What’s To Become Of The Self-Driving Car Market?

Tesla Model 3

SEP 17 2017 BY EVANNEX 45

Self-Driving Car

Tesla Model 3

IN THE SELF-DRIVING CAR MARKET, WILL THE WINNER TAKE ALL? [INFOGRAPHIC]

We all know that, when a new technology becomes established, there are winners and losers. However, the picture is more complex: some companies will be small winners and others will be big winners. Where a particular company ends up depends not only on the quality and popularity of its products or services, but on where it fits into the overall market – its position on the “value chain.”

Self-Driving Car

Tesla Model S (Image: Tesla Shuttle)

Many of the products and services that we use every day would be useless on their own – they depend on an “ecosystem” of related offerings that all work together to solve a particular problem or deliver a particular benefit. One way to describe this is to refer to each individual product or service as a “layer” in a “stack.” The classic example is the smart phone stack, which has at least three layers: the phone itself, made by a hardware company (Samsung, Apple); the OS running on the phone (iOS, Android, Windows); and the network that provides connectivity (Verizon, T-Mobile, et al). There are other layers as well – third-party apps, accessories, network hardware, etc – but these three are required for the phone to work.

This hierarchy of products and services is not unique to the high-tech world – a grocery store, for example, has its own stack of individual layers: the physical store; the chain that runs the business; the food wholesalers that provide the products, etc.

*Editor’s Note: This article comes to us courtesy of EVANNEX (which also makes aftermarket Tesla accessories). Authored by Charles Morris.

Not all layers in a particular stack are equally lucrative. Companies that sell hardware or network services tend to become commodity providers. They can be described as small winners – they enjoy a large market for their offerings, but they have no way to differentiate their products from those of their competitors, and little control over the overall market. The big winners are the companies that provide the system that ties the whole ecosystem or stack together, because they have much more power to control the user experience, and thus more opportunities to extract fees. The biggest winners are those that achieve a monopoly or near-monopoly, because of “network effects,” which make a product more valuable as more people use it. In some markets, the network effects are so strong that a “winner-takes-all” situation develops.

Self-driving technology is just beginning to emerge, and it’s still anybody’s guess who the big winners and small winners are going to be (although the holders of some of the losing hands are already pretty clear – taxi drivers are not in a good position). As we have already discussed several times in this space, Tesla is in a position to be one of the big winners – but how big? Is a winner-takes-all scenario likely? Will network effects allow one or two companies to squeeze the rest out, or will there be room at the top of the value chain for several competing companies?

Tesla fleet lined up at Supercharger station

In a recent blog post, Benedict Evans considers these questions in great detail. He notes that there are several dozen companies vying for a slice of the autonomous market, across several different industries: vehicle OEMs, automotive suppliers, major tech companies and startups. What will the future balance of power be? None of these companies wants to end up as a commodity provider – they all want to be the big winner that takes all. As Mr. Evans sees it, the companies that can define a lucrative business model will be those that can find “choke points,” as Microsoft and Intel did in the PC world, and Google did in smart phones. What are the choke points in the vehicle autonomy market?

The hardware and sensors that enable autonomy are likely to be commodities. There are strong manufacturing scale effects, but no network effect, because there’s no real reason to use one company’s LIDAR (for example) rather than another’s. Likewise, a vehicle’s operating system will probably be pretty generic. Windows beat the Mac in the PC era, and iOS and Android trounced Windows Phone, because of the virtuous circle of third-party software developer adoption. However, cars won’t need to run a wide variety of apps – just the apps that enable transportation as a service (Uber, Lyft, etc).

The place to look for network effects, Evans writes, is not within the cars but further up the stack – in the autonomous software that enables a car to safely navigate the roads, in the optimization and routing software that enables all cars to be automated as a city-wide system, and in the on-demand fleets of robo-taxis that will ride on top of these software layers.

According to Evans, these three layers (driving, routing & optimization, and on-demand service) are technically independent – for example, a GM autonomous car could use a Waymo autonomy module and have its movements coordinated by Lyft. However, Tesla doesn’t appear interested in playing this game – it has a bundled service model (aka a walled garden) in mind. It builds its cars to use its proprietary autonomy hardware and software, and has said that it plans to forbid people from using its vehicles with any on-demand service other than its own Tesla Network.

Self-Driving Car

Unlike other automakers who may outsource their self-driving vehicles, Tesla maintains a tightly controlled ecosystem (Instagram: marysia_mendakiewicz)

Evans foresees strong network effects in the data that enables autonomy – maps and driving data. In this area, as others have noted, Tesla has a huge head start over other industry players. When an autonomous vehicle (AV) drives down a road, it simultaneously compares the road to an existing map and updates the map. The more AVs you have on the road, the better your maps will be – the definition of a network effect. Driving data brings a second network effect into play. An AV must not only navigate the road itself, it must also figure out what the other vehicles on the road are going to do, and react appropriately. Machine learning is the key to this, and the more data you collect on how drivers (both human and robotic) behave, the better your software will be at reacting to real-world driving situations.

Tesla is amassing a huge database of both maps and driving data, thanks to its Autopilot hardware, which is constantly beaming information back to the mother ship. Will this advantage in data allow the Tesla Network to dominate the transport-on-demand market the way Microsoft dominates PC operating systems?

It’s far too early to say. There are several very important questions yet to be answered. Who will end up owning the all-important data, and how much of it is really needed? Tesla has its trove of data, but as other OEMs develop their own AVs, they will amass their own data empires. Is there a point at which adding more data has diminishing returns? Once competing players have generated enough data to get the job done, will Tesla’s advantage disappear? In other words, how strong is the network effect? Will autonomy data be bought and sold? Shared among OEMs for the good of all?

Another critically important question: how soon will true Level 5 autonomy – full self-driving capability – be available? Futurism has created an infographic that illustrates the wide range of predictions that various industry players have made (some of these predictions are already outdated). In July, Audi unveiled the 2018 A8, which includes a feature called AI Traffic Jam Pilot that the company says is capable of full self-driving in certain situations (it has not said when the feature will actually be enabled). Elon Musk has promised a demonstration of Level 5 autonomy with a coast-to-coast drive by the end of this year (although he expects regulatory approval to take another year or two). Google has predicted that its autonomous technology will be available in 2018, and Toyota, Nissan and Honda have all promised self-driving cars by 2020. The Institute of Electrical and Electronics Engineers has predicted that up to 75% of vehicle will be autonomous by 2040. Exciting times are ahead.

Infographic

Self-Driving Car

===

Source: Benedict Evans / Infographic: Futurism

*Editor’s Note: EVANNEX, which also sells aftermarket gear for Teslas, has kindly allowed us to share some of its content with our readers. Our thanks go out to EVANNEX, Check out the site here.

Categories: General

Tags: , , , , , , , , , , , ,

Leave a Reply

45 Comments on "What’s To Become Of The Self-Driving Car Market?"

newest oldest most voted
Lou Grinzo
In many ways, the transition to autonomous vehicles is much more interesting than the move from liquid fueled vehicles to PEVs. In my time working in computers (yes, I’m an old geek), I’ve seen many examples of promising new technologies that [1] took longer to hit the mass market than expected, [2] took over the market quicker once they arrived, [3] had knock-on effects virtually no one predicted, and [4] did not, in themselves, look and work the way “everyone” expected. I think autonomous cars will be an extreme example of this unpredictability. In general, I expect we’ll see a rush through the early levels of autonomy, but going from 3 to 4, and then 4 to 5 will be much more difficult and take much longer than many are expecting today. The technical and legal issues are staggering. As I so often point out, consider how often in your daily driving you encounter situations that would require science fiction-y levels of AI, sensors, etc. for a car to navigate safely and acceptably efficiently without human intervention. Just the interactions between autonomous and human-guided vehicles presents immense challenges, let alone unexpected police checkpoints, detours, etc. I desperately want to see… Read more »
Pushmi-Pullyu
“As I so often point out, consider how often in your daily driving you encounter situations that would require science fiction-y levels of AI, sensors, etc. for a car to navigate safely and acceptably efficiently without human intervention.” What everybody seems to ignore is the possibility — I’d call it a near-certainty — that the environment will almost certainly be altered to make it easy for autonomous cars to navigate. Before the motorcar, we didn’t have highways with clear lane markings, and traffic signals (which developed into the modern stop light) were exceedingly rare. Society adapted to the motorcar, and it will adapt to autonomous cars. Where it’s necessary to put in certain signs or markers which the autonomous car can recognize, to help in places where navigation proves difficult, then that will be done. “Just the interactions between autonomous and human-guided vehicles presents immense challenges… Waymo clearly has mastered those challenges, since according to reports, Waymo self-driving cars can operate on most roads at 35 MPH or less, and there are only a few special cases, such as bridges, where they can’t go. Interacting with human drivers is clearly a problem which has been solved. “…let alone unexpected police… Read more »
Will

Too invasive

Nick

Why? Cars are already unbelievably invasive.

Alonso Perez

Really? I’ve never had a car that reports its every move to the “authorities”, and can be controlled by them remotely.

James P Heartney

Welcome to the brave new world.

Pushmi-Pullyu
If you are driving, and you hear a siren coming up behind you, and you see flashing lights in your rear-view mirror, you are supposed to pull over and stop at the side of the road. If you don’t, you are subject to the penalty of a traffic ticket. Autonomous cars won’t be allowed on the road unless emergency vehicles have some sort of ability to cause (or “force”, if you prefer) such autonomous vehicles to mimic this behavior. Will we rely on problematic cameras and possibly microphones to “see” the flashing lights and “hear” the sirens of emergency vehicles, to ensure autonomous cars get out of the way? No, I don’t think we’ll rely on something as unreliable as that. Autonomous cars will be built to communicate wirelessly with each other. I don’t think there is any real question this will happen; planning and implementation of that is already in development. They’ll be communicating with each other to facilitate traffic flow, to assign priority at intersections (eliminating the need for stop signs and stop lights), and to help each other “paint” a virtual picture of their surroundings. Since enabling autonomous cars to do all that is inevitable, doesn’t make… Read more »
Jason

I can just imagine the confusion over this wireless system. Let’s face it, they can’t even decide on CHAdeMO vs CCS vs Super Charger vs whatever else is out there.
In Australia I thought we had a good design guides for vehicles, but you can get into a car that had the blinker and wiper stalks on the opposite sides, how confusing is that?
So I think it will be a far flung future before such common standard is reached, I hope I’m wrong.

James P Heartney

Pushmi, I often disagree with you, but I think you’re right on the money with this.

Beyond the likelihood that the driving environment will be adapted to enable vehicle autonomy, the next question will be, once true autonomy is achieved, how long until they start restricting human drivers? Obviously some areas (rural and newly-developed areas, for example) will require/allow human drivers for the long term. But in cities, it’ll quickly become apparent that the human drivers are the weak/dangerous link, and there’ll be benefits to eliminating them. This could happen more quickly than people expect.

Priusmaniac

Skynet woudn’t say better.

Beware of what you wish for.

Pushmi-Pullyu
James P Heartney said: “…the next question will be, once true autonomy is achieved, how long until they start restricting human drivers? …in cities, it’ll quickly become apparent that the human drivers are the weak/dangerous link, and there’ll be benefits to eliminating them. This could happen more quickly than people expect.” An excellent question, thanks. There will be two competing social forces at work: 1. The desire to reduce or end the danger of human drivers on the road as quickly as possible, once autonomous cars are proven to be much safer. 2. Protection of the economically disadvantaged, who need a car to get to work, and can’t afford to buy a new autonomous car. It will be interesting to see how that plays out, if I live to see it. Certainly it could start happening sooner than we expect. I have been startled to see talk of banning gasmobiles from metropolitan areas such as central London. I never expected to see that seriously proposed this soon! Perhaps we’ll see similar things with autonomous vehicles. Perhaps we’ll see them banned in certain areas of very high traffic density, with the ban gradually extended outward as the years pass. Rather than… Read more »
orinoco

Applying (traffic) rules is simple. Detecting the exemptions from the rules is the tricky thing.

In traffic there are often situations where the rule does not make sense e.g. in a narrow street, you have priority but it does not make sense to insist on it and therefore give way to the other driver. Then you need to realize the situation correctly, predict the different outcomes correctly and start communicating with the other driver. And this happens within fractions of a second.

As a result you always drive anticipatory, predicting different outcomes of a situation.

I think this is the real challange for autonomous driving. And how would an autonomous car sign to a human driver like humans do? ok, headlight flasher or horn, but human drivers rarely use them to solve situations like this as they are not very subtle and may confuse more (and other drivers) than make something clear.

Pushmi-Pullyu
orinoco said: “In traffic there are often situations where the rule does not make sense e.g. in a narrow street, you have priority but it does not make sense to insist on it and therefore give way to the other driver.” I don’t know how Waymo or other companies’ autonomous or semi-autonomous cars handle it, but from watching videos of Tesla’s semi-autonomous cars driving, it seems the self-driving car always gives right-of-way to other vehicles. I think there is a basic misunderstanding, among those who are not computer programmers, of how autonomous cars are intended to work. Programmers are not trying to fully mimic human behavior. In particular, programmers are absolutely not going to program cars to compete with each other for priority. Self-driving cars will be programmed to cooperate, to facilitate smooth traffic flow; they won’t compete in the self-defeating manner that human drivers use; behavior which causes traffic snarls and jams. The short answer to your question, orinoco, is this: The autonomous car will always give priority to the human-driven car, and to drive defensively in all cases. In fact, speaking as a computer programmer, it makes sense for self-driving cars to be programmed to treat human-driven cars… Read more »
Mike

To have truly self driving cars we need a level of artificial intelligence comparable to ours. This has far bigger consequences on our lives than whether we are or are not driving cars. I am not looking forward to such future but even at current pace it will probably not happen in my lifetime.

James P Heartney

Probably not true that vehicle autonomy will require sentience (i.e. “humanlike” intelligence). All current self-driving strategies employ a variant on “expert systems” AI, which is not self-aware. It’s not completely clear that an expert systems approach will be enough, though there are billions of dollars riding on the bet that it will.

The larger question of whether expert systems, applied to big data, will be powerful enough to overwhelm human intelligence and enslave us as individuals, is also unanswered. But I suspect it won’t.

And then there’s the blue-sky question of whether we’ll ever actually create artificial sentience, and if we do, whether it’ll be scalable enough to become artificial omniscience. Interesting times ahead.

Mike

I didn’t even think about self-awareness. If AI can replace us at driving it can replace us at everything. At best, we’d turn into a society driven purely by consumption. At worst, we’d have to compete with AI on price.

We’re still decades away from it, though. People see a self-driving car following the lane precisely or reacting to simple inputs in zero time and assume we’re almost there. But that’s all mechanics, automation. It’s like saying automatic translation is good because the result is grammatically correct and has no spelling mistakes. The problem is that to push it to the next level AI has similar situation awareness to ours.

Ultimately, self-driving cars should be able to pass during test unattended. A rigorous one at that, as they are going to be replicated in millions of units. Before we reach this level AI should not be allowed (both ethically and legally) to drive a car without the human driver remaining 100% in control at all times.

Pushmi-Pullyu

Mike said:

“To have truly self driving cars we need a level of artificial intelligence comparable to ours.”

This absolutely is not true. All we need is fairly sophisticated “expert systems” software to have fully autonomous cars. Self-driving cars don’t need to understand the world on the human level any more than a tax return program needs to understand why you need a job to earn money.

Waymo already has some prototype self-driving cars which don’t have steering wheels. Do you think such cars are self-aware, or “intelligent” in the manner that humans are? Heck, they’re not even as smart as your pet goldfish, let alone a dog or cat.

Let’s be clear: Modern robots are only about as “smart” as a middling-smart insect. That’s not likely to advance much in the next few years, but we certainly will get fully autonomous cars within a few years.

Mike

That’s enough for driving assistance and definitely not enough for self driving cars. No human driver would pass a driving test being oblivious to the surrounding environment and standards for the AI should be even higher, given that a single “driving test” would put millions of identical drivers on the road.

A competent driver should be able to effectively predict actions of other road users based on their behaviour and appearance. Elderly people, cyclists wearing headphones, drivers using cellphones, any obstacles that may affect the traffic in either direction all present distinct hazards which should be accounted for.

For the foreseeable future we can use AI as an assisting technology, or build infrastructure where road rules don’t apply (eg. dedicated roads complete with departure/arrival areas).

Randy Bryan

The advent of fully autonomous cars-trucks will change much of the car market, more toward taxis/uber and platooning trucks. But the idea of personal ownership of a car, car type, and control over its interior will survive for most. Maybe we’ll go back toward 1-car families… Thx for the article.

Pushmi-Pullyu

EVs and autonomous cars will reduce the price of cars and car insurance, and lower the barrier (no driver’s license needed) to owning a vehicle. That will result in increased car ownership, not the fantasy of decreased car ownership which some industry watchers have convinced themselves will happen.

SJC

No, car insurers are out to maximize profit. Musk says all he needs is better software, he needs a LOT more than just new software.

Pushmi-Pullyu

I hope you’re not really trying to argue that insurance rates won’t drop due to lower accident rates for self-driving cars. Insurance rates are based on actuarial tables. You can certainly argue that insurance policies are not priced fairly, but no matter how much we may hate insurance companies, it’s not rational to argue that insurance prices won’t drop even if the accident rate does.

Nick

Why own a car, when you can have one at your door in minutes?

Car ownership is a huge pain.

BillT

I completely agree with Nick: I think fully autonomous cars could vastly reduce car sales and are a much bigger threat to automakers than the transition to BEVs. Some back of the envelope math tells me that Uber or Lyft could make a hefty profit selling rides in even $50K autonomous Teslas for far less than I pay per mile now when I factor in depreciation, loss of investment income due to money tied up in the car, insurance, fuel and maintenance even on a fairly frugal car kept over 10 years.
At the very least they could vastly reduce the number of multi-car households. Certainly if reliable, safe and cheap autonomous taxi service were available we would immediately get rid of 1 of our cars and just keep the other until we become completely convinced we didn’t need it or when we got the urge to drive ourselves.
I suspect many others would do the same and that would have a very large impact on car sales.

Priusmaniac

Because your car is the one that you like, the one in which you have your specific items, the one you know and trust, the one with memories of good and bad times. To me such a question is as strange as someone asking why getting a house when you can sleep at a hotel or even why getting married when you can have a women in minutes.
If a car is reduced to from A to B, than indeed you may as well take a bus, a metro or a taxi. Knowing it is more than that is the start of the answer.
If you think about it there are other similar cases, sometimes as simple as your razor or your handbag, it is just yours and not a stranger’s, it is a referee, a like to yourself, an identification, an adequation, a recomfort in recognition up to the finest details and imperfections. Globally it is part of what makes a you instead of a him. So going from A to B is just a hapen to be sideline of a car.

Pushmi-Pullyu

“Why own a car, when you can have one at your door in minutes?”

Well, why doesn’t everybody give up their car right now, and depend on taxi service?

I’m pretty sure you know the several answers to that question. What mystifies me is why anyone would think those reasons are going to disappear just because cars will become self-driving.

James P Heartney

The big reductions in cost for autonomous EVs in fleets will come from the fact that they can be in use 40-60% of the time rather than <5%, as will happen if they are privately owned. Those heavily-used EV robotaxis will be able to take advantage of the EV half-million or more mile lifespans, unlike the private car spending most of its time parked.

Pushmi, you just can't get your mind around the idea that lots of people would happily stop owning cars if they didn't need to. But it's true.

I'm sure I won't convince you otherwise, so we'll just have to see how it plays out in the next decade or so.

Pushmi-Pullyu
James P Heartney said: “Pushmi, you just can’t get your mind around the idea that lots of people would happily stop owning cars if they didn’t need to. But it’s true.” Oh no, I absolutely agree that many people would happily give up owning their own car if it was practical to do so. I just don’t agree that autonomous cars are going to make that practical for large numbers of people. In fact, as I’ve pointed out, the overall effect, between BEVs being cheaper to buy than gasmobiles, lower insurance rates for autonomous cars, and no need for a driver’s license, will almost certainly be to increase personal car ownership, not decrease it. That’s basic economics. Your arguments appear to me to be based on wishful thinking and ignoring reality. For example, your assertion that autonomous cars will increase the amount of time each car is being used from ~5% to ~40-60%… while ignoring the fact that using the car 8x-12x as much will cause the car to wear out 8x-12x as fast, and also will cause much higher insurance rates for that vehicle. You’re also ignoring the fact that the demand for cars peaks at certain hours of… Read more »
Pushmi-Pullyu
I quit reading maybe halfway down. Wow! I guess that Evannex writer was getting paid by the word. He had very little to say, and said it at very great length! The idea that cars will become just commodity devices, merely rolling appliances, is ridiculous, and repeating that doesn’t make it any less so. People choose the car they buy on the basis of style, utility, comfort, and/or the luxury features. Or they buy a certain car because they’re loyal to a certain brand and their old car is worn out. The idea that somehow EVs and autonomous cars is going to change all that, is pretty silly. If cars come to be seen as commodities, with little differentiation between them, then the market will support only a few models. The reason the car market supports hundreds of different models, from many different auto makers, is because people have a wide variety of what they need and what they want in a car. This isn’t going to change just because they’re EVs, nor because they’re self-driving. No one auto maker is going to capture a monopoly of the market. Ford managed to capture 90% of the market at one time,… Read more »
ffbj

It’s like looking into a misty valley where the road disappears and arguing which way it goes after that. Well it goes across the valley floor, that’s about all you can say.

Victor

For a car to be fully autonomous the computer in that car has to be able to reason. To have rational thought is a lot more than staying in ones lane or reading street signs. The level of artificial intelligence necessary to drive a car is to make that computer human, minus the emotions. There’s something about that that’s a little scary.
When computers become that intelligent, there will be other applications (unintentional consequences).

Doggydogworld

I don’t agree. You can drive with your subconscious. Ever found yourself driving to your old house a few weeks after you moved? How much reasoning is involved in that?

Pushmi-Pullyu

Victor said:

“For a car to be fully autonomous the computer in that car has to be able to reason. To have rational thought is a lot more than staying in ones lane or reading street signs. The level of artificial intelligence necessary to drive a car is to make that computer human, minus the emotions.”

You only need to look at what Waymo’s self-driving cars have achieved to realize that simply isn’t true. All self-driving cars need is a good suite of sensors plus an unusually sophisticated “expert systems” software program. The car doesn’t need to reason any more than your tax return software needs to understand why you need a job to earn money.

Ocean Railroader

There is a possibility that the self driving car market might burn out or become a fad or not really grow.

The reasoning is if this self driving feature costs $5000 dollars more over a regular car then I wouldn’t buy it in that I would rather have the car be $5000 cheaper.

Bacardi

I’m sure people said the same thing when the automobile started to replace the horse and buggy…”Autos aren’t as reliable, it’s a fad”…

There is incredible global demand with the new generation of tindering/netflix kids leading the charge with more and more “old people” being flipped to the technology every day…It’s superior technology which can mirrors a luxury experience, valet…Can drop you off at works or a restaurants front door and never have to worry where to park…

Priusmaniac

It is not especially a matter of cost but simply of people not being interested in a self driving car.
For a taxi it is obviously interesting although many customers will prefer the presence of a driver to be able to speak with him about different things and to get a help with the luggage, but for your personal car it is not evident at all and perhaps even a push back considering the implications in some kind of silicon valley dictature, privacy infringement, hacking risk and just plain malfunction. Being wounded because of the mistake of a fellow human being is one thing, being wounded because of a software or a hacker is yet much harder to accept.

Pushmi-Pullyu

I can’t imagine why anyone would want one of them newfangled motorcars. Why, your horse knows the way home, so if you fall asleep or you’re feeling no pain, then your horse will get you there. Let’s see one of them motorcars do that! Not to mention which, it’s liable to get stuck in the mud at any low spot in the road.

These motorcars are just a passing fad. GET A HORSE!
😉

Priusmaniac

Well, I am pushing for EV’s but not at all for removing people’s freedom to drive.

James P Heartney

Other than the sensors (which most cars are starting to have anyway), the most important piece for autonomy is software. And software has a near-zero marginal cost to duplicate once it’s written.

The first few versions of autonomous driving software may be pricy (as they recoup development costs). But within a few years, it’ll be another cheap commodity.

Bacardi

http://mashable.com/2017/08/09/tesla-self-driving-computer-update/#OT9qK0HQmsqU

From a Tesla spokesperson:

“Tesla still expects that HW 2.0-equipped cars will be able to reach Level 5 eventually, but admits that there is a chance that the system will need to be upgraded in the future. If that’s the case, the spokesperson told us that the automaker will provide upgrades to the required computer system “at no cost.”

Pushmi-Pullyu

So Tesla keeps claiming.

Maybe that will turn out to be so. But in my opinion, Tesla is in pretty serious denial about not using active long range sensors in a 360° circle. It’s not the car’s computer which needs upgraded, it’s the sensor suite. Cameras and optical object recognition software pretty clearly don’t cut it now, as shown in Tesla’s own demo video. I don’t think it’s possible to improve the software sufficiently to be truly reliable, and even if it were, why use cameras which have the same limitations as the human eye, regarding darkness and being blinded by fog? Lidar and radar don’t care if it’s light or dark, and at least in some configurations can see thru fog to a reasonable degree.

Doggydogworld

The blog this article links to is much better than the article itself. But even the blog post the biggest network effect.

The biggest network effect is local robotaxi fleet size. A bigger local fleet offers quicker pickup with fewer deadhead miles. An incumbent with faster service at lower cost is almost impossible to dislodge.

This makes for an interesting situation. Waymo is preparing to launch in Phoenix. They have ~600 cars there now. How long would it take to ramp to 6000 once they go live? A month? Then 6000/month after that until they’re too big to attack? What would stop them?

Beyond Phoenix it gets interesting. It’s one thing to add 6k cars/month in one city, quite another to add 600k/month over 100 cities. That’d be be buying half of all cars sold in the US! Now expand that to 1000 cities worldwide and it becomes obvious logistics and local politics play an even bigger role than technology.

Waymo can choose to ramp slowly in all cities or quickly in fewer cities. Either approach leaves them open to competition. It’s a fascinating game theory exercise.

Doggydogworld

That should say “But even the blog MISSES the biggest network effect”.

Pushmi-Pullyu

Thanks, Doggydogworld! Yes indeed, that blog post is far more interesting, and unlike this Evannex article, I actually learned some things by reading it! (The writer also makes a couple of the same points I’ve made in comments on the subject.) Not that I agree with everything it says, but whoever wrote it seems to be pretty well informed.

Here’s that link again:

http://ben-evans.com/benedictevans/2017/8/20/winner-takes-all

Priusmaniac

When the dust settle we will see robo taxis and that will be about all once the hype is over.
We have seen that with rocket cars, battery exchange, flying cars and now self driving cars.
Fortunately the transition to ev cars with at contrary stay and live on.