Standardizing on sensors and communication protocols can only be a good thing, and the talks about making an "open" self-driving platform are enouraging. Time will tell if it's going to pan out and whether the need for quotes around open.
We are also going to need standardisation on reporting of car use to Governments. If we are no longer going to fund roads through fuel taxes, we need some other mechanism to do it. Proper charging of road use would unlock efficiencies for the road network overall.
Dunno about other places, but in New Zealand commercial trucks mostly have voluntary GPS tracking/logging to estimate their road mileage for road levies. (IIRC because they can subtract mileage travelled on private roads, but the per-km rate might be lower too.)
For most purposes though, a few linked to odometer reading, measured at mandatory "car health check-up" time and maybe amortised over the intervening period would be just fine, and require no new tech or tracking. Maybe tie it to insurance somehow?
Self-driving kind of seems like the type of thing that would benefit from a single standard protocol. It seems almost reckless to have a dozen or more organizations trying to invent it on their own, each discovering or missing pitfalls that can cost lives. The problem is that the asymetry of balkanized development leads to lucrative competetive advantages, at least in the short term as the first to "get it right" can reap a windfall. And I suppose they'd deserve it for the risk of the investment they spent.
Is there an alternative they could compete on instead of the core tech? Maybe their coverage of localizations to different sets of traffic laws? Even with a dozen parallel developers, eventually the tech will become commoditized. When that happens to a technology, often the competition surrounding it focuses on services instead. What would those services be with self driving tech?
> Is there an alternative they could compete on instead of the core tech?
To add to what tim333 said: The core tech doesn't exist yet, and it will be very expensive to finish developing, so the companies must be incentivized to pursue it, which pretty much necessitates competition. And indeed, the history of science and tech development shows that competition is exceedingly beneficial, and in the vast majority of cases far outweighs the minor costs of cooperability layers. The advent of a standard protocol generally marks when technological progress has stopped (although this can beneficially allow tech progress on tools built on top of the protocol).
Trouble is no one has really cracked self driving yet so it's probably premature to focus on one flawed system. Even Waymo it seems can't merge with heavy traffic https://news.ycombinator.com/item?id=19011890. We may need better AI beyond current machine learning algorithms.
Right. If the solution was just to put more miles on their vehicles. We'd see them with 10x the number of vehicles out on the road. They're not just a few months of "learning" away from cracking full self driving.
The financial incentive of getting there first is too great for them to not just throw money and hardware at the problem if that's all it took. That we see they're not doing that tells us they know they're missing some critical component. Which means all the cars out on the road now are mostly there for PR. Waymo is trying to get people comfortable with the change and to change regulations in states across the country, but they can't be "learning" anything (or they'd just scale that up and launch full self driving sooner).
No real breakthrough is needed. At this point, the current machine learning architectures and models we have are enough. The pains of R&D have been dealt with, and now it’s all about implementation and scaling, at least for the companies that have reached this point. I wouldn’t be surprised if these German car makers are still heavily behind the state of the art.
I don't think this is the case - both the breakthrough and your statement about where the car manufacturers are.
First, current machine learning is flawed. I provides reasonably accurate predictions for simple to medium complex problems with manageable amount of input variables, but requires a lot of data already at the high end of that. For high confidence and high dimensionality it's already uneconomical to even get the data in amount and quality needed. These models fail ultimately with complex scenarios like day to day driving.
As you can see with Waymo and Uber even the simple grid like traffic systems in the US with wide lanes and lots of space to cross are already a problem. Here we haven't even looked at organically grown traffic systems like in Europe or the sheer craziness they call traffic in countries like India. Effectively none of these are self-driving cars as they all come with a driver who regularly has to intervene as the cars prefer to drive around the block rather than doing a left-turn. These are problems they will not be able to solve even with 100x the mileage and cars they have today - scalability is not the problem here. Or rather it's uneconomical to solve it like that - by the time you gathered the necessary amount of training data the traffic systems will likely have changed significantly.
The underlying problem is that machine learning is impossible to debug and deep learning is based on the understanding of the human brain from the early 60s. These fields has advanced significantly since then and we will need both, new algorithms and new hardware, of what will eventually become sophisticated neuromorphic networks that truly mimic the human brain capacity. As it stands today, a human needs only a couple of examples to train, a machine needs millions. This is not only a compute problem but also a sensor problem. Today's sensor don't deliver the data at the necessary fidelity fast enough. We need algorithms that can learn more effectively, not harder.
I wouldn't be surprised if the German car makers are basically where everybody else is, but just waiting until the systems gets to a level where it's economical, truly safe and well accepted to get into mass production. Until then, they are probably fine with customers of Uber / Waymo play guinea pig.
There is a huge level of information assymetry that currently exists in the self driving industry. It is very much not an equal playing field. You have to understand that certain companies are so far ahead of other ones that they are pretty much competing in distinctly different leagues at this point.
You’re perspective on what the cutting edge of machine learning is shows your understanding of the current technological landscape is lacking. It’s understandable as the developments are happening incredibly quickly and the majority of the research is done privately, and unless you follow the release of academic papers very closely, it can be very easy to fall behind. Nevertheless it’s all happening regardless of whether you’re aware of it or not.
There are many required services: testing/certification, charging/refueling, parking, mechanical maintenance, inspections, cleaning (assuming ride sharing will be the top use case), dispatch, redistribution, picking up stranded vehicles, etc. On the commercial vehicle side, things get more interesting with logistics (loading/unloading of cargo, dock operations, etc.) and to-the-customer or to-the-door delivery.
It would seem a good idea for self driving cars to talk to each other to coordinate movement. Eg four way stop signs. So they’re all not just trying to go and seeing what happens.
Also one car part suppliers (Bosch) are making components interoptablity become easier.
This would be huge, and Cadillac has already deployed this in some of their models. The basic idea is simple: your car broadcasts its position and velocity vector multiple times per second (and your "car id" is periodically randomized for security), and the signal can be detected by any receiver within 500-1000 meters.
The simplest receiver could then trigger a flashing red "collision imminent" light if things look dicey. A complex receiver could employ active controls to "platoon" a set of vehicles, co-ordinate lane changes, or negotiate intersections.
The missing links, I think, are the need for
a) 1-2 meter position accuracy for platooning, and
b) 100-200 cm accuracy for lane-keeping/lane changing.
You'd still need to be an active, engaged driver: more "auto-pilot" than autonomous.
Carmakers will probably adopt the ZF system, as they did it with gearboxes. ZF's products are known to be expensive but are extremely reliable in general.
Imagine if all the car companies were competing against eachother on tech for brakes, seatbelts or how a steering wheel works.
Or Boeing and Airbus competing on fundamental safety aspects of flight.
Self driving should be a singular science with canonical baseline requirements and standards to how it should function and interact/react to the human environment.
So - this shouldn't be "news" - it should be the requirement to even play in the space.
> Imagine if all the car companies were competing against eachother on tech for brakes, seatbelts or how a steering wheel works.
Except for the seatbelt, That’s how it went. Especially with Airbus and Boeing who have chosen completely different approaches to flight safety system.
IIRC, in Airbus planes the pilot has no direct control. He gives inputs which are then accepted or rejected depending on what the computer thinks is safe; at take off you can pull the stick as much as you want, the computer will not let the tail hit the ground. In Boeing you have direct control but with better feedback on what’s going to happen.
What's interesting is that modern crash-rate statistics for the two companies are essentially the same, meaning there doesn't seem to be a superior choice of the two approaches
It's an often repeated misconception, but in Airbus planes the pilot has full control of every single control surface of the plane. Disabling the flight enveloppe protection system is probably necessary if you want to tail strike, but it is perfectly doable.
Which is eactly what we have learned to NOT repeat. My statement is still valid - We now know that letting many people die in vehicle safety trial-and-error is a non-starter.
A big reason for this is probably to ease the way to government regulations. Germany is very strict on these regulations and compared to, e.g. Arizona, reluctant to give experimental technology a pass for the street. If they pull on the same string, they have a much better chance to lobby for the required legislation to actually test and finally sell their self-driving technology in the important domestic market.
Historically the German auto industry has held a lot of influence over politicians. They've successfully lobbied against things like a national speed limit and for weakened emissions rules in the past.
More recently, they have blocked and delayed the EU's proposed emissions targets for 2025 and 2030 which would have required 40% fleet-wide CO2 reductions by 2030 and set minimum targets for zero emission vehicles.
(It should be noted Sweden, France, and Spain - who also have significant auto industries - are in favour of the new rules).
Not to mention that germany did absolutely fucking nothing against VW when dieselgate happened. Still waiting to see Winterkorn & friends under the guillotine for scamming the entire world.
What happened instead? Normal engineers got blamed instead of executives. Fucking capitalism and fucking CDU, SPD and FDP. The sole reason for their existence is to sack money from lobbyists.
When you have people like Schäuble, a piece of shit that was convicted for corruption in 1999, who was the minister of finance for the past couple of years in germany, it makes sense that germany does nothing against big corporate.
> Not to mention that germany did absolutely fucking nothing against VW when dieselgate happened.
VW (and Audi etc.) top executives are under criminal investigations. Their houses and offices have been searched multiple times.
> Still waiting to see Winterkorn & friends under the guillotine for scamming the entire world.
Sorry, we don't have capital punishment. And the guillotine is a bit too French for our taste, anyway (also, it was used by the GDR, so not a good role model).
While it is true that Germany has legalized and not so legalized corruption in politics, it makes no sense to single out VW. All large companies have "government affairs" departments which sadly provide great return on investment. If only the people had a government affairs department...
He failed to do anything about cum-cum/cum-ex.
He came up with a tax on fuel for nuclear power plants, that I am convinced was deliberately designed to be in violation of the constitution, leading to a 7 billion tax-revenue loss.
Not to speak of his views law and order and how the wrecked Greece because he doesn't understand economics.
I think this is great news. I think it’s a valid point that it’s hard to compete with Waymo (and probably Cruise from GM). I’d also worry what would happen if Silicon Valley cornered the market.
Your SelfDryv subscription has expired. Please renew via the SelfDryv app (available on iOS only) to continue using your car. Subscribe today for only $99/year (for our 5k Mile Package) + $20/year per each passenger.
Notice: After an incredible 12 months, SelfDryv will be ceasing operations in 30 days due to unforeseen circumstances. Please use up any SelfDryv Mile Credits you have within those 30 days, before your SelfDryv vehicle stops functioning. We apologise for any inconvenience caused, but we are currently unable to offer refunds or credits for your vehicle. Please see the attached document for alternative vehicles you can purchase going forward.
We apologize for the SelfDryv experience disruption experienced by some[0] customers earlier this week[1]. We experienced technical difficulties during a server migration that may[2] have degraded customer experience. The server migration exposed an incorrect dependency between the SelfDryv consumer hardware the the SelfDryv cloud services[3]. We have issued a fix[4] and offer 5 miles credits to any customer who was effected.
[0] About 100%
[1] 54 hours spanning Monday through Wednesday.
[2] ^H absolutely most certainly did
[3] Turns out part of the car's A/C UI dynamically fetches JavaScript on startup and our outage stopped serving JS, so A/C module failed to load. The worker thread that spawns the rest of the car's functionality freezes if the A/C module fails to respond. I guess we never tested that, I mean our internal network is pretty reliable, you know?
[4] We moved the JavaScript to a CDN so that means 100% availability. Presumably eventually we'll rewrite the entire client stack anyway thus implicitly killing this horrible dependency.
Also due to the fact that you forgot your password once and your last name sounds vaguely Ukrainian our AI has determined that your account should be shadowbanned, which means the app will appear to function as normal but the car will only drive in small circles around your home.
When you contact customer support our reps will explain that it must be something in their system but they can’t see your account and will transfer you to a manager at which point the call will disconnect.
We also apologize for remote-updating your car while you were driving on the highway and for the subsequent pile up but do note that we have disclaimed such events in your 238 page EULA that you clicked through after you purchased the subscription.
Har har. I do appreciate that this whole area is a rich vein to mine for humor, but I think remote updates will continue to be installed only when cars are in park.
> I think remote updates will continue to be installed only when cars are in park.
What you think doesn't really matter, and you are wrong about the 'continue' bit.
But it would be good if we had some laws around this stuff stipulating that software updates of moving vehicles are strictly forbidden, it seems like common sense, unfortunately not everybody has it.
People joke. But wouldn’t surprise me if whatever self-driving systems eventually come to the market will require subscriptions. And most likely subscriptions well in excess of $99 per year.
While this is said in jest, it might become the reality. Cars are designed for life time of more than 10 years, much longer for trucks, buses and commercial vehicles.
Not 100% sure if any startups are planning this far ahead in the future.
Level 3 autonomy works much better and level 4 becomes possible and scales faster earlier, if all carmakers cooperate and lobby for the same geofencing infrastructure and protocols in those roads where it's possible.
Level 5 is far away pipedream and research project that would boost global productivity and car sales for everyone. Cooperating in basic research is good.
Level 3 shouldn't exist. Multiple self-driving car makers have studied the issue and have concluded that Level 3 is too dangerous for humans, because human reaction is too slow to deal with the imperfections of a Level 3 system, all the while "expecting it to work flawlessly 90% of the time."
You can't just tell humans, "Hey, this works 90% of the time - it's up to you to decide when to take over that 10% of the time. Oh, and we've legally covered in the ToS footnote, just in case you think about suing us later! Best of luck, though! Thoughts and prayers for you and your family."
This can't be accurately called "collusion" since it's not a secret, it's still a cartel thought but lucky for them EU regulators have eyes only for US companies.
There's been increasing consolidation in the automotive industry, nothing new here.
Big challenges need big investments and it makes sense to work on open standards. Car makers don't want to become the next smartphone industry, with a race to the bottom among OEMs and Google reaping all the profits.
German groups know very well that creating distrust in the industry is not a winning strategy when everyone works with the same suppliers.
Disclosure: former HERE employee, executives always insisted on the fact staying neutral was key to long-term success and that we should welcome investments from other OEMs.
I admit that European countries have conflicting interests in the matter, however I'd rather have them look at more pressing issue first, like measuring and enforcing emission standards. "Unfair" competition against Google and Uber... I can live with that.
You're suggesting that the law shouldn't apply to the competitors of US companies because you worked for one of them and feel sympathetic?! That says a lot about the community makeup and voting pattern on HN.
I'm not suggesting anything, and I don't feel sympathetic towards my current or previous employers, this would be quite silly.
I'm sure German car makers take part in cartel-like behavior, like all big companies. However I don't see how building an open platform and working together on setting open standards could be seen as one of them. By the very definition of open standards, competitors can adopt them.
These are seriously big and well funded companies. Why do they need to share development? If Cruise can build a working prototype as a startup so can BMW. What is going on?
Part of me wonders if self driving tech ended up being harder than expected. These companies can't get it to work but also don't want to be left behind. Collaboration makes perfect sense.
Cost control. They are already working together on various different things. In the high end market, it doesn’t matter if you have your own self-driving software or you co-develop. People who purchase a Mercedes vs BMW select base on different things like brand loyalty and perception. To me (personally), I would always pick BMW over an Audi or Mercedes, even if they all came from the same groupation / factory, because it’s the brand I care about.
>In the high end market, it doesn’t matter if you have your own self-driving software or you co-develop.
Maybe. But what if it turns out that everything other than the core intelligence, communication and software platform features turn out to be low margin components?
What if these key functions turn out to be the only differentiator of any substance and the rest is just marketing blurb and a bit of component assembly?
Then companies that traditionally make their money with hardware will have a hard time accepting that. There are millions of lines of code in a modern car, but car manufacturers don't see themselves primarily as software companies.
Exactly. And let's not forget that car makers are also losing their key powertrain competency to electrification.
If they don't commit to really owning some of the new tech then there will be nothing left of them.
In my view, these traditional car makers are a slow motion train wreck of epic proportion.
How on earth can BWM and Daimler decide to pay more than 5% dividend instead of investing every last cent they earn into surviving the current technological revolution?
Apparently they have decided to die and pay out what's left to shareholders. This may be a legit strategy for a private equity fund, but if I was a German polititician or trade union official I would be horrified.
I'm not talking about licensing some basic patents. The question is, what will be that high margin thing that's worth optimising the heck out of for years and decades to come?
And who will own the customer? Who will define the software platform that creates all the stickiness and benefits from all the network effects?
If car makers continue down the path they are on now, they will own none of it. Like publishers, they will be reduced to go begging for the help of regulators to impose some form of Google tax.
There are not many differentiators on cars now either and we still have a range of "different" cars/brands. What makes you think the electrification will change anything(considering all cars will run the same self driving software)?
Electric motors are far simpler than ICEs. A lot of expertise that traditional car makers have will become obsolete. Many jobs will be lost, factories shut. The highest margins will no longer be in anything that car makers have traditionally done themselves.
I don't doubt that there will be many brands, probably far more brands than ever before, exactly because it will be so much simpler to make cars from off-the-shelf parts, or just order them from some white label manufacturer.
But where will the highest margins be? Wouldn't it be wise if car makers were asking this question and then invest in whatever that is with the greatest sense of urgency?
>considering all cars will run the same self driving software
If all cars run the same self driving software then I suspect that the maker of that software will benefit from some of the highest margins in the industry. I also think that there will be a lot of money in the network effects of linking the data that all those cars generate.
The car companies should be forced to share the traffic data(i.e a kind of traffic data neutrality law) so it can benefit all. After all it should belong to the people not to a single car company. I'm pretty sure there is enough left to innovate and increase the margins. Self driving should be a standard/norm, not a premium proprietary feature.
Absolutely, but these differentiators already exist while a lot of the high margin work that car makers are performing today is moving somewhere else in the value chain.
The key competency of the car manufacturers is integration and supply chain management. Sure they also develop some of the software and engines in house. But they would do so for electric cars as well. So there is not a huge difference. They are also very aware of the threat of being commodified software platforms and they are actively invest to prevent that. Both in ASICs to be not solely dependent on NVIDIA, as well as in software and maps.
Compare an $80,000 Mercedes to an $80,000 BMW and you're not going to find much in the way of objective material differences. At that point your choice will mostly come down to personal preferences about dash layouts, precise seating styles, the sound of the engine, the precise feel of the pedals, etc, which all pretty much fall under "brand".
Those things are real differences in how the car ia designed. If it was the brand you cared about, then you would buy BMW even if you didn't prefer those things in BMWs.
They're real differences in how the car is designed, but barring some outliers, they're not differences that are objectively better in one or the other. I consider those kinds of design choices to be things that fall under "brand" as a whole.
Why not?
I mean this companies already have many parts of the car designed and built by third parties, there is no need to have everything designed and built by themselves.
So if you have 3 big companies collaborating you will have more money and maybe more important more training data for the self driving program. Now Uber,Tesla and Google do not share their data(probably the data is not compatible either) but if this data was shared progress could have been faster and lives could have been saved.
We hear a lot those days in Europe how Germany rested on old school industry during the last two decades, and failed to anticipate the big changes coming. They relied a bit too much on « good old german quality » without trying to rethink itself.
But it could be just jealousy speaking, since germany has been doing so well compared to the rest of us, i don’t know.
Those people don't understand that a lot of "german" manufacturing is happening in low cost countries. The bulk of the manufacturing remaining in Germany is highly automated.
I suspect it's mostly an optics/ignorance issue. If you never set foot into a modern factory, you could legitimately believe that german industry had not changed, simply because it's the same names as decades ago. But once you get the chance to take a peek behind the curtains, you quickly realized how high-tech those "dinosaurs" are.
I’ve heard so many cities claiming to be the capital of europe’s start up scene. I’ll believe it once one of those city manage to build a competitor to any GAFA.
They aren't tech companies. It's like asking why satellite makers don't build their own rocket. The algorithms used in most cars whether it's deep RL or something even more specific isn't easily understood and worst yet scalable by many people. Say nothing for the custom hardware google uses in the form of TPUs.
While they are surely no leading innovators in deep learning saying they are not tech companies is a bit ridiculous, as they innovate in many areas of technology. Sure, VW and Daimler don’t produce their own chips (as far as I’m aware) as they rely on partners like Bosch to do it for them. Bosch sensors and ICs are found in many devices btw (even iPhones I think). So I’d argue there’s plenty of innovation here, just nothing that is very visible for most folks in “tech”. Btw Tesla’s cars use electronic and mechanical parts from around 40 different German suppliers if I remember correctly.
They absolutely design their own chips, typically the precise details are outsourced. But they are actively developing chips from MCUs (motor control units) up to simple things like seat heating sensors. At the moment they have pushed the German government to invest billions in German made inference accelerators for cars, to not be solely dependent on foreign chips by NVIDIA and Intel for that.
That's a very narrow definition of tech companies. I wouldn't be surprised if VW, BMW and Daimler together employ more software engineers than Google or Apple. Don't forget that modern cars use over a hundred micro-controllers.
That's true, but only indirectly. Since these companies mainly assemble cars from components designed and built by suppliers, the software engineers are working for suppliers like Bosch. If you talk to the tech people from, e.g., BMW, their roles are managing budgets and projects implemented by external partners.
This is also why they want self driving to be just another standard function that they can buy from a supplier, while letting the suppliers compete away their margins.
VW (in this case AUDI) founded AID (http://aid-driving.eu/) as a software centric powerhouse to work on autonomous driving for all of VW. They employ lots of software engineers directly (I work there)
They don’t, and the parent is right that they are no software companies. Cars carry an enormous amount of software, but most of it is contracted and built by tier1 suppliers like Bosch. I worked at one of the mentioned companies and was the only person in my department (of >500 engineers) who had production software run software running inside cars.
In the meantime they realized they need to get more into software, but it takes a lot of time to change course for tankers.
Nobody who is sane uses deep reinforcement learning for autonomous vehicles, far too dangerous. Remember, the first 99% is easy, it's the last .999% that's hard.
- Commoditize the stuff you don't want to compete on. I don't think German companies want to compete on self driving - they are positioning their product as a premium product and thus compete on performance, quality and emotions.
- Have a common interface to the outside - there is more to self driving than sensors and software: Legal framework, insurance market, road standards, parking standards, home garage standards, refueling standards. These external partners need to change and if too many competing priorities and possibly contradictory demands are coming nothing will change.
Self driving is certainly harder than it looks. If your market is the world and not just the relatively empty and wide grid streets in the US things are even more complex.
I'm sure there is some fear of losing out. More however I think about the limited upside: The cost for developing self driving vs the later ability to charge for it is very limited from a car manufacturers POV: Sensor's are bought in and software is either bought or very expensive to develop and run. Differentiation is minimal: Either considered safe or can't be sold. Self driving here has different economics from EV.
