Seeking Alpha

Google's Eric Schmidt continues to hype the development of a self-driving car, this week taking...

Google's Eric Schmidt continues to hype the development of a self-driving car, this week taking up the case at the annual Allen & Co. media confab in Sun Valley. According to the exec, the innovation works splendidly as drivers equipped with a "Google Maps on steroids" - along with computer-linked cameras, radars, and lasers - see increased safety and fuel efficiency. Automakers quietly back the idea of manufacturing models, with Ford going on record as saying self-driving cars of some variety will be here within five years. Increasingly, the question on the tech is not why or if, but when?
Comments (50)
  • Early reactions by scoffers claim that a self-driving car is an accident waiting to happen. I would say that about human drivers with a hundred years of evidence to back me up. The self-driving car is a brilliant example of applying Google thinking and technology to a novel situation that will be a game changer sooner than most people think.
    14 Jul 2012, 11:22 AM Reply Like
  • Self-driving car or AUTOPILOT is an accident waiting to happen. But human-driven cars are 35,000 deaths per year and 1 million ER visits already. Does anyone REALLY believe that Google is going to be worse than that?
    14 Jul 2012, 12:11 PM Reply Like
  • A self driving car is pushing the initial implementation too far. They should start with a car that can take over in cases where the driver is clearly about to hit or be hit.

     

    So if you are within 20 meters of a car and you are going 50mph the computer will take over to either drive you around the obstacle or break as safely as possible. The benefit of a fully road-aware system is that the car would be able to calculate what is the safest move in that situation given road conditions, physics, what other cars are around, traffic lights, road signs, etc.

     

    I'd argue that a significant number of bad accidents today are by people who are texting or daydreaming or playing with the radio, etc. If we could simply remove THOSE accidents then the roads would be a lot safer. And i think this system by Google is uniquely well suited to those situations where the driver is about hit something purely because they aren't paying attention.

     

    The problem, I guess, would be that for this system to really be effective it would have to remove the driver from control in emergency situations. It would be applying power and breaking to each wheel as well as steering. That is the only way it could work. If it is also accepting input from the driver then it could actually make things even worse. I mean, most people already have a form of this in Traction Control Systems. But it doesn't feel that way because you still have control of the steering and most of the breaking.
    14 Jul 2012, 11:36 AM Reply Like
  • Mikeurl,

     

    There are many mistakes you make. First, Google self-driving cars are not an "initial implementation" so your whole argument falls apart. See Cadillac Super Cruise, Ford Lane Assistance etc. Each automaker has a semi-autonomous mode they are developing. This is the intermediate step you haven't heard of or seen before.

     

    Your argument that a significant # of accidents are by people texting etc misses the point. Those distractions are here to stay. We can track people, make them stop texting through law enforcement interdiction (bad on many levels, reduced freedom, takes police away from high priority crime) or we can remove human judgement from driving. Perhaps you have an example of where draconian police interdiction is the optimum solution to a problem (see WAR ON DRUGS) but me (and many others) see technology increasing safety as well as providing the driver with more time to do as they please without endangering others.

     

    Finally, the system is intended to remove the driver from control in emergency situations. A driver is ALLOWED to control the car when/where they want, but the computer, antilock brakes, traction control already does a better, safer and more efficient job than a human operator.
    14 Jul 2012, 12:20 PM Reply Like
  • Makes sense ,but all cars are not created equal, and have different tires ,suspensions, etc. and therefore would react different regardless of who, or what is in control. Like concept,but nothing is easy.
    14 Jul 2012, 04:32 PM Reply Like
  • > They should start with a car that can take over in cases where
    > the driver is clearly about to hit or be hit.

     

    Perhaps they will:
    http://bit.ly/Mrr8nQ
    15 Jul 2012, 10:08 AM Reply Like
  • Thanks D_Virginia, that is exactly what I mean.
    15 Jul 2012, 10:54 AM Reply Like
  • Definitely impaired drivers of some sort should be the first adopters.

     

    You can imagine the following apps:

     

    1) DWI/DUI offenders being forced to drive something 100% computer controlled.

     

    2) Insurance companies forcing reckless drivers to use these vehicles if they want to be insured.

     

    3) First 1 or 2 years of being on the road for teenagers.

     

    4) Senior drivers.

     

    5) You can imagine special freeway lanes being reserved because more cars could be packed per unit length while under computer control.

     

    6) The interesting thing would be if the car cameras could be accessed to hunt for scence changes indicating crime or the license plates of people on the lam.

     

    Naturally in return for some cheap technology license, GOOG will own or have access to where the car stops for marketing and data mining purposes.
    14 Jul 2012, 11:59 AM Reply Like
  • for self driving cars to work would seem to require all cars to be self driving, otherwise you have the problem of human error for those driving their cars, hey I think its great, no more taking the car away from old Mom and or Dad, no more auto insurance, no more DUI, no accidents, no law suits, no restrictions on texting, phone calling, sleeping, no traffic delays, and the end of cab drivers, its nirvana.
    14 Jul 2012, 12:13 PM Reply Like
  • Actually the reverse is true. Even where a significant portion of the cars are on autopilot, the remaining self-driven cars will find reduced traffic, more predictable outcomes from other vehicles and greater safety.
    14 Jul 2012, 12:39 PM Reply Like
  • The reverse is TRUE Really, tell me how you feel you can make such a wild statement, as long as humans will be driving you cannot remove human error. Sounds like your already a resident of Nirvanaville
    14 Jul 2012, 01:02 PM Reply Like
  • Enigmaman,

     

    This is obvious and self-evident. That it is a predictable outcome of autonomous cars is a function of how autopiloted cars work. They are not distracted. They see (using radar, lidar, cameras, GPS) further than human senses. They don't have "ego". They don't think they are better drivers than they actually are. They obey posted speed limits. This is not "Nirvanaville", but a function of their design.

     

    If you have some EVIDENCE for your assertion that "for self driving cars to work would seem to require all cars to be self driving, otherwise you have the problem of human error for those driving their cars". Um . . . we have human error right now to the tune of 35K dead and 1Million ER visits due almost entirely to human error. If you removed SOME human error as you posit, then you remove SOME loss of life and damage. If your argument is that not all harm or errors will be removed, only some, then you concede my point. I accept your apology.
    14 Jul 2012, 01:17 PM Reply Like
  • TL- stop pontificating and indulging in self masturbation, and think, read and comprehend before you respond.

     

    My point was only that as long as there are humans driving any cars there will be "human error" end of point.
    14 Jul 2012, 03:08 PM Reply Like
  • Engimaman,

     

    Only one of us is "indulging" and it's the person who wrote "for self driving cars to work would seem to require all cars to be self driving,"

     

    My objection is to that FACT FREE statement. Think before you write silly thoughtless things like that.

     

    If your point was that human driving contains errors, then why bother to write anything at all.
    23 Jul 2012, 10:31 AM Reply Like
  • Unless you take away the threat of lawsuits this will never be practical to implement.

     

    Every computer/machine/human eventually suffers failures or breakdowns. The liability from even a rare 'bad event' from deep-pocketed providers of the service would end up with them being bankrupted by the mass tort lawyers.
    14 Jul 2012, 12:29 PM Reply Like
  • Goog contemplates (and legislation in Nevada) requires the equivalent of a "little black box" which will reduced lawsuits. Cameras and other imputs will record the actual causes of accidents and not very unreliable witness testimony will not provide much if any weight to a claim of malfunction (unless malfunctions actually causes the accident).
    14 Jul 2012, 12:41 PM Reply Like
  • I already have a self driving car. I get constant input from my wife. She commands. I steer.
    14 Jul 2012, 12:44 PM Reply Like
  • Sounds like your autopilot works extremely well.

     

    And you didn't have to replace your vehicle.
    14 Jul 2012, 01:42 PM Reply Like
  • @billddrummer Actually, I have a similar versin to @The Goeffster, and I find it a bit annoying.
    14 Jul 2012, 06:50 PM Reply Like
  • what kind of money are we talking about
    14 Jul 2012, 12:45 PM Reply Like
  • It is probably true the self-driving cars on "autopilot" would result in far fewer accidents than normal human driver stupidity and inattention, on top of the rare instances of equipment failure.

     

    Already automobile manufacturers deal with hundreds of millions of dollars in lawsuits in alleged (and on occasion some real) defects; for example brake or steering failure, runaway throttles, fires, etc.

     

    Then there are tort lawsuits from folks that are killed or injured when crashing at high speed into bridge abutments and flipping over, while driving drunk, texting with their peeps, and not wearing seat belts. Somehow that becomes the manufacturer's fault for failing to prevent it from happening.

     

    With some 6 million car crashes a year in the US, resulting in 3 million injuries and 40,000 fatalities, folks are always looking for someone else to blame, and someone to collect from.

     

    With cars that drive themselves, it is only a matter of time until a relatively minor computer malfunction causes an SUV full of a family of 7 to crash into a school bus filled with 60 children, resulting dozens of fatalities and a grizzly crash scene. And a multi-billion dollar lawsuit against the deepest pockets: Google (or other supplier) and the unfortunate manufacturer that installed it.

     

    It only takes one. All the other statistics and probabilities become meaningless when the judge and jury is hearing the testimony from dozens of weeping soccer moms.

     

    Yes commercial airliners use the autopilot all the time. But they have many orders of magnitude less traffic and other stuff to deal with in flying along carefully designated and choreographed routes, with no unauthorized deviations from the flight plan. And triple or even quadruple redundancy in the flight controls and computers, so that catastrophic failure resulting in loss of lives and the airframe becomes "extremely unlikely" events - one in a billion type stuff. And automatic and immediate release (with loud warnings) as soon as either of two pilots moves a control to regain control of an errant autopilot.

     

    That said, there is something else to be said about automatic cruise controls and other driver-assist devices, that do intervene if an inattentive driver drifts out of lane, or towards the ditch, or fails to slow down when closing on another vehicle.

     

    Haven't seen any stats yet, but even the simple backup warnings that "beep" progressively more insistently when approaching an obstacle aft, which have been around for 10+ years, have most likely resulted in countless fewer accidents involving backing into something.
    14 Jul 2012, 01:15 PM Reply Like
  • Tdot,

     

    Good thoughtful response. Just as a point of clarification, autonomous vehicle systems rely on multiple redundancy (similar to or more robust than aviation systems). The systems are akin to space flight redundancy.

     

    GPS is one layer.
    LIDAR is one layer.
    Stored maps and images is one layer.
    Front and rear cameras are one layer.

     

    Each layer provides confirmation of the other layers. If one layer fails then the system may sense failure (may, not sure, haven't seen code or designs) and return control to the driver.

     

    A catastrophic failure of all systems is always possible and eventually likely. But with redundancy, you reduce the incidence of failure AND even if there is failure, you provide a backup (human operation) with warnings.

     

    That you assume this very rare failure would occur in an SUV on schoolbus collision, is the rankest speculation, and isn't born out by this happening with any frequency with human drivers.

     

    It's all going to rest on a "black box" concept generally acceptable in aviation, that provides the very best evidence for WHY an accident occurred.

     

    This will not eliminate manufacturer liability, but it will diminish it, and with (what I assume) will be millions of miles of accident-free driving as compared to countless deaths of human-driven cars, I, unlike you, am unconvinced that a jury (many who ride in ubiquitous autopiloted cars) will break Google's bank.
    14 Jul 2012, 02:15 PM Reply Like
  • Well folks seem to crash with those bright yellow school buses all the time these days - it is in the news on a fairly regular basis. Although admittedly, rarely are there massive fatalities among the school bus riders. The raised floor in school buses, being some 4 feet off the ground, seem to protect the kids from most routine crashes with cars and light trucks.

     

    But if something big and heavy comes along on the freeway out of control at high speed and collides with and forces a school bus off the road and over an embankment, those unbelted (!) kids are going to get tossed and kicked around like so many hacky sacks.

     

    At the same time, one of the most likely candidates for early adoption of full authority digital autopilot systems in road vehicles would be commercial trucking. Then come the taxis, limousines, shuttle buses, luxury vehicles and work trucks, where the additional costs can be "worked in" more easily. Eventually they arrive in commuter and family vehicles. It would be interesting to see how driver training and road tests for a driver's license might be changed. Again - probably easy for commercial truckers (CDL) and chauffeur licenses...

     

    The point is, once these robot drivers are on the road and actively guiding the vehicles under their control (and probably chatting with each other in negotiating for positions in traffic and warnings of trouble ahead), and then the one in a million, or one in a billion or trillion failure happens, the confused driver who has been relying on the robot may take several seconds to realize that something is wrong and a crash is imminent. And loud, startling warnings of equipment failure can be just as disconcerting, taking many seconds to comprehend and act on. The less often it happens can mean longer time to properly respond.

     

    I think we are almost on the same page. You must know that triple or quadruple redundancy in the guidance system does not necessarily refer only to three or four layers or sources of information (eg: GPS, LIDAR, optical and IR cameras, inertial guidance with gyroscopes, magnetic compass, dead reckoning with stored maps, etc., some more useful than others) Each guidance system must be independently considered all the way to the steering wheel, throttle, and brakes, with some sort of weighted voting process to choose the most likely "correct" response path.

     

    You don't necessarily want a GPS loss of signal, or a splash of mud on a camera to trigger an abort mission fault that suddenly returns control to an unwary driver. You also don't want an outdated map or temporary detour to send the vehicle careening onto an abandoned road or into a construction zone. I've personally seen multiple incidents where a GPS navigation system has indicated a turn onto a nonexistent road. Once it was onto a horse trail (if that) through the washes in the middle of the Mojave desert. Even Google Maps can be months or years out of date with new construction and temporary detours going on all the time.

     

    The bottom line is, if it is even remotely possible for something to go wrong, with some 100 million cars on the road on any given day, then that one in a billion probability of failure is statistically going to come up on a fairly regular basis. Yes it will be orders of magnitude less often than the 16,000 crashes a day caused by human error and stupidity. But when it does happen, and the bewildered driver and passengers head for their destiny in a runaway out of control robot vehicle, it will get a lot of news coverage, and gather lots and lots of actual and punitive damages for Plaintiff's Attorneys.

     

    The Severity in the Failure Mode Effects Analysis is really high on this one. You really do have to plan for the absolute worst case scenarios, and provide multiple countermeasures to prevent it from happening, and multiple more countermeasures to reduce the "damages".

     

    That said, yes absolutely I want one. As a pilot and a driver, would love to be able to explore the possibilities and limitations of the guidance systems. And I want a side-stick controller for steering and speed control while we are at it...
    15 Jul 2012, 09:04 AM Reply Like
  • I agree. One advantage I would point out is that a commercial truck with no driver would actually have an additional option to save a stupid human driver by sacrificing itself. It could take to the ditch and would have the control skill to lay over on its side and slide away from traffic.

     

    Once people can work, view news and entertainment, or nap during their commute this will become instantly must have technology. Two car families can become one car or motor pool families because after a trip the car can go by itself where needed next.

     

    Bus systems could become populated with more, smaller and more frequent buses. There will be more routes covered.

     

    You will be able to pull out your phone, launch the cyber taxi app, touch a frequent trip and time, then wait just a couple minutes for your ride. Ride sharing will be quick and optimal.

     

    Highway capacity will decrease dramatically. The savings in road construction will spur governments to give credits and set aside dedicated lanes.
    6 Aug 2012, 03:08 PM Reply Like
  • I can't wait then I can sleep during the 10 hour drive to see family!
    14 Jul 2012, 01:18 PM Reply Like
  • Driving Miss Daisy...

     

    ... without the racial backdrop.
    14 Jul 2012, 02:10 PM Reply Like
  • Unless tort liability laws in this country are radically altered, this idea is stillborn.
    14 Jul 2012, 02:42 PM Reply Like
  • Sad but true. God knows how many innovations die at the company's lawyer's desk.
    14 Jul 2012, 05:05 PM Reply Like
  • Huh? Less claims will make insurance rates for these cars nominal. Insuring yourself to continue human operation will become a luxury.
    6 Aug 2012, 03:11 PM Reply Like
  • I share the concern about lawsuits. But I think the big positive force will be the massive number of aging Baby Boomers.
    14 Jul 2012, 04:04 PM Reply Like
  • They're making this concept overly complicated...for greater profit, of course. San Diego experimented with a strip on the road that controlled the cars, providing smoother traffic flow, and Las Vegas has, or had buses that followed a strip in the street to eliminate the need for a driver. And many new cars are advertising various accident avoidance systems.
    But the key to this car is the expense. Put all the gadgets and programming into it and raise the cost to $100k plus and the auto makers are happy campers. The fact that this will delay mass implementation by decades is irrelevant. Profit now, and bemoan the "failure" of the system.
    14 Jul 2012, 04:30 PM Reply Like
  • Moore's Law will apply. They won't be terribl expensive.
    6 Aug 2012, 03:12 PM Reply Like
  • I want one!
    14 Jul 2012, 04:31 PM Reply Like
  • I would offer two thoughts:

     

    First, the marginal cost of a fully automated, self-driving car compared to that of a conventional car - once in volume production - will be fully offset by the reduced insurance cost over, at most, the first three years of ownership. (marginal cost something like $1,500; insurance savings $500 per year)

     

    Second, based on personal experience designing and implementing a rather primitive self-driving vehicle system over a decade ago, people will become very comfortable in and around automated vehicles in a very very short time. The system I did drove large, triple trailer class-8 trucks around a track, 24/7 to test paving. Within a week of beginning operation, people working around the system were comfortable standing just a few feet from a 40 mph, 76 ton vehicle under automatic control. This level of confidence was, in our opinion not warranted by either the the system design or the very limited operating experience and administrative steps were necessary to keep people away from the operating trucks. People only need to experience something work correctly a few times before the begin assuming that it "always works". This happens, at a gut level, even among thoughtful, highly analytical people, fully aware of the reliability limits of this kind of system. It is an effect truly profound to see...
    14 Jul 2012, 04:46 PM Reply Like
  • I would offer two thoughts:

     

    First, the marginal cost of a fully automated, self-driving car compared to that of a conventional car - once in volume production - will be fully offset by the reduced insurance cost over, at most, the first three years of ownership. (marginal cost something like $1,500; insurance savings $500 per year)

     

    Second, based on personal experience designing and implementing a rather primitive self-driving vehicle system a decade ago, people will become very comfortable in and around automated vehicles in a very very short time. The system I did drove large, triple trailer class-8 trucks around a track, 24/7 to test paving. Within a week of beginning operation, people working around the system were comfortable standing just a few feet from a 40 mph, 76 ton vehicle under automatic control. This level of confidence was, in our opinion not warranted by either the the system design or the very limited operating experience and administrative steps were necessary to keep people away from the operating trucks. People only need to experience something work correctly a few times before the begin assuming that it "always works". This happens, at a gut level, even among thoughtful, highly analytical people, fully aware of the reliability limits of this kind of system. It is an effect truly profound to see...
    14 Jul 2012, 04:46 PM Reply Like
  • Automated trucks for container handling at Rotterdam:

     

    http://bit.ly/NXVoTk
    14 Jul 2012, 06:41 PM Reply Like
  • Its hard to imagine what the Palemetto Expressway in Miami would be like with no delays or accidents...and where else could we train our nations future Nascar drivers?
    14 Jul 2012, 05:18 PM Reply Like
  • The economics of self driving cars don't add up on the scale of existing vehicles.

     

    Think about it.

     

    Why blow 50k + interest on a vehicle which is going to sit in a garage all day when that vehicle is capable of driving itself? It can be out earning money driving other people instead of sitting in a garage. You just call it 10 minutes before you need it on your smartphone.

     

    So the only market for self driving cars is the taxi market in which this would wipe out the incumbents by removing the driver as a time based cost.
    When you think that a taxi can make 40 journeys per day to a privately owned vehicle's 2, there are some interesting second order effects to observe and profit from shorting market segments: road infrastructure reductions, city center real estate price drops, auto companies and their suppliers, banks
    14 Jul 2012, 06:25 PM Reply Like
  • I like your ZipCar on steroids idea. I would definitely give up my car for that.
    15 Jul 2012, 09:54 AM Reply Like
  • Now if they could just make it fly too...
    14 Jul 2012, 08:01 PM Reply Like
  • Remember, all you nay sayers, you will be old one day too. The day will come sooner rather than later when you will lose your drivers license and be imprisoned in your house (if you live in the US and outside the handful of urban areasthat offer mass transit and plentiful taxis)

     

    Reflect on this tale - "A grand mother is living with her son, daughter in law and grand daughter. The grand mother keeps dropping her bowl of noodles and breaking the bowl. The daughter in law loses her temper and sends her daughter to buy a wooden bowl for the grandmother. The kid returns with two wooden bowls. The mother asks "I sent you to buy one bowl for your grandmother, so why did you buy two?" Replies the kid "The other one is for you when you grow old"
    14 Jul 2012, 08:53 PM Reply Like
  • Exactly.

     

    The self-driving car is one of the very best "demographic bets" you can make.

     

    We've built an infrastructure that demands automobiles -- but humans, who are not great drivers in the best of circumstances-- become dangerous as they age. The Google answer makes sense.
    14 Jul 2012, 10:10 PM Reply Like
  • Self Driving cars will be implanted with chips that the government first uses to tax you on miles driven, then to regulate your speed..... and later to "direct" you to your government assigned activities to help the state.

     

    I don't care how damn old I get - I'll be steering and pushing the brakes and gas pedal until the day I die.
    15 Jul 2012, 12:25 AM Reply Like
  • Gotcha. With 40,000 traffic fatalities in the US each year, your chances of that happening to you are 1.49% or 1 in 67.
    15 Jul 2012, 09:17 AM Reply Like
  • Thats fine - who says life is without risk?
    15 Jul 2012, 02:28 PM Reply Like
  • I sincerely doubt that cars will become 100% self-driven in the near future due to serious infrastructural deficiencies facing our nation. A substantial portion of paved city roads is in utter disrepair. Lane markings in coastal communities often appear faded, exhibiting signs of systematic neglect from state and local governments...
    15 Jul 2012, 01:44 AM Reply Like
  • You are behind. They already deal with this without requiring special infrastructure. They will share data. Precise inertial sensors will permit collaborative adaptation with rapid convergence to perfect navigation of any route that the cars use.
    6 Aug 2012, 03:17 PM Reply Like
  • There are supposed to be lane markings!!
    6 Aug 2012, 03:43 PM Reply Like
  • I imagine a "next step" along this path is establishing "commuter convoy lanes", sort of like the "high occupancy vehicle" lanes seen around many larger cities.

     

    "Smart" autopilot cars would line up NASCAR-style on the freeway, drafting on each other in designated express lanes, transmitting convoy speeds, road hazards, and status to each other; potentially saving enormous amounts of fuel, and avoiding collisions with each other. And virtually eliminating Road Rage for the participants, at least until the whole line comes to a grinding halt because someone panicked and slammed on the brakes because The Robot missed their exit. Or someone snuck into the convoy with an "unSmart" car and really fouled up the skunk works.
    15 Jul 2012, 09:25 AM Reply Like
  • It might be convenient for the government to regulate fleets of these vehicles that would be owned by large corporations, similar to the way airlines are currently run. No private individual would own their own "auto-pilot" car (except for perhaps the rich) and accidents would be investigated by the NTSB. In this way they could be shared as needed, perhaps even occupied by several persons on each trip. The cost of each trip could be determined by the number of passengers, distance & comfort level of the vehicle. So there might be several price classes of these vehicles, each price class determined in a manner similar to the way airlines currently determine price & classes.

     

    Imagine not having to own an expensive, depreciating mode of transportation anymore. No more registration fee's, taxes,Insurance, fuel or maintenance costs either. Of course, the cost of using the vehicle would reflect this, but it would seem that over time the "economies of scale" would make it affordable for all.

     

    In the not to distant future you might use your smart phone to "request a ride", a computer at the vendors local office would calculate the nearest available vehicle and it would be dispatched to your location.

     

    Whats not to like?
    15 Jul 2012, 11:50 AM Reply Like
DJIA (DIA) S&P 500 (SPY)