getting in a car with a drunk driver.
Who's going to sign up for that? Not me, that's for sure.
I always question whether humans are really any better at this calculation. In the gnome example, many people will probably stop to avoid damaging their property, not see the car behind them careening forward, and end up getting rear ended anyway. In an ideal world, ethical issues like the one you presented would be major hurdles but humans aren't going to be perfectly rational in the moment of many accidents, especially the most dangerous high-speed ones when there is physically not enough time to do ethical math.
> we all still get in cars believing we'll be better than other drivers. Who's going to get in a car when any given car has an equivalent chance of making a bad decision all on its own?
Given the fact that plenty of people take taxis, rely on Uber and Lyft drivers they've never met, and take public transportation like aircraft and trains with little knowledge of the operator, this seems like a non-issue to me.
The other thing to bear in mind: I would guess that the vast majority of accidents are caused by people not paying attention -- distracted by a phone, person, great view, or even just because they're looking the wrong direction. Cars don't do any of those things. They will always be paying attention, looking in all directions at once.
So if they eliminate all of the accidents that are caused that way, but increase the number of garden-gnome related accidents...wouldn't that be worthwhile?
Another, less flippant way to say it: yes, we can imagine scenarios in which the number of accidents will increase. But what is the relative frequency of those types of accidents compared to run-of-the-mill rear-end, red-light run, lane change due to inattentiveness?
If given a choice between changing from what you are used to which has drawbacks, to something new that has drawbacks, even if those drawbacks are markedly milder, almost everyone will choose to stick with where they already are. Consciously choosing a position which has drawbacks that you know about makes you feel personally responsible for those when they occur. That is infinitely worse than suffering the drawbacks of the "default" you stay with which you can give yourself a moral pass on because it wasn't a situation you chose to begin with.
There's also the issue of what a self-driving car could do... and what they WILL do. Much technology COULD be amazing... if companies did not function in the ways that they actually do. Because of that, technology often fails to meet any of the promises made. Development is supposed to be cheap and fast, and safety and security get left at the door. No company is held liable for such things, and it would be expensive and bothersome to fix, so they're not about to bother.
And even if the number of accidents increases: Does that increase cause greater monetary loss than the considerable cost savings of not paying drivers.
There's no ethical question here. You either see that it's a gnome, or you see that it's a child.
Humans may or may not be less situationally aware overall, but we have the gnome question down cold. Cars don't. They see "obstacle in front of bumper radar."
The OPs point is well-taken because it illustrates a technical risk that few are discussing: the "last 1%" of self-driving software may be equivalent to solving AGI.
I do think the "last 1%" may be a very hard problem to solve, but the majority of crashes are simply people not seeing things or making bad decisions. "Recognition Error" and "Decision Error" are the two most frequent categories of accidents which basically boil down to not seeing obstacles while performing a maneuver (changing lanes and not seeing a vehicle in that lane, turning and failing to see an oncoming vehicle) and driving too fast respectively. ( see https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/...)
I don't see why the "last 1%" problem needs to be solved when we have hundreds of thousands of accidents every year caused by people not paying attention, not reacting in a timely fashion, and driving unsafely; all problems that computers are likely to excel at.
Because a half-baked product is dangerous (re: handover to human) and unsatisfying. You really want the version with no steering wheel. I’ll take the intermediate one, but let’s work toward the good one.
They are probabilistically modeled already. You keep tracking other objects and compute various possibilities for their movement and depending on the past and present behavior pick the most probable one. If car encounters lower-than-threshold probability, it can always slow down or stop, like what non-reckless humans would do. Even end-to-end deep learning can recognize construction zones and learn to drive around traffic cones.
The edge cases are endless (wheelchair woman chasing a chicken), and without a semantic world model the car will have no idea what to do. Low-level object avoidance architecture isn't enough.
Actually that case would be detected as two objects on a collision course and car would stop/slow down. Mind that sensors scan surroundings with high frequency and some of them can see "beyond the walls" like radar. Ultrasonic sensors can also detect the state of road surface (dry, wet, icy etc.).
Yes, but when can the car proceed, and in which direction? Helps to know what a wheelchair, a woman and a chicken are and how they behave. Helps to be able to speak with them.
Superhuman sensors will be great, but you really can't dodge AGI for true full self-driving.
None of these contrived scenarios compares to the rate at which humans simply fail to stop when they should have stopped. I don't know anyone who got into or was rescued from an accident by pulling one of these maneuvers. I know plenty of people who got into accidents because someone (them or someone else) failed to stop when they should have stopped.
Considering humans can barely stop when they should, I don't give them a high likelihood to also be able to have the situational awareness and make the multiple split second decisions required to both decide to, and take action upon, directing their car towards the least-valuable obstacle in an emergency. By the time they decide what to do they will probably be rear ended.
Edit: your examples also ignores that with self driving cars, people won't be put in these situations in the first place. A driver not paying attention to the road, looking at their phone would be... not a safety hazard.
As long as other cars (or humans, or animals...) are on the road, robot cars will face these scenarios.
The OP's scenario is contrived to make a clear example, but it happens regularly when you're driving: you're forced to make a choice between acting and possibly hitting something, or doing nothing and being hit.
We don't need self driving cars to solve the 80% case of preventing human drivers from hitting things. We need radar and auto-braking, which already exists. Full autonomy requires solving the harder problems.
There are 35 thousand deaths per year as a result of car crashes, and over 6 million total crashes[0]. If autonomous cars reduce those deaths to hundreds and overall crashes to thousands, I think you're underestimating the number of people that will sign up for that.
You're thinking about the tail end of the distribution. If self driving cars can solve 90% of the mass of the distribution, i.e. driving on highways and urban streets, it would be a significant improvement in terms of efficiency. Of course there will always be morale questions like the one you proposed. But quite frankly, the return on investment of solving the tail end scenarios like that is really small.
You can't ship without restricting where the car goes, punting edge cases to remote drivers, or actually solving the tail end with AGI. Actual full self-driving is far more satisfying than "you can go from here to the airport" or "you can drive in downtown SF."
But actually shipping a product that can operate autonomously in a subset of areas and conditions is more satisfying than not shipping a product at all.
The danger in this contrived scenario is still the human driver, who is distracted and careening down the road.
If all the cars were autonomous, it would more likely to be a matter of frustration than safety: the gnome causes a traffic jam because the cars won't endanger the passenger by driving into the opposing lane. Then what? They find an alternate route? They automatically call in a service crew to clear the road? Not the end of the world.
MIT (and I assume others) have collected data on these types of scenarios. For example, swerve and kill a bunch of people, or stay the course and kill a mom and baby?
Hopefully in a not too far future you and all the others that are convinced that are better drivers will be forcefully signed up so your scenario can't possibly occur given that the other car will be self driving and always paying attention.
There is nothing hard about valuing gnome vs child, if you can identify them. Computers can evaluate huge problem spaces in milliseconds. The hard part is the perception: identifying the child and the gnome. Deep learning has allowed us to make gigantic leaps in the quality of perception but there is a long way to go in uncontrolled environments.
I also think people incorrectly think that the moral choice dilemmas that self-driving cars will face are somehow the difficult part. What will happen is that we will provide valuation functions that are somewhat close to our intuitive collective understanding of value of life in different forms. Computers use them to evaluate these situations and we as a society just adapt to the new ethics of the AI age.
I will get excited about a self-driving car on the day that an auto manufacturer fully opens up details about their development process - size of team, experience levels of team, breakdown of salaries, tools they used, methodologies used, etc. I also want a PERSONAL statement from every single engineer who worked on the project saying that they take personal responsibility for the safety of the vehicle and that every single one of their own concerns about safety or foreshortened schedules or testing were addressed to their full satisfaction.
I also want to see the CEO, CTO, and other C-level execs of the auto company to release a legally binding agreement holding themselves personally liable for criminal negligence if the vehicle is later found to have obvious flaws - obvious determined by an independent group of several thousand unaffiliated volunteer experienced software engineers.
Right now, if a car company slapped a 'self driving' car together that was designed to seek and destroy pre-schoolers, no one would say boo. Certainly not the courts. Toyota killed people by having ludicrously egregiously bad development practices (their developers did not even have access to a bug tracker... they followed 6 of 94 industry standard recommended practices... and worse) - and the courts acquitted them of negligence. That is disgusting, and makes it impossible for any reasonable person to trust any auto company doing development of something as significant and difficult as making a self-driving car.
They know that they will not be held liable no matter how bad of a job they do - would you drive across a bridge built by bridge engineers hired, managed, and made to work the way software engineers are hired, managed, and made to work in modern companies?
Nope, I don't believe it for a second. There's no way these can run in the rain, or maybe even after dusk, and absolutely not off of extremely well researched and hand-picked routes. I'm so sick of companies trying to pretend like they've figured out level-5 self driving.
Wow. I asked in another post what people were ready to pay because of all seems like such expensive tech. To me it's worth maybe $2-5k in a new car (so 5-10%) because I usually don't mind driving. I could picture myself choosing between the sunroof/stereo and L5 autonomy...
How much will people be willing to pay for self driving cars? I mean it's nice feature but won't this always be very expensive tech also in the future? How expensive will it be to make this reliable in various regions (road signs, traffic laws)?
Imagine how much money is spent by car manufacturers now just validating the functionality of "simple" tech like locks and wipers at the testing grounds in northern Sweden and similar. All components of these systems will have to go through the same kinds of tests.
My dash display gets funny whenever it's below -20 in the car - I can only imagine what happens to all these sensors.
How cheap do you think this could be made? What price do you think people (consumers, not taxi operators) are ready to pay for decent autonomy in a car?
'self driving' really means the vehicle is being controlled by outside entities who have created technology that enables navigation.
This article is very opaque on how mature this is. The vehicles are the 'last mile' and from everything I've read we are a long way from Level 5 mass market maturity. everything else seems to require you to be able to reflexively grab the wheel and use the pedals at any moment if something goes wrong...
I noticed that that the sentiment about self-driving cars become much more pessimistic (or more realistic) in the last few months compared to the previous hype.
At this point, I'm not really going to be impressed until just ONE self-driving car company even talks about being able to handle snow and fog, preferably their cars knowing when they can't drive safely in the conditions.
(It's not just about seeing through moisture: it's also about traction, something you can't really use sensors to enhance.)
Cool. But sometimes the road is just plain impassable. And having grown up in many Iowa winter blizzards, I still don't trust even electronic systems. They help, but sometimes there's nothing you can do when there's basically no friction.
The safest thing might be for the car to pull over, put on hazard lights, and tell you to wait or call for assistance.
Humans deciding to go continue driving in such conditions is a mistake, leading to more collisions. Perhaps there will be a manual override, like "Unsafe conditions, voice password and waiver required to continue in manual mode; you might die or kill someone else."
Once you get into bad snow conditions pulling over is a really bad option. That said, this is a corner case I care less about than crossing Manhattan in rain at rush hour. Or really lots of tricky busy off-highway situations.
Self-driving car can go slower in those cases. It depends entirely on how beefy its CPU/GPU is to process a lot of sensor data realtime. The more messy situation is, the more time it needs, the slower it needs to drive.
Well, after a few cars get stuck (or crash) it will show up on the mapping system as blocked and the remaining cars will find another route. I'm semi-joking, but human drivers do that more than we would like to admit. Just needs a cheerful "please get out and push" message.
Or take a page from the internet playbook: Destroy the car and re-send from the origin.
Right? There is no better teacher than driving through a few brutal upstate New York winters to teach one how to drive in the elements. No WAY am I trusting a car more than my own abilities to do so.
I don't know if this is allowed - if not, mods delete this.
To the poster who had a throwaway and identified themselves as an employee of Cruise, if the environment you work in is as toxic as you claim, your best bet, right now, is to leave.
Don't stay a minute longer, as this kind of thing has a way of biting those who stay in the rear sooner rather than later.
I read the whole blog post, but not sure what's the difference between this car and the Tesla with 2.5 Autopilot hardware. Isn't that a production ready car that will give you autonomous L4/L5 "some day" ? Or is there some fundamental difference that I am missing out on ?
Whether or not this car drives itself it's still damn ugly.
Tesla is going about making electric cars sexy.
So you won't find me interested even if it sings too.
Never-the-less, for all the folks the that don't mind ugly cars it's good to have competition to gather all the data to prove that driverless cars are safer.
At the end of the day, the mass market does not really care whether a car is gas or electric.
The average Joe/Jane just has X dollars budgeted for a car and a mental model of what kind of amenities they ought to be able to get for X dollars.
Most of the "mass-market" electric cars I've seen in the $35,000 range (i3, Leaf, Bolt) look like $18,000 subcompacts. They just have electric motors inside.
You fundamentally cannot take an $18,000 car, slap an electric motor inside, and call it a $35,000 car.
If the car is going to cost $35,000, make it a $35,000 car.
Not that anyone believes that this car is really going to production anytime soon, but as an additional datapoint: the car in the first image is sporting $40,000 worth of lidar on the roof (5 x Velodyne VLP-16).
I'm sure GM is getting a healthy discount off retail from Velodyne, but production-ready lidar is still eye-wateringly expensive by OEM standards. Some solid-state lidar units (Quanergy) out there in the hundreds of dollars (and one from Velodyne in the works for 2018 release), but the fact that Cruise is still using spinning Velodyne units suggests to me that solid-state lidar needs a few more generations to come up to par.
That sensor assembly on the roof is ridiculous. What if I want to strap down anything to the roof, like a kayak, bike, luggage, or maybe a mattress? Good grief.
The number of times I've strapped anything to the roof of my vehicle (in 13 years of driving) is exactly zero. This particular vehicle might just not be suitable for your needs. Maybe eventually auto cars without roof sensors will come out, or strapping stuff to your roof will become as ridiculous as strapping something to the side of your vehicle.
>Electric self-driving cars will save millions of lives and significantly accelerate the world’s transition to sustainable energy, but only when they’re deployed in large numbers.
Just like cars that we drive, self driving cars enable sprawl. Sprawl will never be environmentally sustainable. We need cities that you move through by foot, bike, and public transit.
I agree with you on this. We need dense cities that don't depend on cars.
I will say that self-driving cars help a bit in getting people to bike, though. The biggest barrier to biking in San Francisco is people's fear of getting killed by cars (the SF Bike Coalition has done surveys on this). Self-driving cars should pose less of a risk to cyclists - they don't run red lights or speed, etc.
Another potential benefit would be reducing the need for parking. The biggest obstacle to physically separated bike lanes is the space they take up - typically that requires removing parking. If self-driving taxis are super cheap, that should lead to less car ownership, and possibly less demand for parking.
Why do people believe that public transportation does not enable sprawl?
A great commuter rail system is what allows people who work in the world's premiere cities to live in strip-mall exurbs and take the train to work.
I live in Washington, DC. Metro is what allows huge numbers of people to live in sprawling suburbs instead of in the city. Yes, some drive, but traffic is rough and would be unbearable without Metro. Fewer people would be able to live outside the dense city center.
Urban sprawl involves low-density areas. Metropolitan areas are going to spread out regardless. What cars enable is sprawl even when there is not enough density/economic output to make mass public transport viable.
Mass transit enables people to live in low-density areas who otherwise could not.
A massive proportion of the DC area population lives in the burbs, drives to a cheap/nearby parking garage at a transit center, and then takes mass transit into the city to work.
This lifestyle would be impossible without mass transit.
What makes you think that density/economic output are the main factors in how much public transit actually gets built?
EDIT: I actually love public transit, but I just don't see any evidence that it does anything but encourage sprawl just like cars do. Also, public mass transit's environmental advantage over cars will vanish with electrification. What public transit does do extremely well is get people from A to B with high throughput. That by itself is highly valuable in any major city.
Self driving cars can enable both urbanism and sprawl, depending on their deployment— if the existence of self-driving vehicle fleets at a per-mile price competitive with a personal car enables the collapse of personal car ownership in cities, that will open up a lot of possibilities for urban infill.
It will make the politics of building mid-rise housing without extensive, expensive underground parking space a lot more politically palatable in American cities. The same is true of replacing parking with pedestrian and bike infrastructure. And that says nothing about the safety benefits that will make cycling and walking a lot more pleasant and accessible in a lot of the US. (There are no technical barriers to building mid-rise urban housing without parking and replacing parking with high-quality pedestrian and bike infrastructure in the US today, but the politics are very difficult as drivers treat free/cheap access to city-owned street space as a right)
Plus, urbanites paying the marginal per-mile cost of a robotaxi trip, as opposed to buying a car as a huge sunk cost and then trying to get the maximum use from it, means that people will consider the car-versus-transit-versus-bike economics a lot more often, rather than buying a car and becoming someone who drives everywhere.
It might not be quite the same sprawl, though, because parking is different. With taxis and automatic parking, individual driveways, garages, and even on-street parking become less necessary.
In cities and dense inner suburbs, parking is a limiting factor to higher density; typically every parking spot is taken, and then some.
Also, automatic driving could greatly improve the cost and convenience of public (or private fleet) transit.
Sprawl is not sustainable in that an organism with unnecessarily long pathways wastes energy. However, many cities are not "done" growing yet and we can't really say if a city of sprawl will eventually become an efficient city as it is refactored over time and as the selection pressures of the environment change (energy cost, maintenance, etc...).
"Despite the hype around “micro-apartments” and other innovations intended to cram more people into less room, many Americans still want space. They want to live in detached suburban homes, or in an apartment with enough square footage and access to outdoor space that it feels like one. Two-thirds of people born since 1997, including those who live in cities, want to live in single-family suburban homes, according to a 2015 survey, but the costs make this aspiration prohibitively expensive in most urban centers."
They want something they can't afford... that the world can't afford/sustain. In my opinion the source for this want is equal parts marketing and subsidy.
We actually did decent low density non-sprawl until after WWII. Find it at the center of any city that existing before that (and wasn't bulldozed).
> Sprawl will never be environmentally sustainable.
Assuming clean, renewable energy powers all electric mobility, why not? I do not want to live in a city/urban area, and I will pay more not to.
EDIT: @InitialLastName
Sorry, HN throttling kicked in, can't reply to your reply so I'm putting this here.
If I have a well for water, septic for sewage, solar on my roof, and a gravel road to town, what's the problem? I understand its super expensive for cities to deliver utilities and services across sprawl, but if I'm not asking for those services and utilities to be provided, it should be a non-issue.
If anything, sprawl will make service and utility delivery more efficient, as we'll have no other choice for those who choose that lifestyle.
Have a look at the graph of the cost of delivering utilities. This isn't about environmental sustainability as much as long-term economic sustainability.
When towns sprawl, utilities with costs that scale with distance cost more per person.
> When towns sprawl, utilities with costs that scale with distance cost more per person. Have a look at the graph of the cost of delivering utilities.
That StrongTowns blog graph is intentionally misleading. It's not a cost of delivering utilities, it's a cost of utilities based on tax revenue a given parcel generates. It's an attempt to make cities "for profit".
Obviously, anywhere significant amounts of people live is going to come up "red", because we don't tax people the same rate we tax corporate offices, and because people use more resources there (most people live / cook / shower / sleep / etc at home, not in their cubicle).
That doesn't make it less worth that cost though (those downtown regions are only able to be "green" because there's lots of red around them). And it doesn't really have to do with sprawl. (The cost does rise with distance, but in mostly trivial ways. And those costs spiral backward again too -- 5,000 feet of new powerline in suburbia is far cheaper than 500 feet of new powerline in Manhattan, for instance).
> This isn't about environmental sustainability as much as long-term economic sustainability.
If that's the argument, then cities are almost entirely ruled out already. Cities aren't even economically sustainable for their current residents today, sprawl is the only thing preventing cities current economic issues from becoming even worse.
Sprawl is the only reason the majority of regular folks have any housing to sleep in at all.
When towns sprawl, utilities with costs that scale with distance cost more per person.
When factored in with total cost of living are these cities cheaper or more expensive? I can think of plenty of dense efficient cities that are more expensive than cities with sprawl.
What if the future of public transit is autonomous cars? Why do you need buses or trains when you could have machines that can take you from anywhere, to anywhere, on demand, for less than the cost of current public transit fare?
Well, I think that there's a human controlling the vehicle is a big part of the problem with each vehicle taking up so much space. If autonomous vehicles were working in unison, arguably:
* traffic problems will be significantly reduced
* you won't need traffic lights like they exist today
* the vehicles can be reduced in size (double occupancy perhaps); a family can just order a couple cars and have communication between them
* streets can be reduced to fewer lanes; possibly one-way because the vehicles can navigate the best route in any case
> Just like cars that we drive, self driving cars enable sprawl.
A self-driving car increases the radius of effect of things like mass transit stations. That can be a useful intermediate step before another transit stop gets built.
Hop into our self-driving car! It was designed by a team of inexperienced developers working under a manager who foreshortened deadlines and delivered ahead of schedule and under budget! The software guys asked for crazy things like 'bug trackers' and 'static analysis tools', but we kept their nose to the grindstone. Some of them said the software required more testing, but you know software guys, always such nervous ninnies. Luckily we maintained the standard business practice of keeping every employee replacable so that there were no worries when we booted those guys.
Don't forget to read the fine print about the driver actually needing to keep their hands on the wheel and ready to correct the car should it do anything improper. You bear full responsibility for anything the car does. And have fun!
Haha! Love it! This is precisely how enterprise software is developed and if you look at the way our industry has trained its managers, I can't think of any reason why the people that make decisions about these things can be convinced to do anything differently. I mean, if you can build buggy software for a Point of Service system this way, surely you can do the same thing with self driving cars! It's just software! How hard can it be?
One would like to think it would be handled by auto manufacturers. But one would be DEAD (literally) wrong. We found this out a few years ago when Toyota killed a couple people with their cars and their "unintended acceleration". The degree to which they were egregiously negligent was astonishing. They were doing development around 2010... and didn't have a bug tracker. Just, nope, no way for developers to keep track of and tell one another about bugs. Who would need such a luxury? Just don't make bugs! The auto industry has a big list of 90+ required standard practices in coding firmware, and 30+ 'recommended' practices. Toyota claimed to adopt 9 of the required ones, but repeatedly violated 3 in the code, nailing 6. The CPU didn't even have error-checking RAM despite Toyotas claims that it did. It was patently absurd how bad - and how utterly standard - their practices were.
And the court looked at it and said "yep, no negligence here. Software has no standards. There are no licensed engineers to hire. Go ahead and do whatever the hell you want, even if it kills people you're fine." No company is going to take the millions of dollars worth of precautions necessary to prevent things like that.
> Cruise employee here. throwaway for obvious reasons. It needs to be said that this is entirely marketing smoke and mirrors. The Cruise platform is significantly behind most other players in the market, and is having many many technical problems because of shortsighted leadership on Kyle's part, including high level people quitting (the head of planning and controls left last month because of Kyle, and more are threatening to quit) The reason I'm saying this is that this sort of public "everything is fine, we are the best" posturing leaks inwards. The opposite is very much true. Employees are very upset with Kyle because of things just like this. I make no exaggeration when I say Kyle is a mean-spirited, selfish person in private, and is very quick to publicly take credit for the work of others that he has emotionally and verbally abused. It's truly one of the most toxic environments I've ever been a part of.
Please don't use URL shorteners as it obscures the destination URL.
Edit to add: Also, if the intent was to reveal what was deleted, I would lean on the side of respecting someone's right to retract what they said. We've all said something that we regret saying. If you think there's something untoward going on, you can contact the HN mods via the Contact link in the footer.
Cruise employee here. I'm sorry to hear you feel that way. In my experience, Cruise is one of the most exciting companies to be working for today, and Kyle is easily the most intelligent, hard working, and reasonable CEO I have worked for.
For sure, there is still a lot of work to be done before we can launch a full L4/L5 vehicle, but I would say with high confidence our platform is ahead of most other players in the industry, and I'm confident that we will be one of the first companies to launch AVs at scale.
Not damage control, just providing an alternate opinion.
Yes, it's entirely possible that the commenter is (or was) an employee in a department other than engineering. We all report to CEO eventually though, so I don't think that would result in a significantly different experience.
> If you're manufacturing the car from scratch just build a digital bus into the frame.
And just how would you do this?
The frame is metal - it isn't plastic; so to move control signals around, you need to have a harness. That harness, on the majority of cars today, is attached to clips that are mounted to the frame (inside the vehicle), behind the interior fascia plastic. Google "vehicle wire harness" and look at some pictures to get an idea; most of the harness is behind the dash, with additional branches leading to the doors (locks, windows, speakers, etc), up the A-pillars (rearview mirror and dome light, or other accessories), up the B-pillars (air bags, more speakers), toward the rear (gas tank, brake lights, etc), and elsewhere. There's a ton of wiring in the interior of a car.
There's a separate harness inside the engine compartment, clipped mostly to the firewall and fender area, with branches going to various spots in the compartment (too many to name); that harness is generally connected in some manner to the interior harness of the vehicle, thru a pass-thru on the firewall. This may be a connector, or just a hole the wires are routed through. Sometimes, the junction will be via a computer in the engine compartment; the point being that there are many ways it is done.
Engineers over the years have worked to reduce the number of wires and such in the wiring harness; one of the more recent ways has been the nearly universal adoption of the CAN bus. It is essentially a robust vehicle communications bus with a differentially signaled system, using two wires. Of course, there are bunch of these in a vehicle.
I think eventually they'll move to a fiber-optic system for signals, with the only wires remaining for power.
Embed channels in the unibody. Seal power+digital rails in the channels and provide replaceable hubs at joints (which avoids some problems with CAN transceiver issues, and simplifies maintenance). Use plugs that can be sealed in place to connect sensors. The only wires, if any are needed, are between the sensor and a point on the frame, which might be a few centimeters apart.
Unibodies already are subject to replacement upon severe damage and the replaced part would include the data rails. Attached sensors could be disconnected and rewired or replaced. Fiber would be just as vulnerable to crash and transceivers would probably cost more, and it certainly wouldn't be less complex, other than perhaps EMI.
A lot of modern car design is a result of balancing complexity against maintenance cost. A lot of these factors will go away in a driverless electic world.
"The launch of the world’s first mass-producible driverless car is a major accomplishment, but we’re not across the finish line yet. Building a few of these self-driving vehicles, or even a few hundred, won’t accomplish what we set out to do. And a self-driving business that depends on humans sitting behind the wheel is fundamentally unsustainable, so that won’t cut it either. We will achieve success by integrating the best software and hardware to deploy truly driverless vehicles at scale."
That's the last paragraph. It makes me think the writer views it as a technical problem - they need a human as a backup. But, there's not enough context, so I could be misreading it.
It's harder to assess the state of the technology from the outside, but this is due at very least to legal limitations -- It's currently illegal to have a car on the road without a human.
Autonomous vehicles exist since quite a while now...
An Italian university tested an autonomous vehicle from Italy to China in 2010: the car was independent the 90% of the time, and semi-autonomous in case of lack of maps or other special conditions: https://spectrum.ieee.org/automaton/robotics/robotics-softwa...
