Fun Friday: Self Driving Cars

I saw this projection from BI Intelligence below. It suggests we will have 10mm self driving cars on the road by 2020. They define “self driving” as “as any car with features that allow it to accelerate, brake, and steer a car’s course with limited or no driver interaction”. The Telsa that my wife and I drive fits that definition.

sdc installed base

It is unclear to me whether this is a global number or not. I assume it is. There are over 1bn cars on the road around the world, so that would be 1% penetration. That seems low to me.

Do you think “self driving” will have penetrated more than 1% of the world’s car population by the end of this decade? I do.

#machine learning

Comments (Archived):

  1. Mark Essel

    That curve looks pretty steep already. Assuming all the big auto companies can deliver models with features like your tesla, it’s probably a reasonable estimate. I’ll guess it’s on the low side.Also depends on whether Apple or Google release and can manufacture a popular vehicle by then. That could swing the market penetration higher. After watching your morning video interview, I agree the tech companies have the advantage.What about higher powered personal drones? Who needs to drive when they can be flown around faster?

    1. pointsnfigures

      The promised me flying cars and all I want is cell phone calls that don’t drop

      1. Jess Bachman

        Next thing you will want is flying cars that don’t drop…

  2. Ben Kinnard

    Assuming all future Tesla’s count as self driving (or is it optional on Model 3?), they alone would account for approx. 2 million of the 10 million vehicles in 2020.Add in every other car maker in the world and 1% seems pretty conservative.Also a lot of companies will probably rent their vehicles out in the future, lowering the barriers to adoption and getting more self driving cars on the road sooner – Chamath Palihapitiya talked about Tesla being one of the first HAAS companies (Hardware as a Service), completely changing the business model for car manufacturers

    1. markslater

      Absolutely – as a city lover i loathe owning a car – i’d simply like to hail a service

  3. Christian Noske

    The number of self driving cars might not be the most relevant number we should look at, it would be interesting to see the miles driven by self driving cars by 2020. Today’s cars have under 6% utilization, self driving cars will have a much higher utilization and will represent a higher percentage of overall driven miles. But I also believe 10mio cars is too low.

    1. ShanaC

      Why miles?So I guess one of the real questions is penetration first for everyday use or for something like mass transit/goods transportation use?If everyday, if you don’t have /have a very mild commute penetration would be roughly linear with miles driven. That group would be my parents. They both work from home, and use thier car for errands that are a little too far to walk/slightly easier with a car,like weekly groceries. That’s it. The miles driven curve for this hockey sticks as units sold hockey stick.Mass transit or goods transport wouldn’t be liberal. That would have hockey stick miles driven, since every unit added goes a lot of miles quickly until saturation. But that is its nature. You could have slow linear growth of units, but miles driven would disguise that.My guess: you won’t see hockey stick growth of units unless the first group is also in it.(People who commute around 45 minutes each way are wildcards in this because I’m guessing thier behavior may or may not change based on which comes first to really penetration, commuter or everyday use)

  4. Ales Spetic

    If the number is compared to all existing cars its in my view quite optimistic. Cars have a product lifecycle of 6y and they exist on the roads for another 10y at least. Since not all cars will be shipped by this features by 2020, only a portion of new cars will count as self driving cars…adding all that, shipping enough cars in the next 4 years with self driving feature to add to 1% of total car population is a lot. i do believe that a significant portion of new cars will have that feature (e.g. 20%), however cars have a long lifespan to shift the numbers.

    1. Ryan Anderson

      Great post, and where my head went immediately.Say 6% of cars on the road are new, you would see ~25% of the total car population churn between now and the end of 2020. Self-driving cars would need to be 4% of all new cars sold between now and then for that to happen and, to my knowledge, only 1 manufacturer with 0.1% of total market share currently offers that feature.

    2. Ana Milicevic

      This is really Tesla’s biggest innovation – the over-the-air car update. Perhaps to that end other car manufacturers will develop ways to ‘upgrade’ their cars without necessarily requiring a full 6yr+ lifecycle change.

      1. Brian Bensch

        The over-the-air upgrade is great, but making a car able to drive autonomously is clearly not just software. The first 2 years’ worth of model S’s cannot upgrade to autopilot mode, because they lack the hardware. I agree with Ryan above, it seems optimistic considering Tesla’s the only carmaker doing so today, and even Tesla’s wildest ambitions are to sell 500K cars/year by 2020, a tenth of the 5M projected being sold in the chart above. Not impossible, but optimistic.

  5. Theodore Mollinger

    Estimates indicate that in 2020 about 100 million vehicles will be sold, i.e. c. 10% of new cars could be self-driving. New vehicles sold is a better metric than total cars. Retrofitting old vehicles with self-driving technology adds a lot of friction and cost to end-users.

    1. Guy Lepage

      Agreed. The number of new vehicles sold is the metric the car companies look at. Unless there was the ability to retrofit a legacy vehicle. Now that would be a game changer.

      1. ShanaC

        It will be doable. If only because there will be demand for it

  6. Antonio Rocha-Ferreira

    They are stupid. Self driving is already becoming a feature and by 2020 it will be available or in the production pipeline of all major players. 1%? I would say 20% at least, but by 2030 ALL cars will have it (NEW CARS). But Ales is 50% right, it will take a while for the old industry to crumble, Tesla is the only one treating this as a software upgrade, all others are years away.

    1. PeterisP

      If every major player equips all their new cars beyond the low-end budget line with self-driving feature, then it comes up to something like the quoted 1%. The vast majority of cars in use are nowhere close to new, and worldwide sales are heavy on very basic car models.Also, autonomous self-driving cannot be a software upgrade – it requires installing extra hardware (sensors and computation, for most cars actuators as well) beyond what the current (including Tesla models) have, so it will be generally limited to new cars. You can improve existing automation capabilities with an over-the-air upgrade, but making the car able to detect things that it currently cannot (e.g.… requires installing additional (and, for now, expensive) sensors.

  7. William Mougayar

    How about special lanes for these self-driving cars- what’s the prediction on that? And will the self-driving cars be only allowed inside special lanes, and required to stay under a certain speed limit?

    1. Matt Zagaja

      I’ve heard that because self-driving cars are so safe they might be able to create special high speed highway lanes for them to shorten commutes.

    2. LE

      That’s one of the things that will impede self driving cars. The fact that they will drive cautiously by design will drive some people crazy I predict. Like having old people in a Buick in front of you or in a parking lot.

      1. William Mougayar

        It could be the other way around.

  8. jason wright

    the electric car is just so much more important than the self driving car.ride a bicycle.

    1. Alex Murphy

      Electric cars where the electricity comes from coal powered plants are not that much better, if they are, than internal combustion engines.Regarding energy, the most important advancements to come are Solar along with amazing storage (batteries).While advancements in Energy are super important and will save the world, Self driving cars will change the world as well. One study out there says the average American drives over 37,000 hours in their life. That is 1,541 24 hour days. Take out 8 hours for sleeping each and it is over 2,300 days, or 6.33 years.Imagine what you could do with an extra six years of your life.

  9. Paul M

    I don’t think they hit this by 2020 (only 3 years, 7 months away). As Tesla ramps up production of the Model III it will demonstrate the usefulness of the technology to a mainstream audience, and I think the 20s will be the decade the technology becomes demanded the way seatbelts and then airbags became required. My selfish interest (as a 48 year old) is that by the time my cognitive skills go into decline in my 70s or 80s I have little doubt that self-driving cars will be ubiquitous.

  10. Todd Savage

    The biggest issue with self-driving cars is their interaction with humans. The interaction with human driving causes problems/accidents (human stops to soon while computerized car does not – this actually causes accidents). I heard from Eric Schmidt in a very small setting that this is the biggest issue with self-driving cars right now.

  11. JimHirshfield

    In time for my kid’s legal driving age, please.

    1. markslater

      Driving? (s)he wont need to 😉 there wont be anything called “driving age”!

      1. JimHirshfield

        And there won’t be 1000% insurance premium increases when they hit that age, amirite?

        1. LE

          Put on the rain gear because the insurance companies will soak you either way.

    2. Jess Bachman

      “You get that new Fast and Furious app for your car? It’s so intense, I turned the drift settings up to 11.” — Overheard 2020 high school conversation.

      1. JimHirshfield


    3. ShanaC

      I knew someone would say that.Teenagers the scary drivers

      1. JimHirshfield


    4. Twain Twain

      They’ll have IoT enabled cars, Jim — where there are so many sensors in the boot …There’s no space for shopping bags!But, hey, by then everything will be Ubered or Amazon-droned to us so no need for us to have bootspace for shopping!

      1. JimHirshfield

        I think we’ll just stay home

        1. Twain Twain

          Whilst Musk is on Mars.

  12. falicon


  13. BillMcNeely

    If I get on with Maven I’ll let you know how it’s working out

  14. kirklove

    Oh gosh I sure hope not.Now if you’ll excuse me I’m off to find where my uncle parked his Red Barchetta#hater

  15. andyswan

    It’s a phenomenal advancement that will save thousands of lives and freshen the smell of taxis.

    1. Mark Essel

      For me the “aroma” is one step closer to home.I’ve been pavlovian trained to think about getting home when I’m in a taxi (I walk 99% of the time) but on late work dinner nights in NY all I can think about is getting home and passing out (about a 2hr commute door to door, and I’m an early sleeper so those days are rough)

  16. pointsnfigures

    I think self driving trucks and buses will grab hold even faster. I heard an interview with the GM CEO last year. She thinks it will happen but might take a bit longer than people expect. Her point was we need infrastructure modification to really make it all work. In Detroit, they are building miles and miles of roads embedded with sensors to test, improve and develop the tech that goes into driverless cars.If you lived in a city with driverless cars that you could page on your phone, why would you even own one?

    1. Susan Rubinsky

      Yes, there is lots of talk in the transit sector on this.

    2. markslater

      Exactly. THe entire concept of transport is going to be re-imagined.

    3. Bill Seliger

      Agreed on trucks – they will be the market leaders here. It will start with platooning and other technologies but fully autonomous will be here very soon. The payback is phenomenal – a transportation analyst at MS released research recently that showed a less than 6-month payback. Utilizing a truck 24/7 versus current allowable hours of service will drive this, and there’s a driver shortage as it is. Trucks also turn over faster – most cars on the road are much older than trucks so the adoption will be much quicker on trucks. This will cause trucking industry consolidation – that will start with electronic logging rules going into effect soon and the higher fixed costs of fully-autonomous trucks will keep it going.This is no longer a technology problem but an adoption problem. The first time someone is killed by a self-driving truck it will be headline news but the other 5,000 people that are killed annually in truck accidents are invisible.

    4. SubstrateUndertow

      <bloclquote>”Her point was we need infrastructure modification to really make it all work.”</bloclquote>Where is that money going to come from? Basic road infrastructure is crumbling with little response at present.

      1. ZekeV

        Self-driving car tax? Yay!

        1. pointsnfigures

          Yes, the multiplier effect of government spending is close to 0

          1. SubstrateUndertow

            You mean like the money spent on the interstate road system after WWII or universal public education or space exploration or military expenditures etc. . .Maybe you are overstating your case ?Legitimate collective expenditures create complex multiplier effects that are often difficult to measure.

          2. pointsnfigures

            not really. where do the dollars come from? Universal public education is a total cluster you know what. Space exploration looks like it can be done better with private companies. The military is the military-can be run more efficiently but read the Federalist papers for the reason why we need it.

      2. pointsnfigures

        governments still spend billions of dollars a year on roads. as they redo them, the upgrade. problem is right now there is no sure path to the right sort of upgrades-hence the test roads they are building in Detroit.

        1. ZekeV

          wonder if there will be some more hwy privatization, convert state highways into tollways run by Google. the new owner invests in self-driving infrastructure, and gets to charge tolls.

          1. SubstrateUndertow

            Doesn’t Metcalfe’s law sort of imply that less than an “all-roads” road network approach would be very inefficient for all those self driving cars? All those optimized/isolated new-road-network-nodes bottle necked by all the surrounding/interconnecting old-road-network-nodes?

          2. ZekeV

            Probably. I’m not a fan of toll roads, myself.

          3. Lawrence Brass

            If self driving cars free drivers from driving, how do you think Google could monetize the liberated time and attention span? Highways may end up looking like Times Square and you would get ads run on the car dashboard, no tolls required.

          4. ZekeV

            Then Google can capture a larger and larger portion of our activity, and sell ads against it. Driving cross-country, Google gets to influence where you go for supercharge, snacks, and motel. I’m really torn b/c on one hand, robotic cars will save so many lives. Driving is stupid and dangerous currently. I feel guilty every time I put my kid in the car to go to the grocery store at significant risk of death or permanent injury. Is that bag of organic chips worth a .001% chance of an accident? Significantly reducing that risk will make our lives so much better. On the other hand, it’s a faustian bargain where we give up our autonomy in exchange for safety or other benefits. I guess that’s a bargain we’ve already made though.

    5. LE

      If you lived in a city with driverless cars that you could page on your phone, why would you even own one?If you could have a car that would shift the gears for you why would you buy one where you had to shift? (I like manual transmissions but most people don’t).On a serious note, because owning gives you an extra level of convenience. For one thing you get the same predictable car that is not going to be altered by others (smell, dirt etc.) For another you get to keep things you own things in the car. It’s there right in your garage ready and waiting for you.Why do people who can afford to do so purchase vacation homes when the cost of staying at the nicest hotels (with more offerings and services even) is typically less? [1]Now of course if you are talking about “a city” like Center City Philly, or Manhattan or Chicago and you are only going to drive around the city and not go out of the city perhaps you don’t need a car but that’s no surprise people take cabs or other forms of transportation (now not available in Austin Texas).[1] To many reasons to list. One is that there is no checkin and checkout time, space, having your own things there, pride of ownership, idea that it might appreciate in value (often dubious) etc. Not having to plan or make reservations being able to go at the drop of a hat. Not having to deal with other people. Otoh there are plenty of people who like to stay in hotel rooms at least for short periods of time.

      1. pointsnfigures

        I could foresee a future with 100% electric cars inside cities that were driverless and owned by an asset company. Imagine that, no cars parked on city streets-and no noise caused by engines or horns. A city dweller would page a car to take them anywhere, sort of like they do with Uber now. No need to build high speed rail since driverless high speed buses would be cheaper and more efficient. No need for the huge amount of parking spaces because driverless assets would be utilized better. Less dead weight loss in the economy because of all kinds of factors (money reallocated from insurance to other things etc) As tech gets cheaper and more convenient, driverless cars take over the burbs. Already I know families that use Uber to shuttle their kids from school and activities.

        1. LE

          So if cars are not parked are they constantly moving (like musical chairs) to the next pickup? I can see this being good for city dwellers I am not doubting that. Of course it’s kind of hard to see that in a dense city like Manhattan there would be enough assets to satisfy the demand (but I guess that’s a good problem for the asset company to have). One thing though. Car engines don’t really make a great deal of noise today it’s not the 1970’s (remember that).

        2. sigmaalgebra

          Nearly nobody really likes public transportation.For decades, I’ve heard NYC people run down cars, like they really HATE cars. Okay, but NYC still has lots of cars. In NYC, really hating cars is just one of those style things.

          1. ShanaC

            Up to a point you also don’t need one here.If you stay local for hobbies/friends/family and don’t have a huge family, it’s doable for your entire life.If I were talking to my family very regularly (long story) I actually could live like this. Everyone who lives within the nyc metro area, barring one person, is commute via public transit only in decent amounts of time. That one person actually also might be as well : I don’t know the area he lives well (in NJ. I know there is bus/train service to that community, but once I’m there I would have no idea if it’s walkable to his actual house)

          2. ShanaC

            Mainly because it’s bad. Not because cars are good. If public transit is good, cars look less interesting. As areas urbanize, good public transit looks more necessary and tends to get funds to implement.So really is the move towards moving towards urban cores, especially in small cities, a permanent long term trend or a fad. If it isn’t, expect more public transit to pop up

          3. sigmaalgebra

            > Not because cars are good.My main car is terrific. I can think of no good alternative.

        3. James Ferguson @kWIQly

          Efficient Bus != High Speed TrainThat is if you have a decent rail infrastructure

          1. pointsnfigures

            America has a socialized passenger rail system that doesn’t work very well-plus airplanes are more efficient for long hops. Because of regulation etc, our rail structure is highly congested. Freight clogs most of the rail in Chicago, the hub of the national rail network. Efficient Bus on existing highways is cheaper. It would be easier and cheaper to add a third lane to entire interstates point to point where you’d think high speed rail would work. For example, Chicago to Champaign.

          2. ShanaC

            Trains work really well for suburbia to urban commutes.I can see the self driving cars at the terminal points (the town at one end and the city at the other)If rail infrastructure improves at the same time, I can also totally see exburb commute times dropping radically in a system like this. It would be a huge game changer for those towns and cities involved (eg:nyc vs Suffolk County in long Island and upper counties in the hudson valley)

          3. Mark Essel

            I’d love this. My commute days are long.

          4. ShanaC

            how long are we talking about?

          5. Mark Essel

            leave at 5:40, get home 7:20. Some good walking. Some pushy waiting. Some work :)Mon/Wed are NY days, work remote the rest of the week

          6. ShanaC

            I think that you are officially considered a mega commuter when you commute…

          7. Mark Essel

            Hah. Interesting that there’s a bucket/label for it.

          8. ShanaC


      2. ShanaC

        With enough self driving cars it isn’t really convenient to own. Same with regular cars. Hence why zip cars exist.Self driving zip cars could be even better. You could have it show up at your door, and then wave goodbye to it. The whole checking in/out the cars would cease to exist, as well as centralized “car homes”. They could Park on the street.Plus this version of zip cars could automatically redistribute the cars in a given metro area to serve traffic needs because the cars drive themselves.I’d go for that in a heartbeat

    6. sigmaalgebra

      > Her point was we need infrastructure modification to really make it all work.Yup.

    7. kevando

      I love this. I always thought Detroit would be the perfect city to test self driving cars. Rich auto history, lots of urban infrastructure, and very little people!!

      1. Alex Murphy

        The people in Detroit are the same size as other people.

  17. markslater

    I am truly waiting for Apple’s move. It feels to me that this collision of Self-driving and “electric” presents an opportunity to birth a whole new mode of transport – The car as we know it (front facing, engine, even things like brake lights etc) is going to be completely re-imagined.Bolting on features and widgets to the existing concept that we know as a “car” is far less interesting.

    1. Eric M. Seitz

      Yes, yes, yes. There are endless opportunities to rattle the concept of what a standard vehicle is and the businesses of selling and servicing them. The more fresh players the better. Holding onto my 2007 Nissan, and riding bicycles more, until serious disruption takes place. Props to Tesla for their progress.

    2. Alex Murphy

      I think Apple and Godot are on the same bus …Apple will not be the innovator here. Apple was not the innovator, it was Jobs. Jobs was the one that drove different thinking. Right now Apple is mostly ‘Thinking the Same.’

    3. ErikSchwartz

      I kind of expect Apple’s car projects will be like Apple’s TV projects.We will talk about them for decades (literally in the case of Apple is making a TV set). Apple will have lab stuff going on. But getting the research productized might not come soon or be what we expect it to be.

  18. MelkiSch

    Industry players always talk about 4 different levels of selfdriving carsLevel 1 – Cruise controlLevel 2 – Adaptive cruise controlLevel 3 – 99% automation, the human is just supposed to watchLevel 4 – no humanThe number one obstacle for a 1% penetration by 2020 is going to be regulation on level 3 and level 4 according to most players.I believe your 1% penetration is low for level 1 and 2 but very high for 3 and 4.

  19. Ana Milicevic

    As with most things on march towards more automation, there’s an interim step of the technology augmented driver which I’d argue we’re already heavily in (e.g. lane shift warnings, break assist etc). I’m puzzled by the dominance of the autonomous car conversations over autonomous public transit and autonomous commercial freight. The biggest barrier to adoption is not so much the technology as the physical infrastructure: I don’t want to subject any vehicle to NYC’s shitty potholed roads and separate lanes (perhaps HOV lanes at certain times) seem to make a lot of sense for automated freight if you don’t take into account what destructive effect that will have on the actual physical road. I don’t think that transportation in the future stipulates individual cars the way it does today so it’s an interesting exercise to peg penetration for something that is undergoing a macro shift.

  20. JaredMermey

    Autonomous Computerized Electric (ACE) vehicles are the future. The convergence of those three features allow for the mobility businesses of tomorrow (when true driverless both exists and is legal) to be built today. When the laws catch up with technology’s progress, the software will already be built, tested and in the market to manage millions of cars in real-time.For example, see Tesloop in LA.DISCLAIMER – I know these guys 🙂

    1. Tyler

      Agreed. It’s the regulation that will slow things down.Down here in TX, some time ago we had politicians trying to pass laws saying a car mfg had to have a dealership in the state in order to sell cars (ie can’t purchase online (Tesla)). I’d imagine we’ll be near last in legalizing it, even though we could desperately used it….

  21. ErikSchwartz

    It is a very long journey from cars with self driving features to autonomous cars.Most of the tech media conflates them.

  22. LE

    I saw this projection from BI Intelligence I don’t put any stock in this guess based on the source. Good topic for discussion of course. My own “guess” is that it will be higher than 1%.The “senior research analyst” that wrote this is perhaps 2 years out of college (per his linkedin profile) Good move on BI’s part (admirable) for them to try to market things like this.It is unclear to me whether this is a global number or not. I assume it is.Just purchase the report to find out.

  23. SubstrateUndertow

    Are we discussing “self driving” or “diving assisted” cars here ?It seems to me that true “self driving” cars will like natural language interpretation turn out to be a much tougher problem than anticipated especially when we start drill down on the full spectrum of contextual complexity inherent in the transition to full implementation?

  24. MHSzymczyk

    BI Intelligence also predicted 21 Million units for Google Glass by 2018 – http://www.businessinsider….

  25. bijan

    More than 1%? Hell, yes!

  26. jason

    Here’s my bad estimate of the upper limit of adoption:According to http://www.worldometers.inf… there were 60MM new cars made in 2012. So if we are somewhere in that production neighborhood in 2016 (say 70MM/yr) and you added self driving to every new car starting today then 4 yrs you could have 280MM cars. If there are 1.x Billion cars left on the road, that gives an upper limit of about 25% by 2020.That said, once the Big Boys develop the technology, the marginal cost of adding it to a new line of cars is probably small compared to the loss of market share from their competitors adding it first. It seems adding limited-self-driving is primarily a lawyerlawyerlawyer problem, as opposed to a technical problem, though.

  27. LE

    Well as long as we are making wild predictions I think that if there are more cars that are self driving, that allow an easier commute, people will have less of a reason to locate in a city because getting to your job from a far out suburb (with more room, lower prices, and better schools) will be more attractive if you don’t have to directly deal with rush hour traffic stress. That’s even when traffic doesn’t flow better (and it will in theory).This is an example of how hard it is to predict the future. There is a big difference between driving in rush hour, and driving the same amount of time on a road where it’s not rush hour and just light traffic. [1] Make sense? This is what self driving cars could bring. More sprawl. Housing could be further out. Just like the interstate system in the 60’s prior to rush hour developing. I remember when I95 was built and how it influence where my parents moved when we were very young. My Dad could get to his office in the same amount of time in a nicer area with less aggravation (until others followed).[1] Most drivers feel entirely different about 30 minutes getting on a highway with little traffic vs. 30 minutes driving through a city (or a combo with stop lights, turns etc.). Different experience.

    1. Lawrence Brass

      Leaving NYC last november I had the brightest idea. Had to checkout at 11 AM from the place we were staying and had to check-in at the airport at 7 PM, plus we were carrying too much baggage. The obvious and boring thing to do was to call a cab and spend the afternoon at the airport, but no. What I did, which for some members of my family is known as “doing a Lawrence”, was to quickly change plans and rented a car so we could go to B&H to get those last things I needed to buy and then have lunch somewhere. What came next was 4 hours of driving through the city under the rain, looking for a place to park near the store that was unavailable. The whole city was a traffic jam, I realized then why the subway is so relevant to new yorkers. Under stress people don’t smile that much, so the smart thing to do by 4 PM was to rush to the airport. Thanks Google Maps and Waze, you saved my ass.

      1. Richard

        Why didn’t you just stow the bags at the hotel, shop / lunch and uber to airport? Or hire a black car / driver for the day?

        1. ShanaC

          Many people don’t know you can do that?

        2. Lawrence Brass

          We couldn’t stow the bags there because it was an airbnb rented place, new passengers where about to move in. I could have stowed the bags at the coworking space we rent in midtown, but I was going downtown. I thought about the black car but it was a bit expensive and not readily available, instead I went to a Hertz place nearby and got a car, I have good discounts there. Not finding a place to park downtown was the breaking point of my optimal plan.

          1. Richard

            I’m a probabilistic and like to assign probabilities to everything. The probability of you finding one or more parking spots was quite low. One option was to rent the car and leave the car at the car rental location and enjoy the day. Though getting to the airport by 7, means you would have to get back to the car rental place by 430 at the latest, hope on no major traffic, refuel, return car…

  28. Ro Gupta

    as christian, who works for bmw ventures, says below..we should probably care more about a chart showing miles driven. consumer autopilot features are an interesting novelty right now, but self driving taxis, delivery and freight is where a lot of the commercial focus is on, and make up an outsized portion of passenger/cargo-miles transported

  29. sigmaalgebra

    A car with a gas pedal that sticks — wow, holiday for the lawyers, huge law suits, massive recalls, really big bucks.accelerate, brake, and steer a car’s course …So, if any itty bitty teeny tiny thing goes wrong and causes an accident, again, holiday, suits, recalls, really big bucks.Well, for the self driving functions, need some servos. They have to be really good, really, really good, much better than the auto industry has been used to making or holiday, …, big bucks.Likely the servos will be electrical, and on gas powered cars that will mean a much stronger electrical system — more money to buy and maintain.Beyond the servos, need some sensors, basically some version of radar and/or vision — more money — more money to buy and maintain.Maintain? Gotta be kidding: In the auto industry? For that electrical and electronic stuff? Those guys don’t understand Ohm’s law yet.Auto insurance? As the auto insurance industry gets some data, I doubt that self driving will make auto insurance rates go down. So, my guess — rates will go up.Utility? What utility? We’re talking a super version of cruise control, and that was never very popular.Old story: Just because you can doesn’t mean you should.A fundamental challenge in business: There are lots of nice products/services can provide, but they cost so much the number of customers is too small. There are lots of products/services that can offer at low prices, but nearly no one wants that junk. So, the fundamental challenge is to offer a product/service that a lot of people want and that can be provided, with a profit, at a price target customers are willing to pay.Just because it’s possible to build in such self driving stuff doesn’t mean that people will actually want it enough and find it useful enough to pay for it in purchase price, insurance, and maintenance. So, to me, self driving fails this fundamental challenge.All of that said, as soon as there are accidents, states will pass some laws restricting or blocking self driving.To me, self driving looks like a fad that will soon go the way of nearly all fads — soon flop.And to me, the electric cars self driving really needs will soon flop:(1) Lithium. Due to a lithium shortage, the prices will be too high or that part of the industry just won’t scale.(2) Charging Times. People will get tired of the long charging times, too long at charging stations and much longer at home. Too many busy people with money enough for an electric car will conclude that they are too busy too have car charging time planning constantly on their TODO list and conclude that an electric car is too much botheration.(3) Greenies. Throughout history, there have been many religions including earth worship. Well, the Greenies will get tired of earth worship, tree hugging, etc. if only because that love seems totally one sided.(4) Climate Change. The claims that humans are causing climate change are just total wack-o, and soon enough people will give up on it and significantly already are.Debunking the CO2 stuff is trivially easy — back 500,000+ years, the record of temperature changes and the record of CO2 concentration changes strongly contradict the claim that the CO2 changes caused the temperature changes. Done. Stick a fork in it.The drivers of the wack-o have been (A) the earth worshiping Greenies, (B) major parts of the media, e.g., NBC, the NYT looking to create anxiety, feed it with continuing articles with the wack-o, and, thus, get eyeballs for ad revenue and some political power, (C) people looking for Federal subsidies, e.g., the people selling solar panels and wind farms, IIRC Musk and Tesla, etc., (D) the UN that wants to be the head of a new world government, (E) politicians looking for power, money (shake downs, kick backs), ways to hurt the US, (F) companies hoping to hurt competitors, etc. Well, too much of this stuff is on the way out the back door and into the dumpster.(5) Gasoline. Currently, and, really, from a lot of evidence, the global economy is now just awash in oil and gasoline. And, if we want to get serious about gasoline, there are ways just to make the stuff. A gallon of gasoline is one heck of a good way to have a lot of energy for a car. Net, gasoline powered cars are not at risk of being replaced with all electric cars.Also, the market for gas cars is very price competitive. IMHO an extra $100 on the price can lower sales significantly.And, adding self driving to a gas powered car is more expensive than adding it to an electric car. So, that there are some self driving electric cars is not much promise that there will be many self driving gas cars — people just won’t want the extra expense.There may be some niche applications for self driving, e.g., maybe for some long haul trucks on long trips on very high quality roads in very good weather.People don’t, very much do NOT, like the hassle of car maintenance. E.g., my wife and I out in Dutchess County, NY bought a new Chevy S-10 Blazer from a dealer near my office in Westchester County, NY. After some weeks, we were supposed to take the car back to the dealer for some initial maintenance. So, at about 10 PM, we got into our two cars and drove the Blazer from Dutchess to the dealer in Westchester, parked the Blazer, got some maintenance request form, checked that we wanted “new car service” or some such, and dropped the form and keys in a slot. Since we got there at about 11 PM, we were in a hurry, and I just checked the box. Normally I wouldn’t do such a silly thing and, instead, would have considered each maintenance item one at a time.A day or two later we repeated the trip but this time during business hours. The charges were too high. So, just the logistics required six one-way trips between Dutchess and Westchester with each trip using a car and a person. And we had to do without the Blazer and write a check. Bummer.But all that was not the worst of it: When we got the Blazer back, at night we couldn’t see. So, soon enough (I have a background in cars; e.g., once was made a Full Member of the SAE) I concluded that the dealer had changed the aim of the headlights.Checking, right, the form for the standard “service” included lubricating hood hinges, aiming the headlights, etc. Nonsense: That maintenance list was busy work. There was nothing wrong with the headlight aiming; the aiming was fine from the factory and still fine. So, the aim change was a case of maintenance induced failure or the reason for the advice, “If it ain’t broke, then don’t fix it.”.So, no doubt the dealer got some doofus trainee and told them to aim the headlights. The doofus had the headlights on high beam but aimed them as if they were on low beam. We’re talking a total doofus. And in this case a dangerous doofus.So, at night on high beam we had low beam visibility, and on low beam we were nearly blind. Well, sure, we could take the car back for a re-do, right, another six one-way trips. Bummer.Well, we lived with the bad doofus damage until Christmas when we wanted to take the Blazer to my wife’s family farm in Indiana. No way did I want to drive the Blazer with that doofus damage all the way to Indiana and back in winter weather. So, the night before we left, with my wife, I drove the Blazer to a lightly traveled, straight road and aimed the headlights myself, just by guessing at what looked right.Yup, at times opposing traffic flashed their high beam lights at the Blazer as if my aim was too high. Okay. Sorry ’bout that.Then one night we were in traffic, and the car ahead was a police car. They pulled me over and were PISSED. They accused me of following them with my high beams on. It took the full story above, with my very credible looking wife agreeing, to get them to believe that my lights were on low beam and get the police to calm down and let us go, which they did.Moral: The US auto industry can’t get headlight aiming right. That the US auto industry can get self driving right is asking too much.I’m reminded of the four wheel drive control on that Blazer — it was mechanical. Well, soon Chevy went to an electrical control. Right, for a few years, it failed frequently. It took Chevy some years to get an electrical control of four wheel drive to work. Chevy to get self driving to work? Gotta be kidding!Eventually that Blazer had some dashboard lights burn out. Well, to replace them, had to remove and reinstall the whole dashboard, $1000 in labor. So, Chevy can’t design a car so that can replace burned out bulbs. And we want Chevy to design self driving so that it can be maintained? Gotta be kidding.It goes on this way: For what the US auto industry can do in design and maintenance is quite limited, really, about as low as it can go and still work at all. And the main way it works is the basic components and systems go for years until the bugs are worked out and then go for decades with little or no change.I’m also reminded of the time I got the NYS building code, drew some house plans, and took them to a general contractor. His response was informative: The people he hired couldn’t read blueprints. All they could do was build again one of a few models they had built dozens of times before.Look, guys, much of the US labor force can’t aim headlights or read blueprints. Work on self driving? Gotta be kidding!Self-driving is a fad. Longer term, that dog won’t hunt. Just not worth the cost or botheration.

    1. Tyler

      Self-driving is a fad?? With so much inefficiency on the road (from long haul big rigs to traffic to the fatalities), I can’t think of many reasons why those won’t or can’t be solved with semi- and/or fully-autonomous cars. They surely aren’t going to be solved by human drivers!Eventually these problems will be solved. And I’d be willing to bet self driving cars will play a large role. And who says the GM’s of the world are going to be the leaders? I’d imagine the breakthrough innovation will come from a smaller, more nimble player – doesn’t it usually?

      1. sigmaalgebra

        Fred was talking about just some optional equipment on a car. You are talking a major redesign and rebuild of most of our transportation system. Biggie difference.For a major rebuild, in the US, mostly people don’t like public transportation.

        1. Tyler

          Changing the way freight is moved by long haul trucks and decreasing fatalities on the road shouldn’t require an overhaul of the transportation system.My point is, change the way people drive (by replacing them with software), and you change the entire system.

          1. sigmaalgebra

            What Fred was talking about is current state of the art — he and GG have one in a garage. Replacing people with software at all generally is much more difficult. Except in some niche cases, I wouldn’t trust that software could replace people within some decades farther than I could throw a car.To have software replace people, really need some major, serious re-engineering of much of our transportation system. That would be a huge effort and highly conjectural.100+ years ago lots of people had lots of ideas for how to spend public money for public/private good. Some of the big cases were water resource projects, that is, dams, canals, etc. to make the desert bloom.So, too soon it looked like the country was about to go broke on such projects. Then a law was passed establishing the process of “cost/benefit analysis”. So, add up all the costs. For the benefits, try to estimate how valuable they would be. E.g., for a new park, what would people be willing to pay if charged them? Then before funding, the benefits had to exceed the costs. That simple process cut out a lot of projects.If we are going to do serious, major re-engineering of our roads, etc. to permit software to replace people in cars and trucks, then we should do some cost/benefit analysis. My view is, it would be super tough to argue so far in advance for such a big project that it was other than a big money sink.Here is an example: At one time Baltimore spent a lot of money and built a big subway system. The system was completed and about to accept riders. A guy the applied cost/benefit analysis, assuming that the construction was for free. So, the only costs considered were operations and maintenance. With this approach, the conclusion was that the optimal approach was to brick up the entrances and f’get about it. That is, the subway couldn’t pay for itself counting even just operations and maintenance.In the US, big public transportation projects tend to operate deeply in the red.So, if we go for some serious, major re-engineering of our roads to enable software to replace people, I sense a big hole in the ground that we pour money into.Gee, how ’bout: It should be much, much easier to have software replace people on the railroads. So, when the major US railroads do that, then we can consider — right cargo ships. When we get that successful, okay cargo airplanes — call up founder, COB, CEO at FedEx Fred Smith and ask him the status of his airplanes flying without pilots; I know Smith and would bet that he’s up to date on this idea. Then ask him about replacing people with software for his larger point to point trucks. Let’s see where DARPA and the US Army stand on replacing people with software for battlefield vehicles. Also, same song, the USAF. And, farmers, check in with some guy with 2000 acres of soy beans, corn, or wheat and also talk to John Deere about replacing people with software for the equipment that plows, seeds, cultivates, and harvests 2000 acres. Ah, don’t forget barges on the Mississippi. And ships on the Great Lakes. See what those cases are doing. Ah, yes, what about replacing people with software on the subway systems. Then we will consider cars and trucks on I-95, in the suburbs, and into Manhattan, at night, with fog, rain, snow, and ice with software replacing people.

          2. Tyler

            I don’t really understand what you’re saying. You think the govt is going to have to dig up every road, replace every traffic light, etc. to accommodate self driving cars?? And that’s what will keep them off the road?? Clearly, as seen from all the self-driving test cars on the road, the software will navigate the existing infrastructure. Companies aren’t going sit around and wait for the govt to update the infrastructure to accomodate them – they’re forging ahead regardless.I’m also confused by the industries you site – all have clearly been radically changed by autonomous features. John Deere tractors? You literally don’t touch anything in the cockpit while in the field, it’s all automated using GPS data to most efficiently use the land. Planes essentially fly themselves now, large ships are largely controlled by navigation software, etc. All those industries are going to way of automation.If you can’t see how the human functions required in each of those industries is being displaced with software, I don’t know what to tell you. Companies and industries are moving quickly; they’re not waiting on anyone, much less the govt.

          3. sigmaalgebra

            > I don’t really understand what you’re saying.I’m just saying that it will be a long time until your goal “software replaces humans” will be able to hold for cars, trucks, and more in general driving. Some special cases? Yes, at least until the insurance companies and/or lawyers get involved. In general driving? No.The CEO of GM claims that infrastructure changes will be needed, and I agree with that.And my view is that the auto industry is not up to designing, manufacturing, or maintaining sufficiently good self driving equipment. And I doubt that many people would be willing to pay for sufficiently good equipment.My view is that software can’t replace humans for cars and trucks in general without massive infrastructure changes because we can’t expect the software to be anywhere near that good for a very long time.The core reason is, in ordinary driving on ordinary roads, occasionally need actual human intelligence, and so far there is no software with anything like actual human intelligence. Some people can talk about artificial intelligence all they want, but the emphasis has to be on the first word and f’get about the second.Yes, my understanding is that essentially all the commercial airliners can land themselves at essentially all the standard airports. I would still be concerned about cases of high cross winds and ice on the runway. But, still the planes have human pilots.Sure, maybe some 18 wheel trucks can be sent without a driver across, say, an Interstate in Arizona in daylight, light traffic, perfect weather, but that’s nothing like going across the Rockies on a two lane road, at night, with fog, rain, or snow and with ice and snow on the road and with some accidents, rock slides, and detours ahead.A lot of people like to talk about AI, *neural*, *deep learning*, etc., but, for software having anything like human intelligence, that is at best 99 44/100% hype and the rest water and often worse.

          4. Tyler

            Ok, I understand a little better now…I simply disagree.I think you underestimate the ability of software to navigate complex situations. Software can already make more computations per second than humans, so it should be, given enough data, better at driving than a human. And that’s what is missing so far, data. However, that’s being solved one mile at a time by Tesla, Uber, Google, etc. The more data accumulated, the better software the software.The amount of waste and inefficiency in the automobile/transportation industry is so large, that the reward for solving it will be massive. And bc of that reason, there are countless people trying to solve it. Someone (probably more than one actually) will solve it.And in terms if GM, Ford, Chrysler, etc. – you’re right. They probably aren’t up to designing and manufacturing self-driving cars. The revolutionary product or technology will most likely be created by someone outside the industry, putting the legacy manufacturers in jeopardy.

          5. sigmaalgebra

            > I think you underestimate the ability of software to navigate complex situations.Depends a lot on the “complex” situation. Since there is no very meaningful or useful definition, characterization, or measure of complexity, tough to say what “complex situations” we can handle with software and what ones we can’t.For a nutshell answer, I urge you to take seriously my claim that at times driving takes human intelligence and that we have no idea how to write software that has human intelligence — we don’t know if that is 10 years, 50 years, 200 years, or 1000 years or longer.Chip away at parts of the whole problem? Sure.For complex situations, there are lotsof examples where just have no hope of handling the complexity directly but can do surprisingly well looking only for some statistical results.A big example is to predict the weather for two weeks. In some cases, we sort of can. But, we know that we have what applied math would call, say, an initial value problem for a system of non-linear partial differential equations. Well, we know some things about stability of the solutions of even simpler systems of differential equations — the stability can be poor. This means that trivial changes in the initial conditions can grow to huge changes later on. IIRC R. Bellman’s Ph.D. dissertation at Princeton was on that subject. So, for weather prediction, after about two weeks, trivial changes might change the prediction from rain to sunshine. This doesn’t have much to do with software for replacing human drivers but is an example of some of the challenges of complexity.So, due to the instability, a prediction about has to be only statistical. That is, do the calculation, say, 500 times with small changes in the initial conditions and collect empirical descriptive statistics on the results. Maybe the results are, for one week, can predict quite well; for two weeks, not so good; for three weeks, can do just as well ignoring current data and just use history given only the day of the year.So, if all we are going to be is statistical, then one approach is cute: To heck with the differential equations. Instead, just take a lot of weather data, what the weather was like one day and then two weeks later, take what the weather is like today, and then just do some empirical statistical estimation from the old data. That is, if have in the old data several cases of what the weather looks like today and for those cases the weather two weeks later looked much the same, then that is the prediction. So, just f’get about the complexity and just be simplistic and empirical — statistical.In advanced probability and stochastic processes, there are some astounding limit results. In some cases, with some mild, realistic assumptions, can give a surprisingly accurate probabilistic description of long term results. Classics include (A) power spectral estimation of second order stationary stochastic processes and the resulting filtering opportunities, (B) Markov processes, especially Brownian motion and the Poisson process, and (C) martingales.An example of Markov and Poissonis a problem I attacked working to support my wife and myself through our Ph.D. degrees: The US Navy wanted a prediction of the survivability of the US SSBN fleet under a special scenario of global nuclear war limited to sea. “Complexity”? Plenty there! They wanted the results in two weeks. Gee, glad they weren’t in a big hurry! There is a piece of old WWII applied math about searching at sea due to B. Koopman. It’s some clever but crude applied math, but seemed roughly okay in practice. What happens is a Markov process where the times of jumps are from a Poisson arrival process. And I had some ways to estimate the rates. Then, just do a Monte Carlo (draw lotsof random numbers, say, like simulating a game of Blackjack). So, each time I ran the simulation was one trial with one sample path. My software ran off 500 sample paths and took some averages and found some confidence intervals. The results looked good and were encouraging. The Navy got their results on time. How ’bout that! My work was later sold to a leading US intelligence agency; I could tell you which one, but then I’d have to …!But, from all the above, I see little hope for software to replace peoplein ordinary driving.> better at driving than a human.In some circumstances, sure: Totally knock the socks off humans. One example is how should a missile steer to run down an airplane when the airplane is trying to evade the missile? So, this is a generalization of deterministic optimal control of some ordinary differential equations. The generalization is called differential game theory. Likely US missiles have been programmed with the results. If so, and if the best solution doesn’t just boil down to something simple, then humans would have a tough time competing. And I’d hope that the same work would be applied to fighter plane dog fighting — I don’t know the situation.There was an example for the F-4 Phantom, how to reach 100,000 feet in minimal time: Part of the solution was curious: Of course, always at full power, climb up to, maybe, 10,000 feet, go into a full dive, get supersonic where, actually the drag coefficient is significantly less, and then pull out and climb as steep as can and stay supersonic all the way to 100,000 feet. I got that from an MIT prof big on deterministic optimal control, M. Athans.So, sure, sometimes software can do much better than humans and in some situations of operating an airplane.It may be that the least time way around the Nürburgring would be from deterministic optimal control software.> what is missing so far, data.In a way, yes: If equip all the roads with enough wires in the pavement, transmitters at the edge of the roads, transmitters for all the traffic lights, etc., then, sure, the self-driving car is basically just following electronic tracks. My guess is that that is some of what the GM CEO has in mind.Then all the car has to do is dodge small animals, larger animals, children, other people, bicycles, motorcycles, cars with drunks, cars that enter the lane from behind an obstacle, cars that change lanes in dangerous ways, handle fog, rain, snow, ice, equipment failures, junk on the road surface, holes in the road, downed trees, accidents, flagmen, situations where need to pass, say, a farmer with a tractor going 5 MPH, need to pull to a stop and talk to police, a flagman, etc.> The revolutionary product or technology will most likely be created by someone outside the industry, putting the legacy manufacturers in jeopardy.Well, sort of. So, maybe DARPA funds some more stuff, and some people get something terrific. Then maybe Samsung puts together a solid product that Detroit can just install. Okay, to me, total long shots, but suppose: Then, who’s going to maintain that? And pay for it — if only due to the servos and the need, from the insurance companies and lawyers, for really high quality and high reliability, I’m guessing that for personal cars the original purchase price will be high.> The amount of waste and inefficiency in the automobile/transportation industry is so large, that the reward for solving it will be massive.I don’t know just where this “massive” expense is, but with some sufficiently big bucks for some niche solutions, okay, might happen. Or, for commercial vehicles — land, sea, air — were we have human operators now, I suspect we will still want human operators. So, maybe the driver of an 18 wheel rig hauling 40,000 pounds of fresh pork from Iowa to Boston gets to sleep until an alarm goes off. The software sounds the alarm at anything that looks even a little risky. So, on the Ohio Turnpike, in daylight, perfect weather, light traffic, etc., the driver gets to sleep. Anything else, he has to take the controls. He is still on the job. Maybe the ICC will cut him some slack on how long he can be in the truck. Otherwise he is still there.The system that lets the driver sleep better be just darned good, basically better than a driver (who might fall asleep at the wheel), or the lawyers and insurance companies will scream. A few little mistake-ys on the Ohio Turnpike, and Ohio may pass some laws saying “No way”.I have a friend who had a company that drove 18 wheel trucks from Ohio to NYC with 40,000 pounds of fresh pork. One day on the Ohio Turnpike, there was a little problem, with the truck and/or ice on the road. The truck jackknifed and slid sideways into a toll booth. My friend was up to his chin in the problem for weeks. Lots of damage. Big mess. The insurance company coughed up big bucks.If there are accidents where the self-driving system can be said to have been a cause, then there is going to be big push back from insurance companies and laws.There is software. And then, very different, there is software where risk big bucks, the whole company, or human life.I don’t see any very significant solution for self-driving that actually replaces a human in any very general case short of actual AI, and I’m not sure that will happen even this century or next. Even then, for personal cars, the system might not be worth even the cost of the servos.

    2. ShanaC

      Tenagers and young adults don’t want to buy cars or drive that much. Most millennials pay more for cell data than cars now. That appears to be a permanent cultural change.Anyone in that category (I’m in it) will go for it.

  30. Andu @

    I’m more interested in the penetration rate of electric cars. Any info on that?

  31. Bret Tobey

    “Self-Driving” makes a great headline but if you called it “self-braking” (accurate in the sense BI used) then 1% might be low. Tesla’s grabbing the headlines but every major automaker is folding some “self-driving” technologies into their general product lines. The total penetration gets skewed because the average vehicle age is so high, 17 yrs in the US, lower in the EU, higher elsewhere.

  32. Morgan Johnson

    My Executive MBA class had this problem posed to us in our recent marketing module. Essentially the question posed was “Should Apple enter the car market? If so, how?”. It involved an analysis of the current car market and a survey of opinions of what the penetration of electric cars would be 5-10 years into the future. The viewpoints across the class were incredibly disparate, but the overwhelming majority thought that Apple should make a car. From a penetration perspective 1% seems low – though a lot will likely depend on execution from Tesla and other low cost producers. I think its more than possible for penetration to be a lot higher in the luxury segment.

  33. Varun

    will believe when applied to this until then we can keep living in a jetsonian fantasy.

  34. Veronica Schmitt

    “my room mate Mary Is getting paid on the internet 98$/hr”…..!jk291etwo days ago grey MacLaren P1 I bought after earning 18,512 was my previous month’s payout..just a little over.17k Dollars Last month..3-5 hours job a day…with weekly’s realy the simplest. job I have ever Do.. I Joined This 7 months. ago. and now making over. hourly 87 Dollars…Learn. More right Here !jk291e:➽:➽:➽➽➽➽ http://GlobalSuperJobsReportsEmploymentsSilvaGetPay-Hour$98…. .✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸✸::::::!jk299e…..

  35. MickSavant

    I think the estimate is probably correct. On top of that, regardless as to how they define self driving, I do not think we will have truly autonomous cars for decades. To be specific, I do not think it will be possible to get in a car on the East coast and have it drive you all the way to a residency on the West coast for at least 15 years. Even if this is technically feasible, it will be a long time before it is implemented.

  36. Roelof Reineman

    From a technical perspective it is low, from a commercial perspective and looking at the rate of replacement, it is high. When taking the legal aspects into account, I think that will be the biggest slowing-factor in this development.