George Hotz: Fully Self-Driving Cars Are a 'Scam' and Silicon Valley 'Needs To Die'
The hacking wunderkind thinks Big Tech's approach won't work. He built a $999 autonomous driving system that runs on a smartphone.
HD DownloadIt seems like self-driving cars have been five years away for at least 15 years. But now, major players in the industry—like Google spinoff Waymo, GM self-driving unit Cruise, and upstart Zoox—are promising that fleets of fully autonomous taxis are just about to roll out.
"It's a scam," says George Hotz, the 30-year-old hacker-slash-entrepreneur best known as the first person to jailbreak the original iPhone when he was 17. "No one's close."
Hotz points out that every system on the road today requires the driver to pay attention at all times and be ready to take over. He says that companies touting fully self-driving cars without human safety monitors—and sometimes without steering wheels or pedals—are offering nothing more than a "press demo."
Hotz started Comma.ai in 2015 to upend what he views as Big Tech's wasteful and shortsighted approach to self-driving vehicles. Instead of building specialized cars that rely on expensive sensors and follow laser-mapped routes, Comma.ai has created an autonomous driving system that runs on a smartphone, works on most vehicles sold in America, and requires no additional hardware.
The company's first truly all-in-one device, the $999 Comma 2, packs a modified smartphone and other hardware into one slim plastic casing, which Hotz 3D-prints in the garage of Comma.ai's office in San Diego. Mount it to your windshield, plug it into your car's OBD-II port, and Comma's OpenPilot software can take the wheel. The Comma 2 uses the phone's cameras and taps into the built-in RADAR and drive-by-wire systems contained in cars built after 2012, automatically turning the steering wheel and operating the gas and brakes.
Hotz says his company has spent $8.1 million thus far and is profitable, while the big players in the self-driving space have spent billions without offering an economically viable product. While his competitors vie to dominate the market with proprietary technology and ridesharing platforms, Hotz is focused on building an open-source, decentralized ecosystem for driverless technology.
When I first spoke with Hotz in the summer of 2017, he predicted that by 2020, cars would take their own wheels for large stretches without humans paying attention and that by 2022 they would achieve full self-driving ability in limited areas.
"None of that's true," he says now. "Profitable robo-taxis are still a decade away."
Hotz thinks the automated vehicle industry has surrendered to what he views as the vices of modern Silicon Valley, focusing on growth and hype rather than delivering truly innovative products. To find out why he's soured on the space and to take the Comma 2 for a test drive, I caught up with Hotz at the Airbnb Comma.ai rented off the Las Vegas Strip during CES, the largest consumer electronics show in the world.
Produced by Justin Monticello. Camera by Monticello, John Osterhoudt, and Jeffrey Cummings. Graphics by Lex Villena. Music by The 126ers, Matt Harris, MK2, Quincas Moreira, Jingle Punks, and Silent Partner.
Editor's Note: As of February 29, 2024, commenting privileges on reason.com posts are limited to Reason Plus subscribers. Past commenters are grandfathered in for a temporary period. Subscribe here to preserve your ability to comment. Your Reason Plus subscription also gives you an ad-free version of reason.com, along with full access to the digital edition and archives of Reason magazine. We request that comments be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of reason.com or Reason Foundation. We reserve the right to delete any comment and ban commenters for any reason at any time. Comments may only be edited within 5 minutes of posting. Report abuses.
Please
to post comments
I have no idea if Hotz is on the level, but he reminds me of the self-driving projects of the 1960s. Million dollar computers, special specially-marked roads build just for, and used only by, the specially-built cars, and they couldn't manager more then walking speed in clear daylight.
Then some butty professor came along, took high speed movies of crickets and such, and discovered how incredibly dumb they were. Six legs. They'd lift three, move all three forward, let them down, alternate, repeat. If any leg was blocked by a twig or blade of grass, that leg would try to raise higher and try again. Maybe it wouldn't. Maybe two legs were good enough.
He replicated that level of stupidity and single-mindedness with 8 bit microprocessors and made more progress than the experts and their multi-million dollar setups.
Yes, they are a scam. They were always an answer to a question no one was asking. Moreover, they get the proper relationship between man and machine exactly backwards. Humans are great at doing subtle reasoning and communication but terrible at monitoring something. Machines are terrible at subtle reasoning and communicating with other humans but great at monitoring since a machine never gets bored or distracted.
So the future of driving technology is and has always been cars that the human drives but the machine monitors and steps in to save you in an emergency. Machines can greatly supplement our driving skills and make everyone a safer and better driver. But it is pointless and stupid to ask a machine to take our place.
I think a lot of people wanted to believe self driving cars would achieve something they never will for a variety of frankly emotional reasons. I think some are just intoxicated by the possibilities of central control fully functional self driving cars would create and others don't want to face the implications of machines being unable to recreate human thinking and consciousness.
"Computers are incredibly fast, accurate and stupid. Human beings are incredibly slow, inaccurate and brilliant. Together they are powerful beyond imagination." ~Einstein
They are obessed with control. Self-driving cars would basically become self driving public transit, nobody would own one and you would quickly become limited in where you could actually go.
That is exactly what it is. And I can't believe some libertarians are so easily taken in by the promise of new technology not to understand that.
The prospect of being "free" to play with their phone during their hour commute to work, while not even having to pay a small pretense of paying attention to the road, is awfully attractive to an awfully large number of people. Paying attention to the two ton metal death machine going 90 feet per second, is just too much to ask these days.
I see insurance companies mandating autonomous cars, once the bugs are worked out. If they ever are.
Agreed that autonomous cars are going to be networked, limited for all sorts of specious reasons---"Your carbon allowance has been exceeded for the month, Mr. Hands. Please take an electrobus..."---and will be a net negative for personal liberty.
It is a fucking nightmare to contemplate. At most I could see your scenario playing out in certain large metropolitan areas. But no way will it ever be practical or even possible in most of the country. The roads are too unpredictable and the needs of the drivers too diverse.
As a side note, the idea that personally owned cars will be replaced entirely by fleet owned autonomous cars has to be the most comically stupid idea I have ever heard. The people who push it seem to think that every person in the country is single, and uses their car for the single purpose of personal transportation. It never occurs to them that people have kids or use their cars to store things or anything like that. Or anyone would be a tradesman and use their vehicle for work or live in the country. Nope. Everyone is a 20 something hipster doing free lance graphic design living in Slope Park or something.
My favorite piece of logic is how, since you don't use your car most of the time, it is inefficient to own one. You could literally say that about anything you own. Hey, you have all of these clothes just sitting in your closet. What a waste of space. You can only wear one outfit at a time. So, we need to replace personal clothes with fleet owned rental clothes. Literally, that is exactly the logic these idiots use.
"At most I could see your scenario playing out in certain large metropolitan areas. But no way will it ever be practical or even possible in most of the country. The roads are too unpredictable and the needs of the drivers too diverse."
I'm sure, with your quote in mind, that the push to make more jobs service industry-related, and concentrated in urban areas, along with increased high-density housing, is purely coincidental...
I've never seen in any of these fleet proposals, who cleans these cars? I guess that's another of those service jobs. Or how beat up these things will be from the general public?
Autonomous vehicles are going to hit the intramodal market first, over the Teamsters and Longshoremens' dead bodies, I might add. When the robots can handle moving freight boxes around a shipyard or trainyard, we might start to see them on I-80.
The problem is the liability issues. There is no way the trucking companies are going to risk the liability that would come with totally autonomous vehicles. Can you imagine the plaintiff's closing argument in such a case? "Ladies and gentleman if the evil trucking company hadn't been so cheap and had someone there to step in when this machine made a mistake, my client would still be alive today". No amount of "but its overall safer than human driven trucks" is going to save a trucking company in those circumstances. So, they will always have to have a driver there to step in. And if you have to pay a driver to be there anyway, there is no point in having it be autonomous.
We're not all lawyers.
Agreed, but see elevators as a formerly piloted mechanism that gave way to automation. A lot less litigation when that happened though.
I think the industry and insurance will ask for liability caps, and the legislature will be all too happy to give it to them.
Elevators are not cars. And the elevator operator was not there for safety. He was there because the elevators couldn't operate themselves. His operating them didn't make them safer. If an elevator falls down a shaft today, having an operator there will not help. So the two are not analogous.
If the elevator would otherwise not stop level with the desired floor, the presence of the operator and his ability to ensure it would, does make the elevator system safer. Presumably the operator also ensured that accidents involving crush injuries with a rising or falling car and the building floor didn't occur. Eventually, the system was automated with a very high degree of repeatability and precision, such that the operator position could be eliminated.
It's a tortured comparison, but not that far off. If peoples' preferences are that autonomous cars become commercially available, and if the money makes sense, the law will follow some way, some how, to allow that to happen. It may be a liability carveout, it might be damages caps; who knows, the tech might actually improve enough that the robot driver's judgment isn't appreciably worse than the average human's.
It’s a progtard’s dream, and every good American’s nightmare.
Why wouldn’t your own one? Self driving cars combine the convenience of trains with the control of owning a car.
Imagine all of us having to do the exact speed limit? Sammy Hagar will quit driving for sure.
They were always an answer to a question no one was asking.
Really? Unless you really like your commute, why wouldn't you want an electronic chauffeur?
For years I wasted over two hours a day driving to work and back. It would have been awesome if I could have spent that time working on a laptop instead of listening to talk radio.
You like most people want to get rid of your commute. Sorry but being stuck in your car for two hours not driving is no better than being stuck there driving. People have these visions of doing all this work and their lives being better are kidding themselves. If you don't believe me get on a train or a plane sometime and see how many people are actually doing meaningful work. Very few if any.
If you don’t believe me get on a train or a plane sometime and see how many people are actually doing meaningful work.
I would be one of those people. Add that to the long list of my weirdisms.
You like most people want to get rid of your commute.
Notice I my last comment was past tense. I've been working from a home office for years. The commute is great but the Christmas party sucks.
If you don’t believe me get on a train or a plane sometime and see how many people are actually doing meaningful work. Very few if any.
It also assumes the time spent in the car automatically becomes yours. One of my wife's most productive jobs was when she commuted on the train. She got ahead because people around her were commuting to get an 8 hour day while she was effectively getting in a 10 hour day. The first six months recognition (and compensation) was great and it was awesome as hell. Then it became pretty much expected that she would put in the extra 25% and she couldn't ditch both the job and the ride fast or hard enough.
That stinks. I was able to log hours not worked in the office. For example I take my laptop with me to a doctor's appointment and work a half hour while I'm sitting in the waiting room. In that case I would say your wife should have been able to spend less time physically in the office since she was putting in time outside of it. Then again other workers would scream "Not fair! Waaah!" Can't win.
She was working her way up/out of bedside healthcare. Physically in the office was what she was being paid for, filling out paperwork before/after hours was getting her ahead.
Gotcha. Still stinks.
I think "without steering wheels or pedals" is key.
There are massive transportation upgrades that could pay dividends tomorrow if people were genuinely interested in travelling faster and more efficiently and/or leisurely. Transporting humans in a relatively upright position is modestly uncomfortable, aerodynamically inefficient, and completely unnecessary in an age of live video fly-by-wire. Steering wheels, sweeping passive analogue (or even digital) instrument panels, and pedal interfaces... all old, inefficient technology. The idea that a steering wheel is virtually the only or best, most intuitive way to connect humans to 2D land vehicles seems laughably absurd.
What would you replace a steering wheel with? And replacing actually looking at your surroundings with some kind of digital interface makes no sense at all. Moreover, there is nothing uncomfortable about sitting or certainly nothing more uncomfortable about it versus any other position.
What would you replace a steering wheel with?
Me personally? Analogue thumb stick. Predominantly with my left hand but with a backup on my right. Potentially with a set of foot pedals like the anti-torque pedals in a helicopter. I grew up with trucks and farm equipment that all had brody balls (as well as dual stick/dual drive systems) so it's exceptional when I drive with both hands in an automatic and, on the rare occasion when I have to drive with my knee(s), I think the torque-pedal mechanism would offer more control. More my point is that it's all fly-by-wire anyway, so as long as you can demonstrate proficiency with your wheel and I can demonstrate proficiency with a stick, it shouldn't much matter.
Similarly, I agree that the seated position is generally comfortable for the average person and the average commute but I think, regulation-wise, we've hemmed ourselves in to making people conform to the average in order to drive when there's no real reason to do so or, at the very least, no guaranteed reason to continue to do so into the future.
Not saying upright seating and steering wheels should be done away with, just saying that laws and safety standards rather actively forbid anything else without much, if any, demonstrable benefit.
"Analogue thumb stick."
Where is the resistance and feedback?
Where is the resistance and feedback?
You mean the resistance and feedback that we invented power steering to free us from? Maybe you think I'm one of the millions of people who will turn the wheel all the way against the power steering lock and then hold it there when pulling in/out?
Steering aside there's still the resistance I can feel when the wheels travel over something or break loose and the audible feedback I get from the engine as it struggles with whatever bad 'drive over it/there' command I've given it. You are aware that you will still pull Gs on a high banking curve at >150 mph whether you're using a steering wheel or a flight stick and that neither the wheel nor the stick is responsible for communicating the centripetal force, right?
Again, I'm not saying the thumb stick should/shouldn't have integrated feedback, I'm saying that, in the current (regulatory/legal! market?) environment, neither the manufacturer nor the consumer gets to even conceptualize having a say so.
You mean the resistance and feedback that we invented power steering to free us from?
I mean the resistance and feedback when my wheels aren't on a perfectly flat surface.
I’m saying that, in the current (regulatory/legal! market?) environment, neither the manufacturer nor the consumer gets to even conceptualize having a say so.
I get it.
No, the market has conformed to what people prefer. If people really wanted such a thing, they would have made cars with them in the first place. Remember, before 1966 there was no regulation of automobiles. You could make and market anything you wanted. And cars all came with the same controls they have today. That wasn't by accident.
Remember, before 1966 there was no regulation of automobiles. You could make and market anything you wanted. And cars all came with the same controls they have today. That wasn’t by accident.
C'mon John, you're not this naive. In 1966 a microcontroller in an automobile was unheard of, today they're thousands-fold more powerful and ubiquitous.
Are you really gonna sit here and tell me that automobile manufacturers who are famous for ripping off each other's ideas and getting fined for 1-in-a-million defects that don't kill anyone aren't innovation and risk averse? That manual acceleration, "programming your VCR"-style cruise control, or fully autonomous are really the only options for fixed or variable speed control?
There were tons of other ways to control a car. Indeed, the set up that we have today didn't come about until the 1920s. It became standard because that is what people liked and preferred.
Moreover, there is nothing in the federal automobile regulations that says you have to have a steering wheel. Indeed, some high end sports cars have more of a yoke than a wheel. So, there is nothing to stop a company from marketing what you are describing. They don't because their customers don't want it.
They don’t because their customers don’t want it.
Disagree. I think the fact that you can be held criminally liable for emissions faux pas and less than one-in-a-million defects, even when the defects don't hurt anyone are a/the significant contributing factor.
I freely admit that it can also be considered as the regulation locking us into the 'egg' option of a 'chicken vs. egg' debate.
Sometimes I wish my car's high beams were controlled by foot pedal instead of using my hand like that old Ford my mom drove when I was growing up. Does anyone do that anymore? Is it by customer demand or regulation?
Yeah. My '66 Sunbeam Tiger has the floor button for high beams. I'm thinking that the first column controlled high beams I had were with mid-70's Toyotas.
CB
So did my Nova
Is it by customer demand or regulation?
Fancy European cars have superior ergonomics so that Swiss bank executives don't have to stomp on the floor to turn their brights on.
Similarly, only idiots and hicks who have to have one hand on the wheel and one hand on the gear shift would need a suicide knob.
Says the guy who was lamenting about steering wheels.
Good point. I wish my car had handlebars.
I disagree. Truly self-driving cars have tremendous potential to increase liberty and to do things that some of us really don't like doing.
Personally, I consider driving a chore. Every time I've ever had a work/home combination that allowed for public transportation, I jumped at it. Sometimes I worked more but much more often I used the time to read for pleasure, nap or just meditate. I consider driving unpleasant and stressful. I would cheerfully turn the "keys" over to a self-driving car if it was practical and cost-effective.
More than my personal lifestyle choice, however, self-driving cars have the potential to be a god-send to elders everywhere. I love my mother dearly but she should not be behind the wheel of a car anymore. But taking the keys away is a de facto prison sentence. She would be unable to buy groceries, socialize or do much of anything except rot at home. Providing safe transportation options to those who can no longer drive themselves is a huge quality of life issue.
And this brings the key point to perspective. Self-driving cars do not have to be perfect. They don't even have to be as good as the average driver. They just have to be better than the worst drivers to be a net gain for safety. Looking at some of the loons on the road, that's a pretty low bar.
I recognize that not everyone will make my choice. And that's okay. If you really enjoy driving, more power to you. But don't get in the way of my choice just because you it's not personally appealing to you.
They just have to be better than the worst drivers to be a net gain for safety. Looking at some of the loons on the road, that’s a pretty low bar.
Good point. I once worked with a guy who kept driving with a suspended license, and kept getting pulled over. Why? He's a shitty driver.
Also big boon to the disabled. My big brother doesn't have the coordination to drive. It would grant him a lot more freedom and the people who drive him more freedom if a car could drive him on its own. Even if it only works on sunny days, that's more days that he could transport himself than he can now.
The blind would also benefit a lot.
And this brings the key point to perspective. Self-driving cars do not have to be perfect. They don’t even have to be as good as the average driver. They just have to be better than the worst drivers to be a net gain for safety. Looking at some of the loons on the road, that’s a pretty low bar.
I recognize that not everyone will make my choice. And that’s okay. If you really enjoy driving, more power to you. But don’t get in the way of my choice just because you it’s not personally appealing to you.
This, a bit, drives at what I was getting at. The focus is too narrow on 100% self-driving. The technology to make the shittiest of drivers more capable exists and would be valuable to insurance providers and the shittiest of drivers.
A 6-button (programmable touch) panel with big labels saying 25, 35, 45, 55, 65, and 75 is not that big of an ask and would do wonders for grandma's ability to keep the car at a consistent 65 on the highway, *and* still offers the majority of drivers the freedom of the pedal with the ease of the cruise.
Quit trying to make the narrow self-driving car for people who hate driving and make commuting better in general.
I disagree. Truly self-driving cars have tremendous potential to increase liberty and to do things that some of us really don’t like doing.
Freeing you from a chore is not making you more free. One has nothing to do with the other. If it did, then prisoners who were not required to work would be more free than freemen who did.
And the problem is not you making that choice. If you want such a thing and it ever exists, have fun. I don't think anyone here would say you couldn't make such a choice. The problem is that the people pushing these cars want to deprive everyone else of their choice and mandate their use.
While their use has the potential to increase freedom for those who are unable to drive, their potential to destroy freedom through the privacy and autonomy issues created by central control is orders of magnatude greater.
Nothing comes for free. If you want to have the chore of driving anymore, you necessarily give up your freedom and control over that function to whomever does it for you. Right now, you are free to drive in any manner you wish up to and including breaking the traffic laws if you want to take that risk. With an autonomous car, you will drive when and how the company who wrote the software tells you. You can't give up responsibility without also giving up control and freedom.
You are boxing at shadows, John. Nobody is pushing these cars, much less pushing them in a way that deprives anyone else of choice or that mandates their use. Centralized control is a theoretical way to create self-driving cars but it is almost certainly not going to be the actual path that we take. Decentralized approaches like that used by Comma are far more likely to work and be adopted.
You are boxing at shadows, John. Nobody is pushing these cars, much less pushing them in a way that deprives anyone else of choice or that mandates their use.
You are completely wrong about that.
https://www.sharedmobilityprinciples.org/
Ron Bailey has published multiple articles arguing for just that in the name of "safety". I am not boxing at shadows. You just have no idea what the hell you are talking about.
Piling on, if autonomous cars were to become the reality, and accident rates plummet as much as I think they would, insurance companies would de facto mandate their use.
Further, a way to improve the performance of the autonomous car system would be to network them. I trust the privacy and personal autonomy problems of a networked system of vehicles, that were formerly largely anonymous, is apparent to people here?
You would think it would be. Like I say above, you cannot have freedom and then farm the responsibility for doing something out to someone else. It doesn't work that way. If you want someone else to drive you around, then that someone else will necessarily have the freedom to control how it is done and control over your privacy.
You are completely wrong about that.
Yeah, even to the point of lying about how it will speed commutes and eliminate the need for vehicle ownership and parking.
I should add that I'm not necessarily advocating that we relieve anyone of any responsibility while driving, just noting that we, for some reason, are obnoxiously narrow in facilitating the variety of means by which people can tackle those responsibilities. The technology exists to more broadly empower and liberate drivers without automation. The fact that many futurists look past this in support of automation lays bare the underlying fact that they really just want the driver to have less control.
I am all for killing off the highway safety regulations and CAFE standards and letting the car makers make whatever they want.
This needs to be chiseled into 50 foot tall flaming letters on a monument somewhere in D.C., lest its truth slip the legislators' minds.
Well done.
Disagree. If they were only slightly better than the worst drivers, and there were a lot of them, then the average driver quality would approach the level of the worst as more autonomous cars became part of the total.
You don't want a self-driving car?
Cool story.
More for the rest of us.
lets do some math.
More of 0= {*X!>} ?
This. But one of the main problems with computer 'monitoring' is that people rapidly tune out the warnings if and when the warnings prove to be mostly not important. Unfortunately this means that people also ignore the warning when it does matter.
My wife's new car has an interesting solution to this issue. It allows the driver to set the proper following distance. Yes you can set it too close for the driving conditions, but that is outweighed by the psychological effect of the machine only telling you what you want to know. Having set that value, should the alarm sound you know it matters because you already said it did.
To me that is using machines smartly.
Driving isn’t like writing a poem, it’s simple locomotion, something even insects manage to do extremely well.
Libertarianism?
this is awesome. Favorite interview in a while.
Agreed. That dude has forgotten more than I'll ever know.
All this does is make Americans even lazier!
Yep, it's worse than the tractor!
Yes fully autonomous cars are a long way off, if they are even possible without a total overhaul of the road system, but do consumers really want a fully autonomous car? I think we are pretty close to the monitored self driving balance consumers actually want
If we redo the road system with the required ground-based telemetry to truly make fully autonomous driving a reality, it could definitely replace fleet vehicles.
"If we redo the road system with the required ground-based telemetry to truly make fully autonomous driving a reality, it could definitely replace fleet vehicles."
Long distance truck driving and intermodal hauling. No more driver rest, no more driver pay, continuous operation without stopping. I can see a lot of logistics companies wanting to give their first born for something like that.
I can see a lot of logistics companies wanting to give their first born for something like that.
The Union thug standing in the background is having none of it.
Beyond what I want -- It's like riding with my mother in some of these new "safety" standards. Don't use the radio while you drive, beep, beep put on your safety belt. etc... etc... etc...
The "new" standard is too nannyish for me.
Hotz points out that every system on the road today requires the driver to pay attention at all times and be ready to take over.
Is this true? Isn't Waymo running a taxi service with no driver or safety engineer for select riders in Scottsdale?
To use Waymo’s service, dubbed Waymo One, riders must download an app and provide a credit card number, similar to ridesharing services Uber and Lyft. A human driver will be behind the wheel, but only to intervene in case of emergency.
Yes, Waymo still uses a human driver and from what I can read between the lines, the autonomous service runs on a carefully-chosen circuit and... *clears throat* it's Scottsdale, so there is 0% weather effect.
I've told this story before, but it's the kind of story that needs to be re-told.
A friend of mine who travels a lot for work was sitting on a plane next to someone who worked for one of the autonomous driving companies working on their test-tracks somewhere outside of Las Vegas. My friend asked the guy a question about how the vehicle dealt with snow and other weather effects that can obfuscate lines and road markers. The guy laughed and said, "We don't get any snow in Las Vegas. Friend quips, "So no one will ever take one of your cars up to Tahoe?
Guy just sat in silence trying to work out an answer.
Shorter, yes, there are some very limited uses of autonomous cars in Good Weather Regions. We're 1,000,000 miles away from fully, level 5 autonomous driving in randomly chosen geographic conditions and weather with no human driver ready to take the wheel.
Yeah, I don't expect them to solve snow and ice, but it appears that there are truly driverless cars running in Scottsdale.
https://www.theverge.com/2019/12/9/21000085/waymo-fully-driverless-car-self-driving-ride-hail-service-phoenix-arizona
Interesting, so you're right, it looks like they did add a few fully driverless vehicles. I found this interesting:
The first paragraph outlines well the frustrations with self-driving technology. It all seemed so easy when it was first put on a test track. But then you add passengers, and the fickle nature of humanity, and things get tricky real fast.
The second paragraph is also interesting because they introduce a "few" fully driverless vehicles, and seriously geofenced them. I'm guessing expanding that geofence area is probably a long, careful and arduous process. And maybe not even possible depending on what factors they used to choose the geofence in the first place. Like it all works great until there's a construction project or a sewer line needs to be dug up etc which blocks off a througway or has cones with little 'keep right' signs poorly placed on the street etc.
right, I was talking to my dad about this. This might work in parts of scottsdale, but in the bay area and dallas texas I have never, in 40 years seen them be finished working on the roads. Ever. They always have the roads blocked off in some manner. With lane cones, and at night only one lane..all my life..it never ends..ever..
I think maybe these people imagine some future when the roads will be ....DONE! Well...it has never existed in my life, or my dads life, or anyone's life probably.
And my current car, a 2019 Accord gets all screwed up when this happens and starts to break and do all manner of crap. I turned off the 'safety' systems for this reason, because it can't read the lines on the road right because so often there is construction or whatever.
This puts me in mind of all the street construction projects going on in my neighborhood. Daily shifting lanes of cones with poorly marked signage. So poor, human drivers have become confused and have been entering the oncoming temporary lanes. No right turns allowed for brief times, again, poorly signed causing more confusion to the human drivers.
No signage but a construction worker with a stop/slow sign on a PVC pipe waving cars and stopping them because they had to reduce to 1 lane for both directions etc. When you think of the difficulty in machine 'interpretation' of this stuff, you can see just how fast this gets "harder than we thought".
And the human can take over in such cases.
Even if you got self driving that only worked in good conditions and required a human to take over when things got bad, that would be something many people want.
Of course, self driving cars can use precise gps, radar, and other sensors to drive in conditions when even a human can’t drive. They don’t need to be able to see at all.
It’s true, but it’s a legal requirement, not a technical.
It seems like self-driving cars have been five years away for at least 15 years.
Finally, a writer at Reason throwing some much-needed cold water on this shit.
Technically a producer and really just parroting his interviewee's objectivism but, yeah, good to have a Reason writer that doesn't get a hard on every time somebody who knows something about technology says the word 'future'.
"It seems like self-driving cars have been five years away for at least 15 years."
Waymo was founded in 2009. ... So no, this a really dumb remark.
The remark isn't referring to Waymo, it's referring to the generalized 'predictions' of industry and technologists in general. This stuff has been going on since at least 2004. And I'm too lazy to google what the predictions were in the 80s... and even back to the 60s.
and upstart Zoox—are promising that fleets of fully autonomous taxis are just about to roll out.
I have insider information on Zoox. They're done. Finito. Es Todo, no mas.
When I first spoke with Hotz in the summer of 2017, he predicted that by 2020, cars would take their own wheels for large stretches without humans paying attention and that by 2022 they would achieve full self-driving ability in limited areas.
"None of that's true," he says now. "Profitable robo-taxis are still a decade away."
I'm glad that Hotz now admits that he too was full of shit.
Living in the 'test range' for autonomous vehicles, I can tell you they are no where close to ready-for-prime-time.
They still stop behind double-parked cars and the 'manager' has to take control to get them moving again.
Further, pedestrians often confuse them, not least because pedestrians are often confused or distracted themselves.
Be glad of that. The only way auto outs cars wont destroy more freedom is if we get rid of post progressives. And so far few have the stomach to give them the boot
Pedestrians are morons.
actually most farm tractors and grading tractors in general are all ready and many survey drones are as well but they do work in limited arenas so it won't be to long before they reach full potential.
They don't have to deal with idiots driving around the field.
I think one of the major obstacles to autonomous cars is people. If *poof* all cars were made self-driving, and people were removed from the equation, it would be a lot easier. Notice in the video that the system did great until another driver did something stupid and it didn't know how to react.
"They don’t have to deal with idiots driving around the field..."
Nor walking off the curb with their nose buried in their 'phones, nor peddling by on the right of a vehicle making a right turn, etc.
Years a go I worked on a project in the US involving a bunch of Australians. They were always looking the wrong way down the street and stepping off the curb. We had to watch them carefully.
Did you read the next sentence? The one that said "one of"?
Oh, my goodness!
Please, please forgive me for adding some detail.
And stuff it.
Take a sedative.
Pedaling. Unless they are selling something.
Driving a tractor in an open field is a very easy problem compared to driving a car in traffic. I have always said that if there is a future for fully autonomous vehicles it will be tractors and mining vehicles and things like that that do not have to interact with other drivers and vehicles the way road cars do.
Fully self-driving trains haven't been widely adopted and they're confined to two steel rails and tracks in full control of command center operators and ground-based telemetry and sensing infrastructure.
I'm going to guess trains might have a lot more to do with government unions than practically.
It's definitely a factor, I can't deny. I know someone who worked for the railroad and he indicated the Union wasn't happy about the technology.
John McPhee, "Common Carrier", rides along in a unit train, and the union is still quite strong and seriously opposed to automated trains.
actually most farm tractors and grading tractors in general are all ready
This is like saying your Roomba is ready to clean up the streets.
Either you've never done the work or you're grotesquely underestimating the amount of human intervention that still goes on, or both.
This technology will never sell without two contradictory settings; "go to slow in the left lane" and "tailgate that bastard until he moves over".
"Please select the number of lanes you would like to change."
Just hope the almighty government doesn't squash out competition like they have ALL THE OTHERS through some politicians will decide what you get to buy for your own "safety" measure.
Young people are embracing Sanders and Warren and "rights" to the labor of others...and this will be the "decade of decentralization?" It doesn't compute.
Government and big institutions can't be trusted, that's why we need more of it.
Dude! The corporations control the government! We need to give more power to the government so it can control the corporations that control it! Shit! The corporations still control the government! We need to give more power to the government so it can control the corporations that control it! Shit! The corporations still control the government! We need to give more power to the government so it can control the corporations that control it!
.
.
.
I have a self-steering car. After using this thing for a few months, there is no flipping way self-driving cars are even close to prime time.
Completely self-driving in real world situations is a pipe dream. The auto makers and tech companies know this but it sure won’t stop them from duping investors that it is just around the corner-hard to believe that investors can be so dumb. Another thing is that car culture is so entrenched in America that cars are extensions of our personality and driving a freedom that few will want to give up. Even today’s teens who aren’t getting their drivers licenses probably at least want the possibility to be able to drive one day.
I think the death of the car culture in this country is greatly overstated.
"Completely self-driving in real world situations is a pipe dream. "
They will be probably be restricted to specially prepared roads. It's likely that human-driven cars won't be given access, not unlike bicycles today, for example, being unwelcome on the interstate network.
The problem is that the money to build those roads will have to come from taxpayers. I don't see taxpayers being to interested in funding roads that only take cars they don't own. It is a catch 22. Without the roads, no one will want the cars. But without the cars, no one will want to fund the roads.
Yeah, you could build a private road that is autonomous only. The problem with that is that your autonomous car still has to be able to drive from where you start to the road. So, you likely end up with ordinary cars that have an autonomous feature on them that can be used on such roads. That still is not fully autonomous cars as envisioned.
18-wheelers stopping at a depot on the outskirts of town to either gain or lose a driver (based on whether they're heading into/out of the city) would be a huge boon, you realize?
And as the tech becomes increasingly refined in narrow domains, it'll increasingly worm it's way into wider domains.
Ironically, if this all leads to private commercial-only roads full of autonomous trucks, that'll be hilarious, as we'd have basically re-invented the railroad.
Ironically, if this all leads to private commercial-only roads full of autonomous trucks, that’ll be hilarious, as we’d have basically re-invented the railroad.
Yeah, lets put these trucks on a track and hook them altogether. That is a really funny and astute point.
they have these in Germany when I was there like 5-7 years ago.
"truck trains"
a number of trucks would be driven one right behind the other by one driver who was in the lead truck. They looked to be following the lead truck and did not appear to be hitched to my eyes.
They would drive down the autobahn like a train with one driver and then the driver would get off at some stop and meet a bunch of drivers and they would each take the individual trucks to their various destinations.
"They will be probably be restricted to specially prepared roads. It’s likely that human-driven cars won’t be given access, not unlike bicycles today, for example, being unwelcome on the interstate network."
You bet! Trueman here will finance an entirely new Interstate network.
Screw...
ball.
"Hotz is focused on building an open-source, decentralized ecosystem for driverless technology."
Open source. Does that mean that users will have the chance to edit and otherwise monkey around with the car's source code? Changing the response to a yellow light, for example. Could be dangerous.
For their own system? Possibly. "Open source" doesn't necessarily mean "once we sell you the box, you can push your own changes to it" after-all. So it's possible that you'll be able to tweak your own box, but also possible that you might not be able to. I suspect that you'd be able to, but it's a "voided warranty, assumption of liability" scenario.
But changing code in the public repo? Well, you can probably do a pull request for whatever change you want, that doesn't mean they'll accept it.
I thought open source by definition meant that the code can be freely modified and distributed. Perhaps the meaning is evolving and the modification and distribution ideas are no longer applicable.
Ten years ago I said it would be more than ten years before we saw fully autonomous self-driving cars available commercially. I still say it's going to be another ten years. We'll see some rollouts in highly mapped urban areas, but I still don't see any of these cars successfully navigating anywhere else.
I think urban areas would be the last places where it would be permitted considering how dynamic an environment city streets are.
Within 5 years I could imagine rural divided highways being the first places where hands free without need of driver attention will be permitted.
Such roads are already fairly well marked, no cross traffic (or very limited), no pedestrians, and everyone is (supposed to be) going in the same direction.
No high precision mapping needed, no crazy expensive sensors, just forward/rear/side cameras and a front radar.
Only places the driver would have to take over, or more likely babysit, is when you approach construction zones, and when traversing geo-fenced portions of the highway that pass through urban areas.
Country roads are just as unpredictable and more dangerous in some ways. There is a greater danger of wild life being on the road. There are more blind corners and two way stops and things like that. I don't see it working any better there than in the cities.
The only place where it can work is on full fledged divided access controlled interstate highways.
Yeah. HOV and express lanes will probably be first. I'd bet 2-3 yrs. before some green initiative pushes to have them revamped to give priority to self-driving electric vehicles.
Urban vs rural is the difference between identifying and tracking hundreds of objects at once vs a half dozen or so.
To the limited extent that you actually do have stop signs/lights outside of towns, identifying and abiding by stop signs/lights is child's play. Only reason the current crop of semi-autonomous cars can't is because its not enabled or permitted.
For the rural divided highway that does have roads directly intersecting w/o a ramp, detecting vehicles waiting to enter the roadway, or more importantly entering in an unsafe manner is more of a challenge, but by no means an impossible challenge.
As for wildlife its a factor on any road, even controlled access highways. You can only react to what can be seen visually and if the highways around where I live are any indication, we aren't so good seeing animals at night. Cameras can be made to see things we can't. Thermal + visual composite should give a car superior animal detection capabilities.
However long before London can effectively ban all private vehicles except self-driving ones. Once you can plausibly commute from one end of the city to another in an autonomous car, London or maybe Seattle will use lies about bike-friendliness, safety, and emissions to ban private commuters. Once it can be co-opted into a third-ish public transportation option, then it will become legal/mandatory.
Full self-driving is a bit like cold fusion or the climate "tipping point".
Ten years away for the last 40 years...
The guy's a troll. There's absolutely no way a mobile phone camera and the parking sensors on a typical care are even close to being capable of driving a car.
If nothing else, my phone only has one camera pointing forward. My car has a second limited backup camera. Do you seriously think that's sufficient? Personally I'm whipping my head around all the time as I merge, turn, park, etc. and I've got two eyes so I can judge distance.
The Tesla has 9 cameras built in and the new iPhone advertises dual forward-facing and one rear facing camera.
Not saying that the guy is talking about using any/all of this, but that his cries of bullshit stand the whiff test a bit even if not for his specific reasons.
There is a difference between sensing and perceiving. Cameras sense. That is the easy part. Perceiving and knowing what the camera is seeing is the hard part. People seem to forget that and think slapping a bunch of cameras on a car is all that is necessary to make it autonomous.
Their code is simply awful. I went to the GitHub repository https://github.com/commaai/openpilot pulled it down and took a look at some of the C code in openpilot-master/selfdrive/common and the C++ code in openpilot-master/selfdrive/locationd/params_learner.cc . Not a comment in site! It appears that a lot of their code base is Python as well.
★Makes $140 to $180 consistently online work and I got $16894 in one month electronic acting from home.I am a step by step understudy and work essentially one to two or three hours in my additional time.Everybody will complete that obligation and monline akes extra cash by simply open this link......Read MoRe
Hasn't anyone mentioned the autopilot systems that fly planes? However, I think having an equivalent system for cars is a far more difficult problem to solve.
"Hasn’t anyone mentioned the autopilot systems that fly planes?"
Airplane autopilot systems still require the full attention of a pilot who needs to be ready to take over immediately. That's basically where current self-driving tech already is, except it's not as good because as you correctly note, there are a lot more variables involved with driving a car (at least in traffic, weather, whatever)
Is it odd or unreasonable that my first gut-reaction to this awesome new innovation is imagining how the politicians will write the "standards" code portion and legislate out ANY and ALL competition from the market so I HAVE to have a GOV ride-along everywhere I decide to drive ?MY? car?
This guy's product is a toy for tech nerds, to compare it to actual autonomous driving is a joke. This whole article reads like an ad.
"the vices of modern Silicon Valley, focusing on growth and hype rather than delivering truly innovative products"
I've only been employed in SV for a few months, but I can attest that this quote is highly accurate.
I live in the Rockies and drive a 15 mile canyon to work. On one side is sheer mountain and the other a drop off into a large creek/river mere feet away. In some spots if you left the road you are going several 100' down the cliff to the river. No way in fucking hell I am trusting something like this that sounds like it needs wifi to work as well which guess what you don't have consistently in the mountains! Even if you monitored it 100% if it decides to suddenly turn you will not be able to react in time to correct and you are likely DEAD. Not to mention weather and icy snowy slush roads. What is it reading to navigate when there is nothing to read but SNOW?
As I mentioned before, the roads will have to be specially prepared, such as having sensors embedded in them.
I work with automation on large ships, and here is my take:
On ships, we have the ability to pretty much completely automate the voyage. But the "almost" is the key word. Maybe things run great for days at a time, but then something malfunctions, or an object appears in your path that is not showing up on the radar or AIS, or there are weather-related problems.
When those things happen, it is critical for the bridge crew to take over seamlessly and quickly. That means that they are alert and attentive, and keeping track of the "prevailing conditions and circumstances". There is not time to hear the alarm, then figure out the speed, heading, traffic in the area, desired track, and proximity to shoals or other hazards. They need to already be well aware of those things to act quickly and properly. It is really hard on the bridge crew to do that, maintaining the required level of alertness for hours or days while nothing of interest happens. But it is necessary.
People in cars are never going to have that discipline. They are going to get bored, and very likely be reading from their phones or even asleep. Even if the vehicle is 99.9% accurate, when the driver needs to suddenly take over, they are not going to be prepared to do so. There are going to be horrible accidents, which will cause death and damage, but also unbelievable delays and jams.
And you have to figure in mischief. Some people are going to try to cause mayhem just for kicks, or for larceny. An automated cargo vehicle traveling through isolated territory is going to be a target.
Maybe we should figure out how to automate the trains first. That seems achievable.
Stay At Home Mom From New York Shared Her Secret On How She Was Able To Rake In $1500 Weekly From Online Work Just 3 Weeks After Losing Her Old Job...... Read more
But the thing is, it could be possible due to on-going advancements in AI's and it is surely gonna possible in near or in the future and I think many software development agencies are working on many advanced technologies. In the future, our roads will have AI-based cars.