1) Google / Waymo is by far the furthest ahead in self-driving tech.
2) Self-driving is possible in at least certain circumstances in the next decade or two (already is to an extent today).
3) Transportation / automobiles is an absolutely massive industry. Even companies like Dana, Inc (maker of axles) have $4bil valuations. It's easy to see how a self-driving tech maker would be worth far more than an axle maker.
4) Software tends to be a winner-take-most industry.
Given 1-4, $1.1B doesn't seem like much. Even if by chance it doesn't turn out to be a positive ROI, it seems crazy to think the $1.1B was a bad bet.
And given current prices for self-driving startups (example, Gm's acquisition of Cruise for >$1B), the current market value for Waymo would likely be >$10B.
> 4) Software tends to be a winner-take-most industry.
That's true of software generally, but not will not be true of automated driving software.
Windows, iOS, Facebook and YouTube have strong network effects. Google search and Microsoft Office offer a familiar user interface and superior features.
Self driving software will have none of these benefits - there is no network effect, no significant user interface, and no feature differentiation. As soon as there are 3 self driving software products on the market that meet minimum safety standards, competition will force the price down to little more than the cost of development. What's more, that cost of development will diminish over time: If in 2027 a company decides to develop self driving software, the development cost will be much less than the $1.1 billion Google has paid. So over the long term the economics will be similar to airlines - suppliers offering a commodity product and generating thin margins.
Although the long-term trend is bleak, Waymo does have two short-term advantages: 1) possibly being first to market; 2) its suite of patents on self-driving tech. Personally I don't place much weight on these advantages so I would value Waymo much less than the stock market. As an Alphabet shareholder I very much hope that they spin off Waymo so I can sell my interest in self-driving tech.
Agree for the most part, but question the lack of network effects here:
Is there not a potential interoperability / safety gain if all the cars on the road know how all the others will react? In a game of chicken they know they should each veer right to avoid an accident, so to speak.
I guess the answer depends to some degree on just how much inter-car communication is part of the future and whether those will be fully standardized. And even if standardized whether additional proprietary communications will be permitted between the same species.
Finally, isn't there a race on to collect as much detailed driving data as possible to train these systems? So the more a system is deployed the faster it will improve and therefore be more attractive?
I guess my question is really whether meeting minimum safety standards is going to be the sole question that purchasers of self-driving tech ask before buying. It's not too hard to fathom different techs being a "smoother" drive or maybe even able to handle more side roads (and thus more shortcut options) that other techs aren't ready for yet.
That's a good point - I didn't think about network effects of having autonomous cars communicate with each other. No doubt it will boost safety, but I strongly suspect that such technology will be heavily regulated and standardised. If Boeing and Airbus each developed a mid-air collision avoidance system, wouldn't regulators force them to interoperate?
Of course there will still be an autonomous car company with a superior safety record and perhaps better navigation routing, but these are not strong advantages. Consider that budget airlines can compete effectively with established airlines even when they have a worse safety record and fewer direct routes.
If you are first to market with selfdriving cars AND manage to transform transportation into a ride-hailing service then you might get a natural regional monopoly wherever you build out the fleet support infrastructure, similar to other utility services. Plus brand recognition for that service.
That is, if somebody figures out how to make good autocars without absolutely massive amounts of training data. For now it looks like whoever gets their software out in world first will have a huge advantage just because they get much more telemetry than the competition.
I agree you need massive amounts of training data. But the marginal value of that data diminishes rapidly after a certain point.
Let's say Waymo spends $2 billion to get the training data necessary to produce a fully automated car. A few years later Uber has also spent $2 billion and has a the necessary training data. By that time Waymo has 10 times as much telemetry data, so its cars might be somewhat safer. But as long as Uber cars are safe enough to get regulatory approval they can compete effectively with Waymo. Beyond a certain point people generally care more about cost than safety - look at the airline industry.
Also consider that Tesla seems to be getting telemetry data for much less than Waymo or Uber - they have thousands of drivers who are willing to pay for sensors and test automated driving software for free.
People are not very rational. Almost all trips in cars are very short, yet people are still very concerned about range in electric cars. Being able to cover more exotic driving conditions ("Our cars also work during a severe hurricane") could be a very effective marketing strategy.
> People are not very rational. Almost all trips in cars are very short, yet people are still very concerned about range in electric cars
As they should be... What good is a car if it can't be used for the occasional long trip? I expect my car to be up to the task to go visit friends or family 200km away on a whim, after coming home from work, instead of sitting idly while my battery is charging for a couple hours... I'm sure many users have the same expectations. Are we irrational?
Yes, the absolute worst case for most people is you get a rental car.
Having family 100 miles away that you visit regularly is reason to get a 100+ mile range electric car. Having family that lives 100 miles away that you visit every 10 years is not.
NYC is a great example where many people making 100+k per year don't own a car because transportation != car.
Sure, most trips are short. But short trips are best accomplished via walking, biking, public transit, or taxis. The only reason I want a car is for long trips. It can make short trips more convenient, but also if I were just buying it for short trips there's no point in spending the money, at least not living in a major city. Long trips are usually cost-prohibitive without a car that has decent range, at least relative to the cost of a gas-burning vehicle.
There are massive economies of scale. If Google is first to mass market, they will have a large information lead. From there, their cars could take better routes and handle more edge cases.
There's a similar network effect and economy of scale that applies to Waze + Google Maps' lead in mapping.
It seems difficult to tell who is the “furthest ahead” because it’s a pretty complex race.
For starters, we can’t even say with certainty who all of the competitors are. Some are very quiet or at early stages but could be working on a breakthrough problem that turns out to be a pivot point.
Some are specializing. Who is going to scale up the best and most cost efficient solid state lidar? There’s a whole list of key problems being worked on.
Do you mean hardware or software? Just the algorithms are a huge piece. Just hardware is a huge pieces.
Beyond price and performance, it’s common that unforeseen factors end up being important in determining who becomes the most successful in selling traditional manufacturers components or whole systems. Surely we don’t realize all of them yet.
Also there is so much still being held close to the vest. These guys are keeping lots of secrets about how far along they are truly, roadblocks, etc.
I don't really trust any of the marketing (including videos) or friendly articles about how far along companies are. The best way to judge I've seen is by the mandatory reports companies testing in California must supply. By those, Waymo is far, far ahead as of 2016. They've driven 635,868 miles (two orders of magnitude more miles than their next closest competitor) and their miles driven per disengagement was 5,128 (an order of magnitude higher than the next closest, with many still in single digit miles per disengagement). I'm eager to see 2017 numbers. Perhaps one of the other major companies has caught up but I'm doubtful. That's a wide gap to close in a year.
Sure, the numbers could be gamed a little (just drive on the same road you can do perfectly every time every day all year) but doing that will never let you improve your real world performance (important for the big companies heavily invested in this, not so much for the smaller companies looking to get acquired).
I think it's reasonable to just ignore all the small companies working on this in stealth. It's all about testing for self driving cars. They can't perfect it just sitting in a garage and thinking really hard about what problems they may encounter and maybe a few prototype vehicles. They need as many cars as they can on the road driving and collecting data on real world situations.
Isn't Uber testing in Pittsburgh and Phoenix? They wouldn't report those miles to California.
What about R&D outside of the US?
I don't think California reports are an adequate measure of progress. Additionally, I am not sure mileage matters as much -- you even suggest why in the next paragraph.
>They've driven 635,868 miles (two orders of magnitude more miles than their next closest competitor) and their miles driven per disengagement was 5,128 (an order of magnitude higher than the next closest, with many still in single digit miles per disengagement).
Making the assumption that "disengagements == bad" is a bit suspect though. We have no way of knowing if the reason for disengagement among competitors is purposeful, or due to more strenuous testing environments. It's not like this is a performance metric we know they are trying to optimize. Without knowing the testing methodology, it's meaningless to compare these numbers.
Who is going to scale up the best and most cost efficient solid state lidar?
Probably some big auto parts company. Continental [1] and Denso [2] are both building solid state LIDARs. Quanergy seems to be all hype.
Maybe automatic driving is just going to be a set of components Tier I auto parts makers sell to auto companies. That's how ABS braking and stability control works. Those are the companies that can make electronics work reliably in the automotive environment.
It may not be a race, either. Many major auto manufacturers expect to ship some form of automated driving in the 2020-2021 model year. There's no reason to expect one big winner here.
As for "transportation as a service", Avis, the car rental company, is getting ready to do that. They're servicing Google's self-driving test fleet in Austin, so they'll be ready to go into that business when the time comes.
They started early and seriously, unlike manufacturers who embraced it with marketing in mind and then put "beta" products on the road. Google were the first to gather massive amount of data on SDV.
Maybe they're not the first right now, but they're safely in the lead, waiting for the others to fumble.
I work for one such "early stage" startup - all we do is bullshit VCs by showing them research done by other people, and then passing them off as our own.
Indeed, the only comparison we have for Google's progress in most cases is what they claim. The few times other parties have stepped in and spoken up, there's been some less rosy truths in there.
It's in all of these companies' best interests to look like they're in the lead.
The tech we can judge the progress of is the tech we can actually test and use.
Consider that at the time, an engineer told him that the car was dependent on a level of mapping detail that was impractical at a nationwide level, that the car would run a red light if it wasn't on the car's map.
The article also cites that the cars would have issues in bright sunlight, particularly determining the color of stoplights, that they couldn't handle construction sites consistently, etc.
Also, consider the Slate article mentions Google Self-Driving Cars don't know how to park, two full years after the PR stunt which shows a blind man doing so in a Self-Driving Car. (Which was almost certainly a heavily staged operation.)
It's safe to assume Google has improved upon some or probably all of these issues in the last couple of years, but I think that article shows a stunning difference between where a press team says a technology is and where it actually is.
The point of not having a steering wheel was to experiment with a platform where it is assume disengagements where user intervention is needed in an emergency won't happen, because Google's own research indicates that relying on disengagements is dangerous.
Level 2-3 cars are frankly, too dangerous for the road IMHO, and the rush by automakers to ship these out to the public is, I predict, going to lead to some serious disasters, class action lawsuits, and regulation, that imperils the whole enterprise.
If there's anything that you don't ship "until it's ready", it's a device that's lethal to human life when something goes wrong. Tesla already killed someone, and while you could argue he violated the beta test agreement, shipping cars which requirement drivers to be attentive and keep their hands on the wheel while self driving -- ready to be aware and take control at a moment's notice -- effectively are engaging in bullshit marketing, because we all know that people who use self driving don't want to sit there "engaged" but want to be free to do somethings like check their phone.
Public data on disengagements shows Waymo is far far ahead. Are disengagements important? Yes. Every disengagement is a failure of the car's automation to handle road conditions that would require emergency intervention by the driver. But keep in mind, Waymo drives off the highway as well, whereas many of the competitors with worse disengagement figures are triggering them on mostly highway driving.
Taking time to do this right might be frustrating to those who want products shipped immediately, but the first time a level 2-3 car runs over a schoolyard, people's minds will change very fast.
Yes, but computer vision has moved forward by leaps and bounds since 2014, and so have GPUs.
AD pipelines uses object-detection for traffic-sign detection. RCNN was state of the art for this back then and ran at what 0.2 fps and ~40 mAP ? SSD came out in 2016 and ran at 60+ fps at ~80 mAP. Google's mobilenets implementation probably runs atleast twice as fast and apparently has no noticeable loss in accuracy.
There are more performant and more efficient network architectures than mobilenets. Don't get me wrong, Google's DL research is good, but it's not significantly ahead of field SOTA like they would have you believe.
I don't think mAP on VOC is indicative of real-world performance - especially given COCO mAP is just now cracking the 40s with similar architectures. But the gains in proposal based detection/instance segmentation from 2014 to now have been staggering for sure.
Seriously, they'd make trillions from self-driving tech. They'd end up spinning off an Uber-like self-driving ride service, and then they can track you and serve ads to you wherever you go. The data mining potential alone is huge.
Don't care to participate? Your steakhouse will show up on Google Maps if the user knows precisely where to zoom in or the exact name of the restaurant. Otherwise, "Hey Google, take me to a great steakhouse" and off we go..
> 4) Software tends to be a winner-take-most industry.
Yeah. I think this is it. I remember when google paid $1.6B for Youtube 10 years ago, people thought they were insane. Look at them now. Youtube dominates the web as video platform.
Pretty much the successful self-driving AI platform will probably dominate the market and will be installed in most cars. Like how Android dominates the smartphone market.
In tech/software, it seems to be winner-take-all market.
I agree that 1.1B isn't that much given the potential size of the market (and given the size of Google's profits, it's a surprisingly small bet they've made).
But how do you arrive at the conclusion that Google is way ahead? I cannot buy a Google car with self-driving capabilities. I can't even buy a car from any other manufacturer with Google's self-driving software in it. I can buy several other cars with some degree of self-driving tech in them, most notably Tesla, but also Volvo and Audi and others.
Already now, you have tens of thousands of Teslas driving around in either self-driving mode or in what is in essence a "recording" mode where they record and simulate. They drive in all kind of traffic, roads, rules, weather, light etc.
Meanwhile, you have a few hundred Waymobiles mostly driving around the perfect, dry and pristine Northern Californian roads with 25 mph. They have merely logged in 3M miles in almost 10 years. That's the equivalent of 10,000 Teslas driving 10 miles per day for a month. In a couple of years you will have a million Teslas driving on the roads. And you probably still just have a few hundred Google cars driving around engineers.
Google is losing the number game with its boutique approach. Waymo is WayBehind.
pfft...self driving cars aren't like smartphones. If you think of our road and traffic networks across the world the places this stuff is going to work is going to be very very low, i.e. there are upper limits on scalability. It's going to be costly and in the hands of a few for a long time to come.
Apparently Tesla is the furthest ahead in terms of having a sellable product. Google however, no matter how fancy their technology maybe, they don't have a window to demonstrate yet and need to rely on existing auto manufacturers, which is developing their own arms of self-driving rapidly and they have vast advantage as to collect data.
I wonder if this includes the 'Fuck you money' engineers on project Chauffeur received for passing certain milestones. Earlier in the depositions it was disclosed that Levandowski was paid a 120 million dollar bonus, though it wasn't revealed how much was given to others on the project.
The bonuses were paid out at the end of 2015. It was shortly after that that Levandowski revealed Otto to the world. Plus Chris Urmson started Aurora, and Brian Salesky went on to become CEO of Argo, which is Ford's autonomous Driving subsidiary.
They all got some kind of huge bonus. It supposedly amounted to a 14% rise in the aggregated quarterly expenses of Alphabets 'other bets', which lumps together a bunch of Alphabet's moonshot subsidiaries.
$94 billion in cash burning a hole in their pocket. $20x billion in new cash pouring in annually, expanding by 10%-20% per year.
In another two years they'll be net cash richer than Apple is (as unlike Apple they pay no dividend and have almost entirely avoided taking on debt).
There aren't very many large companies they can buy with that money, due to perpetually increasing anti-trust concerns. I suppose they could maul ~$25 billion buying Snap and probably get away with it due to Facebook (but comeon).
So what to do? Burn $1.1 billion on self-driving tech. Will it pan out? Doesn't really matter. Their search monopoly isn't going anywhere near-term (probably) and they'll have $200 billion in net cash in another 4 or 5 years. They could vaporize $11.1 billion on self-driving tech in the next couple of years and it would not matter, either to shareholders (oh, some would pretend to be upset) or to their operations.
I can't remember if it was Lazlo Bock or Eric Schmidt, but in one of their books (both?) someone reports a board member opining "You've created the first self-replicating talent machine." Having worked with those cats, that might actually be true.
The top German car suppliers and car makers have accrued 10x the patents on autonomous driving Google has [0]. Bosch alone has 3x the number of patents. Yet hardly anybody talks about Bosch and Continental and Mercedes/Audi/BMW etc.
To put this in context, $1.1b is about 3 weeks of operating profit for Google. Considering the upside for Google if their experiment pays off, this seems like a pretty smart bet.
Seriously, I'm mean what else are you going to do with a bit more than a billion dollars? Develop a full line of Google rockets and disrupt the space industry or something? :-)
But you raise an interesting point, what should we expect from 6 years of effort and 1.1B$ ?
The same tech that is being used in self driving cars could be repurposed for autonomous robots, and end up with a market much bigger than that of transportation. SDCs are just the easiest form of autonomous robots.
> SDCs are just the easiest form of autonomous robots.
That's a claim I have never heard before and I don't agree with. You have some kind of argument to support that drive-by claim? Since there are no fully autonomous SDCs yet you'd have to explain away any existing autonomous robots.
I was thinking that wheeled locomotion is still much easier than legged locomotion, artificial arms, grasping and fine motor manipulation. Robots still can't compare to a dexterous human, but a SDC is arguably a good driver.
Roads are a fairly structured and rule-based environment already designed with machines operating on them in mind. And those aspects that are not designed for the physical properties of cars are still designed to be easily perceptible by the human operators, e.g. signs, lane markings.
That can't be said for many other environments in which autonomous robots would have to operate.
100% of it would be tax-deductible. Corporations don't pay tax on any money spent on salaries, equipment, expenses, etc. Whether there would be ADDITIONAL subsidies, I do not know.
Google made 3.5 billion in their most recently reported quarter. It would have been even higher, except for a nearly 3 billion dollar 'unusual expense'. It's not exactly chicken feed, but 1.1B isn't an unexpected number given their focus on 'other bets'.
But am I the only one thinking that maybe we are trying to solve the wrong kind of problem?
If we had spent the same amount of money into creating a grid of mini-train I believe that most cities would already have some working prototype...
And mini-train is a control problem that we could solve yesterday...
(With mini train I mean something roughly big as a car but that is constrained to move in some predefined path, it will means that they could move way faster than cars and be more space efficients...)
Maybe because Google's self driving tech is fundamentally different than that Of Tesla's. Tesla relies on sensors and vision based system whereas Google relies on LIDAR. So I don't think they'd able to integrate the Tesla's self-driving tech into Waymo in any ways.
I think while the technical differences could be resolved the biggest blocker would be cultural. Musk has been super aggresive in rolling out more and more self driving features, where as Google and everyone else seems to be much more cautious.
Google already has a huge amount of power over where people go on the Internet. It seems having power, however indirect, of where people go in the real world is their next goal, and it is quite deeply unsettling to consider. Imagine self-driving cars that will be cheaper but take you to "sponsored places" before going to your destination, refuse certain destinations "for your safety", and play ads continuously while you're helplessly transported around.
Given that amount of power they will have if they succeed, spending $1B on the (very real) possibility of getting to that situation is definitely expected. And very very scary for everyone else.
1) Google / Waymo is by far the furthest ahead in self-driving tech.
2) Self-driving is possible in at least certain circumstances in the next decade or two (already is to an extent today).
3) Transportation / automobiles is an absolutely massive industry. Even companies like Dana, Inc (maker of axles) have $4bil valuations. It's easy to see how a self-driving tech maker would be worth far more than an axle maker.
4) Software tends to be a winner-take-most industry.
Given 1-4, $1.1B doesn't seem like much. Even if by chance it doesn't turn out to be a positive ROI, it seems crazy to think the $1.1B was a bad bet.
And given current prices for self-driving startups (example, Gm's acquisition of Cruise for >$1B), the current market value for Waymo would likely be >$10B.