Hacker News new | past | comments | ask | show | jobs | submit login
Honda will use GM’s self-driving technology, invests in Cruise (arstechnica.com)
351 points by BooneJS 6 months ago | hide | past | web | favorite | 245 comments

What's interesting is that Honda is a company that has always prioritized being in control of its own destiny. What made them interesting in the past is that there are many examples where they'd engineer their own solution to a problem, rather than license IP.

Quote from Mr. Honda:

"We [Honda] refuse to depend on anyone else. We will not copy foreign products nor pay royalties for the use of other companies’ patents. We don’t intend to get support from the government, either. I’m making it clear that we will do it our way." [1]

Notable examples:

- Hondamatic transmission [1]. For better or for worse, it was only until recently that Honda started using transmissions from ZF like other automakers. Arguably because automatic transmissions are on their way out, and they're focusing on other drivetrains.

- Honda's hybrid drive: They developed their own two-motor system rather than license Toyota's (Rather brilliant) solution like other companies. Their first system in the Insight (gen 1) not withstanding.


It seems like there's going to be an inevitable commoditization of car manufacturers in the future. Equally interesting is GM ceding some ownership of potentially game-changing tech.

[1] https://world.honda.com/history/challenge/1968hondamatic/ind...

Commoditization of self-driving tech, yes; I don’t think car manufacturers will become (more) redundant, but eventually everybody‘s autonomous driving systems will either be hard to distinguish (like gasoline motor tech), or even completely regulated by legislation. Can legislators really allow that cars drive around with a system that is inferior to a competitor‘s? This will put eg every pedestrian‘s live at risk.

I‘m pretty sure self-driving tech will not be a big differentiator, just like airbags - almost every car has them, they are a given.

> Can legislators really allow that cars drive around with a system that is inferior to a competitor‘s? This will put eg every pedestrian‘s live at risk.

They already do this with everything else. Vehicles like Mercedes and Tesla are significantly safer than some competitors. Legislation sets the floor, not the ceiling. Or at least it shouldn't create a ceiling.

And it's also the case that SUVs and other large vehicles pose a greater risk to pedestrians and other motorists than smaller ones, but legislators haven't done anything about that.

> They already do this with everything else. Vehicles like Mercedes and Tesla are significantly safer than some competitors. Legislation sets the floor, not the ceiling. Or at least it shouldn't create a ceiling.

I think the difference is those are, for the most part, internal safety features for the car's driver and passengers, whereas self driving effects mostly people not in that automobile. Sure, there are some features in cars that aren't, but for the most part, a driver knows (should know) their vehicle, and someone driving a 1978 station wagon and someone else driving a 2018 Volvo won't make that much of a difference on the roadways. Driving a car with a bad autonomous mode vs. a car with a good autonomous mode could be a significantly bigger difference for pedestrians and other drivers than older and newer cars.

2018 will be much heavier, bigger, with poorer visibility thanks to huge crash and rollover resistant A, B and C pillars. It will also have passed the pedestrian impact tests and have enormously better brakes and handling. So even though heavier it'll stop a lot quicker, and do less serious damage to a pedestrian at town speeds.

That's quite apart from any safety features like air bags or ABS.

The big A pillars make it harder to see pedestrians though. Wonder if net injuries are actually reduced?

What about safety differences that don't affect pedestrians and other such (nigh-)stationary objects?

For instance, I could easily imagine that some cars might have more powerful kinematic-prediction software+processors and therefore are able to use smaller safety margins, allowing for more aggressive/liberal driving-behaviors than would be possible to safely perform in self-driving systems that aren't so well-endowed.

(Then again, depending on how good the first batch of general-purpose self-driving systems are, it may be possible that traffic safety margins among the self-drivers will be reducible enough that even the base models are "good enough" for anyone who doesn't encounter too many meatbag-drivers on their commute.)

I'm pretty sure the 78 car is way more likely to kill someone

>Sure, there are some features in cars that aren't, but for the most part, a driver knows (should know) their vehicle, and someone driving a 1978 station wagon and someone else driving a 2018 Volvo won't make that much of a difference on the roadways.

Brakes, handling, visibility - better car - better response - safer for others. Also being lighter lowers potential damage impact. Not to mention stuff like sleep at the wheel detection etc.

Except that cars have mostly been getting heavier since the 80s... And electrics even moreso.

Since this touches on the safety of old vs new cars, I'd just like to leave this link to an "old vs new" crash test video here for anyone who might not have seen it before. https://youtu.be/C_r5UJrxcck

Historical vehicular deaths: https://en.m.wikipedia.org/wiki/Motor_vehicle_fatality_rate_...

As an aside, strangely the Great Recession apparently improved safety during, even when adjusting for miles traveled.

The 2000s were a big era for improved safety. The NHSTA began studying and testing a wider variety of different crashes events and used ATDs (crash test dummies) designed to simulate women and children. Side-impact airbags became mandatory equipment in the USA in 2007 and electronic stability control in 2008.

It's not something people realize because they think car safety improvements slowed/stopped once airbags became widespread. But really, the mid 2000s are when cars began to get very safe. By 2011, the NHSTA had to make their star rating system much more strict because nearly every car had a achieved a five star rating.

That's also when Euro tests changed significantly. The difference between cars this century and earlier is night and day.

A Volvo 940 from about 1993 head on at 80mph with a 2004 Renault Modus - the first small car to get 5 stars in the new Euro tests.

I would have bet on the Volvo. No. Dramatically no.


Just from a metrics geek POV: why didn't they add more stars instead of renorming the scale? Progress in car safety would have been more apparent had the scale gone unchanged.

The wanted the current standards to seem normative. This makes it much harder to roll them back.

Couldn’t the floor be a bit higher?

I’m also not sure why everyone hasn’t adopted the Toyota model of including basic safety tech in all trims...

> Couldn’t the floor be a bit higher?

Sure, but that's what actually happens over time. First airbags are introduced, then we get some evidence that they actually work, more manufacturers introduce them and we get even more evidence, then eventually the law requires them.

It's necessary to be deliberative before mandating things because not everything is a good idea, e.g. automatic seatbelts in the 90s.

How long has that been going on? I'm looking at buying a Sienna but needed to buy new because they introduced their Toyota safety sense on all their trims for 2018 only.

It’s a new thing, maybe since 2017? Regardless, I think it’s the right way to do it.

I’ve been wondering how much of a business decision this was, rather than a moral one. A cost-benefit analysis must have been completed that found the sweet spot for a bulk purchase of the safety cameras and how many more additional customers it would attract to purchase their cars.

There seems to be a total lack, from our regulators/politicians about even establishing a floor on self-driving car. That certainly will go well.

The only Tesla safety numbers I have seen have come from Tesla. Do you have any independent numbers?

Look at car safety. You have to pass a benchmark test but after that it’s up to the OEM to go as high as they want.

> Can legislators really allow that cars drive around with a system that is inferior to a competitor‘s? This will put eg every pedestrian‘s live at risk.

As if anyone actually cares about people outside the car.

I know the prevailing view on HN is that drivers don't give a shit about pedestrians and cyclists, and I'm willing to believe it's true that drivers are not often penalized for injuring them, but are most drivers really that bad?

I'm not a terribly conscientious person, but I'm definitely worried and raise my level of caution any time I see pedestrians standing near the street or when I pass a cyclists. I know that they will be seriously injured if I hit them, and I really don't want that to be my fault. even if it isn't my fault (or if it is, but not in a way that a court will find me responsible), I am quite afraid of having to go through the whole legal process that could ensue.

>I'm willing to believe it's true that drivers are not often penalized for injuring them, but are most drivers really that bad?

Most drivers? I'm guessing not. Many drivers, however? I believe so. It doesn't take that large a percentage of the driving population for uncaring/bad drivers to cause havoc, and make it very unsafe for you to be around them. For instance, if you're a bicyclist on a busy highway, many hundreds of cars could pass you on one trip. It only takes one of those who hates cyclists to run you off the road (yes, this actually happens sometimes, in some places).

People really don't care. It's not so much a callous -- "I could see this guy die in front of me without batting an eye." It's more that all the stuff you mentioned: injuries, an incident being your fault, court dates, etc. legitimately never enters their head. They do not picture that someone might be on a bicycle in their blind spot until someone is.

I think it probably changes when people have children, because they can imagine their child being hit, which helps to make that connection. For most drivers without children, yes really, they do not think of that stuff.

i guess it's just my anxiety showing then. probably not a bad thing in this case.

If you, knock on wood, end up accidentally hitting a pedestrian hen you suddently care A LOT.

Additionally it spreads risk/reward. You’re using the ‘standard’ industry system, not some bespoke thing. So you’re negligent because you’re ‘behind’. You can’t be ‘behind’.

You can’t leap way ahead, but what are the chances of any one automaker doing that at this point? Unless you’ve already had a program going for years under very heavy investment it may be too late to catch up let alone lead.

Not just commoditization of self-driving tech. Many people will switch from buying cars to hailing self-driving taxis. And nobody really cares about the brand, styling or horsepower of the taxi they take to work. To that degree each manufacturer faces erosion of their competitive advantage.

Companies change. Mr Honda died a quarter century ago. Is that still the company philosophy?

I asked my friend last week if his new generator was a Honda, and he told me it was a Subaru -- which was basically just as good since Honda just contracts out to Subaru for building all their small engines, anyway.

I won't claim to know what their current philosophy is, but I do think that going their own way is in large part what made them notable to begin with. They were known as the quirky Japanese automaker that made fun-to-drive (for the $$) cars. They did things like:

- S2000: A powerful and lightweight naturally aspirated car with an insane redline of 9,000 rpm!

- VTEC was Honda's solution to keeping displacement down, without sacrificing power and efficiency, and what made their engines so buzzy.

- They were first to the hybrid game with the Insight (gen 1), beating Toyota's Prius to the market.

- Until the early 2000s they were putting double-wishbone suspensions in their econoboxes. This is no longer the case, to the disdain of car enthusiasts and modders.

You'll notice that they've stopped doing/stopped coming up with quirky innovations, leading them to become just another automaker.

I think the Fit remains a remarkable car from a design perspective.

Modern car companies fundamentally do 3 things: make engines, make transmissions, and assembly parts into cars. Nearly everything else is outsourced. These are all expensive complex things that little companies cannot make. They don't make their own seats, alternators, tires... (Honda is an exception to the above rule in general).

Tesla makes their own seats. They've also had four seat-related recalls in the past 5 years.

Which wasn't such a great idea if you ask me. WHy bother with seats if they can just be bought from suppliers specializing in seats? Same for manufacturing. It puzzles me why Tesla did not outsource Model 3 production to a company like Magna or Valmet. Both of them are producing Mercedes, Ford, Porsche, Peugeot, Opel, BMW at a lot of others on the same assembly lines. And you can be sure that this production is running smoothly without a CEO living on the shop floor or meeting volume targets at the cost of quality. If that would be the case, these sub-contractors would be out of business already.

And there are reasons for that as well, usually the risk share form these outsourcing activities is very favorable for car makers. As a sidenote they also outsource a lot of engineering work, at leat in Germany. And that implies a risk for traditional caqr makers way beyond competitors like Tesla. FOr an EV half the activities performed by tradintional car makers is not needed anymore. That impacts head-count, unions, sites, you name it. At the same time the externally produced value per car is further increasing, giving the electric drive train supplier a lot of power through the automotive supply chain. So the biggest challenge for traditional car makers is to remain the most important automotive players while developing EVs of there own. As long as that position is maintained they will do just fine, production is not their problem, neither is money (at least so for).

Many car companies outsource transmissions now.

ZF makes an 8-speed that’s used in nearly everything from Jeeps to Astons to Fords.

The ZF 8HP is used by probably half of the premium cars coming out of Europe.

It's in the Rolls Royce Phantom.

The Bentley Continental Flying Spur.

The Aston Martin V8 Vantage.

The Maserati Quattroporte.

The Lamborghini Urus.

The Jaguar F-Type.

The Range Rover Sport.

The BMW 1 series, 2 series, 3 series, 4 series, 5 series, 6 series, 7 series, X3, X4, X5, X6, Z4 and the M5... the Audi A4, A5, A6, A7, A8, Q5, Q7...


So was their old six-speed.

The ZF 6HP automatic was a game changer when it was released. It had six forward gears with the ability to lock the torque converter in any gear and it was cheaper to produce than most four speed transmissions. They ended up dominating the RWD automatic transmission market.

I have a 335i with a 6HP21, and the converter lockup behavior was the first thing I noticed when I got the car. It usually locks up in first gear, which I’m sure helps with fuel economy a lot.

That said, the overall shift quality could be a little better, IMO. I’m no transmission expert, but I think it basically shifts “clutch-to-clutch” instead of using sprags like a traditional automatic, so it’s really sensitive to timings. If the second clutch engages too late, it’s in neutral for a bit. If it engages to early, it’s in two gears at the same time.

For example, if you’re accelerating lightly and it’s about to shift, then you suddenly have to hit the accelerator harder, sometimes it feels like it has already committed to shifting under the lighter load and it almost feels like it’s in neutral for a split second before banging into the next gear.

Not sure how much of this is related to the BMW shift maps or the transmission itself.

All that said, the transmission has proven to be very reliable. People in the 335i community are putting well over 500 ft-lbs of torque through them and they seem to be holding up, and my car has 211,000 miles and it shifts the same as it did when I got it with 40,000 miles.

Also, people with newer 5.0 Mustangs don’t seem to have many complaints, and it’s basically the same transmission.

What made Toyota's solution brilliant?

Interesting history about Honda, thanks for sharing!

It's genius in it simplicity. The car doesn't actually have a transmission and uses planetary gear sets and two electric motors to fake one.

The gasoline engine is used as the "baseline" power, and the motors then either take power away from it to charge the batteries, or complement it to help in acceleration, or just bypass it completely.

When the car is running in pure electric mode, the gas engine is stopped, and the motors work together to spin the output shaft at the desired speed.

When the car is using mixed power, the gasoline engine is working at a desired power level, one of the motor/generators is adding additional torque, and the second motor is used to make up the rotational difference between the sun and planet gears. This blending of gasoline and electric allows the wheel output speed to be independent of the gasoline engine speed.

When the car is charging the battery, the gasoline engine is run harder than it needs to be to move the car, and one of the motor/generator units is fighting against it, using the excess torque to generate power.

So, there is no transmission, no "shifting", the engine and mgu's are permanently connected to gears and simply spin them at the right speeds for the desired power and speed output.

This is standard since decades in non-electric train engines. https://en.wikipedia.org/wiki/Diesel–electric_transmission

That is incorrect. Toyota's Hybrid Synergy Drive is able to power the wheels with either 100% electric, or 100% gas, or a combination of the two using planetary gears. This allows the use of a more efficient engine running the Atkinson cycle, with the electric motors filling in for the reduced low-end torque.

At highway speeds the engine is driving the wheels directly, which is more efficient than converting torque to electric power to drive an electric motor. Honda's two-motor system runs much like the transmission you linked, but again drives the wheels directly at highway speeds.

I must have misunderstood the GP then, sorry :(

The Toyota system seems fairly different than a diesel-electric train system to me. As I understand it in a diesel electric system the three components (diesel engine, generator and electric motor) all do one thing the whole time and there's no batteries. In the Toyota system however the electric motors can flexibly act as either motors or generators. The gas engine in Toyota also doesn't need to always be on to provide power.

Toyota's hybrid system differs from trains in that there's a direct mechanical connection from the engine to the wheels, boosting efficiency by not requiring all of the power to go through the motors.

It's really devilishly clever. You have to see it to appreciate it.

This site has a nice overview: http://prius.ecrostech.com/original/PriusFrames.htm

This kind of mechanical device where there are more than two shafts are always cool. Classic video about differentials from 1937 (!): https://www.youtube.com/watch?v=yYAw79386WI

Traditional automatic transmissions use a set of planetary gears. The idea is that you can spin the ring gear and planet gears and different rates, which will alter the speed of the output shaft.

Traditional automatics use a two input shafts capable of spinning these pairs at different speeds. A series of clutches select different pairs of ratios for each "gear". A four speed auto might have two ratios for the ring gear and two more for the planet gears.

The Prius transmission takes this idea, and uses electric motors to vary the speed of these gears to create a continuous change in output speed. It's brilliant because it's so obvious when you know the trick. The end result is something that seems really complex, but is arguably even more simple than a manual transmission.

It also eliminates most of the components of an automatic that are prone to wear and breaking, which is a huge boon for reliability.

The Japanese have a track record of not being able to develop complex software on a short timescale. Their culture and bureaucratic institutions are an impediment to new disruptive technologies. That is a big factor in why the Koreans and Chinese are eating their lunch in consumer electronics.

Just like Japan, Sony as a prominent example, did before with Europe and the US?

I think it's a bit of understanding what your core competency is. Software and AI is substantially different than automotive.

That approach worked back when cars were simple, and mostly mechanical. But now the R&D cost of developing major automotive components has risen so high that smaller manufacturers just don't have the resources to go it alone any more. So they have to outsource or form partnerships in order to achieve sufficient economies of scale.

Much like the 1970s gas crunch, when American car makers couldn't innovate small cars fast enough, and bought Japanese car companies to bridge the gap.

By investing, they co-own the company (at least partly), right? Seems smarter than trying to license the tech in 10 years.

I mean, they'll still have to license the tech from Cruise, they'll just be licensing from a company they have a minority stake in.

You can't just use tech from some other company even if you are part-owner. Heck I don't think you can even do that from a fully-owned subsidiary. They have to have structural/legal independence or it's really just a different branch of the same company.

If it's a fully-owned subsidiary you could draft a contract of licensing in exchange for a token $1 or something like that. (Or is there regulatory oversight?)

That would constitute tax fraud in many legislations.

That's one of the principal points that tax auditors check in a group, although it's oftentimes hard to determine a market price.

I saw on Bloomberg TV 5%...though I think that is 5% of Cruise and not GM. GM bought cruise for $500 million cash. This investment puts Cruise at $15 billion. Nice return...

Is this because Softbank, a japanese group is invested in Cruise?

> automatic transmissions are on their way out

Did you mean to say manual?

I think he’s referring to conventional automatic transmissions being replaced by CVTs due to size/weight/efficiency.


And/or direct drive electrics.

Manuals have been out for a very long time, it's just some regions haven't cottoned on to it yet. Autos are also on the way out as newer transmission styles are rapidly replacing them...CVTs for example. But the latest gen hybrids don't even have that and most electrics are basically direct drive.

With a post money valuation of 14.6 billion dollar, Cruise surpasses Dropbox as the third most valuable company that YC has invested in. The valuation of Airbnb is 31 billion, Stripe 20 billion, Cruise 14.6 billion, Dropbox 11 billion.

YC doesn't hold shares in Cruise anymore, since the company as a whole was sold to GM for 1 billion in 2016. After the acquisition, GM invested another 1.1 billion, Softbank Vision Fund invested 2.25 billion and now Honda invests 2.75 billion. So more than 7.1 billion dollar has been invested into Cruise. It's the best capitalised YC company. Airbnb has raised 4.4 billion, debt financing included.

The cruise valuation is pretty suspect, as reported by The Information [0].

> When SoftBank agreed to invest $2.2 billion in Cruise, the self-driving unit of General Motors, most of the attention was on the valuation of $11.5 billion. But a closer look shows that could be inflated by terms. SoftBank’s initial $900 million investment in the firm is at a lower valuation of $8.3 billion; it is only if products are ready to be commercially deployed that it has agreed to invest more money at a higher valuation.

With unicorn-scale deals, the topline number is sometimes not reliable, as it may come with high liquidation multipliers or other restrictive agreements.

[0] https://www.theinformation.com/articles/softbank-asserts-mor...

Honda only invested $750 million in Cruise itself, the 2.5 billion is the total amount of what they are investing in the project, which includes the development of a new vehicle. As per the linked article.

That's a pretty insane multiple on the acquisition price in 2 years, is there anything you can read into that or is it pretty meaningless because of all the post-acquisition investment money being pumped in?

Almost exactly half of its value is represented by the money that's been pumped in (7.1B was the sum GP arrived at). The rest...

> It's the best capitalised YC company.

"YC doesn't hold shares in Cruise anymore, since the company as a whole was sold to GM for 1 billion in 2016.".

It's no longer a YC company.

Contextually, it's pretty clear that they're defining "YC company" as "a company that YC invested in".

I hope they roll out augmented driving.

I don't need full autonomy everywhere, but there are a few things that definitely could be rolled out now.

Give me a virtual bumper. Never let me get closer than a meter to the car in front of me. If it stops, stop my car too. Get moving again? Go.

Stay in the lane at low speeds. Combine with above, get a far more relaxed experience in stop and go rush hour traffic.

Extend the virtual bumper to parking. Rather than video and beeping or full autonomy, just stop the car from hitting anything, 5cm distance control.

Or does this exist? Haven't seen it in my Tesla test drive.

>Give me a virtual bumper. Never let me get closer than a meter to the car in front of me. If it stops, stop my car too. Get moving again? Go.

This is Honda's Adaptive Cruise Control (ACC) with Low-Speed Follow


I have ACC, and it’s great. But adding the Low-Speed Follow makes it SO much nicer. I’ve driven a Bonda like that.

Otherwise ACC is only useful on freeways that aren’t too crowded. Nice, but not nearly as nice as being able to hassle traffic jams.

I'd like this to be simple, watching the video this only starts at low speeds. Virtual bumper would be there, always.

But, ACC seems way better than adaptive cruise control in my KIA, which disengages at crawl speed rather than stopping.

Most of these features exist already in vehicles that have adaptive cruise control and parking assistance. I don't think any of them are fully flushed out but they are getting better.

A lot of cars have Lane Keep Assist and Automatic Cruise Control, but most are not nearly good enough to rely on.

Comma.ai ships a plug and play device that will override the cars software with better software and allow you to do just what you are asking. It's wonderful for stop and go traffic.

Works right now on most newer Hondas and Toyotas, as well as a random assortment of other vehicles. You can even port your own car if it's not supported and has the appropriate hardware.

A lot of cars have Lane Keep Assist and Automatic Cruise Control, but most are not nearly good enough to rely on.

Yes. Here's a road test, from 2013, of three anti-collision systems.[1] These things are all over the place on performance. As many Tesla crashes and deaths have demonstrated, Tesla's system won't stop for stationary obstacles. (Road barrier, street sweeper, stalled car, fire truck - the list goes on.)

There needs to be a minimum performance standard for these things. A reasonable minimum goal is to mitigate collisions down to a 35MPH difference or less. With seat belts, air bags, and crush depth, almost anything below 35 MPH is survivable now.

[1] https://www.youtube.com/watch?v=PzHM6PVTjXo

All the features you described are available right now. E.g. VW Golf has them off the shelf, and I‘m pretty sure the same is true for 2018 Kia/Hyundai cars. Honda is notorious for lots of assistance systems, I guess they don’t lag behind.

And those things also brake for other cars and pedestrians if they project that a collision would happen. There will be a lot less deaths caused by road accidents once everybody upgraded their cars. Exciting times!

One thing I'd add: Head-up-display. Project the most important information up to the bottom of the windshield so I don't have to look away.

This would be particularly useful with a lane-keeping solution, where it could show exactly where it thinks the lane is - overlaid right on top of the actual lane, using perspective skew to match the driver's view.

I don't know how well it works in stop-and-go traffic, but a friend's 2-year-old subaru has the ability to follow the car in front and keep itself in its lane. And i can't imagine subaru is the only one with that technology.

Subaru's tech is fine, but lacks the radar that Teslas bounce underneath the car in front of you to get a heads up that traffic is either moving or not ahead, which is essential for eliminating the jerky stop and go behavior.

Subaru's system was designed for maximum safety at minimum cost. I don't think comfort was a primary design goal.

There’s that subaru commercial that implies it can see 2 cars ahead

All exists already. "Virtual bumper" so you don't hit pedestrians, too.

One thing I dislike with Tesla hype is that practical technology from other companies are buried. Not Teslas fault...really the media.

I'm in the market for a new car. Many manufacturers have similar tech, here is Toyota. https://www.toyota.com/safety-sense/animation/lda

I'd like to see self-driving first appear for limited-access highways: the driving environment there is far simpler than other places and requires far lower computation. Just get on the on-ramp, hit "self-drive", and lie back and relax or sleep while your car drives to your exit.

Features that add extra security are great, but I am against any feature that lets you stop paying attention. When it comes to self driving, it must be all or nothing. Half self-driving doesn't work.

Adaptive cruise is a life changer.

Sure but adaptive cruise only helps you relax your leg, you're still paying attention to the road. I'm specifically talking about anything that lets you not pay attention for some time.

That's rather surprising because pretty much everything you are asking for is already available in Tesla cars (it has been in mine for quite a while).

Will a Tesla (or any car) really prevent you from running your car into something if you try to?

I think what the comment was about would be trying to park your car and as soon as your bumper is 6 inches from theres, stopping you completely. If you push down on the accelerator, nothing would happen. Inertia aside, it would be impossible to hit another car.

>Will a Tesla (or any car) really prevent you from running your car into something if you try to?

I guess it depends on what you mean by "try to".

I drive a VW; if I'm inching out of a parking spot, it will automatically stop the car if it detects traffic. I suppose I could tromp on the gas at that moment and blast out into oncoming cars, but I'm not sure why I'd want to...

Same with catching up with slow moving traffic at highway speeds. It will detect the vehicle (or object) in front and apply the necessary braking pressure. The slower the object in front is moving, the faster it will apply harder pressure (and flash a warning to apply the brakes manually).

Essentially, if you don't react, the car will. You can then override it.

I know first hand that my car will not prevent a crash if I press on the gas pedal near an obstacle (that's how I hit my garage wall).

The GP was asking about a scenario of stop and go traffic (if I understood it correctly). I've used autopilot on a couple of occasions on I-5, and indeed, my car stopped when the car ahead did, then it resumed when the car ahead moved forward, maintaining distance and keeping lanes.

It's a different use case from the one you are describing.

Honda and Waymo have been 'in talks' for years, with the first formal announcement of a vaguely defined partnership arriving in April earlier this year. Supposedly they were going to build a purpose built delivery vehicle together.

The odds that that deal has gone up in smoke I'm guessing are high.


I've noticed in SF at least Cruise seems to have removed branding from their cars (or I believe they are their cars). I'm curious what that implies -- either about the sentiment of self driving cars or future branding partnerships.

perhaps it's just about collecting more accurate test data - if self driving cars are obviously marked as such, the behaviour of other drivers around them may change.

The turret of sensors mounted on top still might be a bit of a giveaway.

Well there's many companies doing AV testing with sensors on their marked + unmarked cars. Not just Cruise.

I have not seen this at all.

Interesting. If GM were super optimistic about Cruise's technology based on internal metrics, you'd think they'd invest their own money, or at least take on debt financing at GM's cost of capital. That they are taking outside investment from Softbank and now Honda suggests there is enough uncertainty that they want to limit their downside.

A quick search suggests that GM is mostly debt financed, with a weighted average cost of capital of only 5%.[1]

[1] https://www.gurufocus.com/term/wacc/GM/WACC-/General-Motors-...

I would argue taking outside investments to finance Cruise is a great strategy for both Cruise / GM. Because:

1. as a self-driving module supplier, the more adoption from the market the better - it's an endorsement and it helps Cruise to build better products with more data points and more customers.

2. it reduces GM's financial risk if Cruise folds (unlikely); also it also improve Cruise's Survivability if GM is in trouble. As an auto module vendor, diversification is key.

3. The money may help develop the tech faster or make it cheaper than would have otherwise happened - both benefiting GM

Bringing Robotaxis to market is very capital intensive. I don't know how much money GM has to throw around, but I'm pretty sure it's a lot less than their nearest competitor, Alphabet.

Waymo announced purchasing orders for 60,000 FCA Pacificas and 20,000 Jaguar IPaces to be converted into robotaxis. If we're to assume a crude ballpark estimate of $150K apiece for those vehicles, then we're talking a $12 billion just for that. GM can't go up against those kinds of capital investments without help.

Argo.ai, Ford's answer to GM Cruise, is also seeking partners/funding from outside Ford. Wish them luck.

> Waymo announced purchasing orders for 60,000 FCA Pacificas and 20,000 Jaguar IPaces to be converted into robotaxis. If we're to assume a crude ballpark estimate of $150K apiece for those vehicles, then we're talking a $12 billion just for that.

You do know that GM makes automobiles right? If GM wants 60,000 Chevy Cruises to use for robotaxis, they can just keep their lines running for a few extra shifts. Their cost for the same vehicles is COGS, which is something like half the MSRP of an automobile. Doing so would actual increase the profit margin on their products that are sold to outside customers. GM also has ~40% more revenue than Alphabet and could easily borrow $12 billion if it were strategically important.

Man people really don't understand the weight of traditional autos.

Also, if you buy 20k cars from Jaguar (mind, 20k iPace is more than a month worth of Tesla Model 3s) you will be paying a lot less than 150k per car.

There's a lot of extra hardware that goes into a robotaxi. Waymo doesn't just sprinkle pixie dust on them to make them drive themselves. The vehicles, their sensors, compute, adn the 100 miles of wiring needed to hook it all up is just hardware costs. There's all sorts of logistics and real estate that needs to be acquired. Every new area needs to be mapped and tested. They run call centres, and pay thousands of people just to do data annotation.

The plan for all these companies is to scale up to deployments of millions of vehicles, they won't see ROI for a decade or more, and that's only if everything works like it's supposed to, it's still technically and unproven business model built on an unproven technology.

And sure GM can raise funds if they really want to, but they've gone from 0 to $60b in debt in the last 6 years, with $20 accumulated just over the last year. They have to deal with converting their fleet to electric over the next decade. Very expensive. Lots of risk. GM needs to pick and choose which risks are worth taking.

GM raised $3B in debt last year at nominal rates of 4%-5%.[1] In real, terms, that's maybe 2-3%. GM can raise billions from banks if it wants to. The fact that it would rather raise money by selling Cruise equity rather than selling bonds indicates that (a) they value Cruise at less than $15B or (b) they are risk-averse or (c) they see some other benefit from bringing equity partners on-board. It's not (d) they have no other good way of raising capital. GM is a huge company with a relatively low cost of capital (~5%).


"Bringing Robotaxis to market is very capital intensive". With the front and center approaches rooted in the DARPA Grand Challenges yes, with much of that money being spent on propping up simplifying assumptions that break at the tail end and on technologies that will be obsolete in a couple years. HD mapping curated by humans, human annotation for training/testing data, Lidars, and all that superfluous real world testing, will soon be replaced by approaches that are orders of magnitude more capital efficient. The less efficient approaches that currently dominate will sputter at 1% market penetration at best.

Why does GM need to put up their own money when Japan is falling over over itself to fund their R&D program?

Right? Letting your customer fund your R&D in order to better engineer the product that you and them will both use seems like a complete no-brainer.

The deal was $750 million cash and $2 billion in resources (whatever that means). Looks like it was more to get resource commitments for the launch... And cash because why not.

I live in a neighborhood that has Cruise vehicles frequently driving the streets. It's anecdotal, but I've noticed a fairly dramatic decline in the quality of on-road interactions. For example, having a lot of trouble navigating around Lyft/Uber cars pulled over to the side of the road or cyclists in the bike lane.

It's interesting to me to see their valuation be so sky high since I feel like it's not manifesting in the streets.

Sounds good to me. I’ve heard VERY good things about Cruise. And I like their approach much better than Tesla’s.

How is their approach different from Tesla's?

They use 3D sensors (lidar) and dense, 3D maps and scans of the operational area. Also they plan to own the vehicles and operate in a designated service area rather than sell a vehicle that works everywhere.

Can somebody explain why GM would want to provide competitive advantage to its competitor for a fixed price?

Either GM has licensing fees in its future, or the project is proving to be too capital intensive. Or could there be an effort towards a consortium?

Most of the major talent in the field is gainfully employed. I simply do not fully understand the objective of all of this capital raising. I look forward to better understanding the strategy in the rear view mirror.

I think one issue is that, like VR, one bad actor sours the consumer appetite for the entire concept. I think a safer road for GM is to become the de facto A+ supplier of tech that works well than for 2/3 of the legacy industry to have janky half-working sorta-dangerous solutions in 2025.

GM has provided engines for other manufacturers for years, so is it really that different?

Honda cars and gm trucks

It's interesting that American automakers are ahead of the Japanese automakers in the areas of self-driving and electric cars. Not really what I would have expected if you had asked me 10 years ago.

10 years ago, or "that time I pretend was just 10 years ago"? I confuse those two all the time.

Actual ten years ago, in 2008, Japan had long lost the halo of being the dominant economic force of the future it had through most of the 1980ies and 1990ies. In 2008, Tesla was already selling the Roadster, the iPhone wasn't new anymore, the The Matrix sequels were already older than the original movie had been at the time of their release. Another twenty years earlier, we all watched Bruce Willis chase German terrorists through Nakatomi Plaza, that's when Japan was "The Future".

In 2008 the iPhone was only a year old, Japan was still the world's number two economy, and two of the big 3 American car companies was bankrupt by the end of 2009 and got bailed out by the government.

So you were referring more to a particularly weak moment of American car makers than to the strength of their Japanese counterparts at the time, makes sense that way.

Eh, seems plausible to me. The Roadster was just a cool toy. Toyota's Prius had become a symbol of the future of cars (not necessarily hybrid, but low consumption) while GM was preparing to file for Chapter 11 and Ford had reported the largest annual loss in company history just a couple of years before.

Well, to me this is not really surprising since America is leading the way technologically speaking. Computers, the internet, the smartphone, Google, Apple, Microsoft, Amazon, Facebook, etc.

What is a self-driving car, really? Software on wheel. Who's leading the pack in software development? America.

For every one of those companies you mention there are Chinese ones that are arguably bigger and more successful.

And I don’t think you can chalk smartphones down to an American invention really.

What, like Huawei? I don't think you can call the companies (I'm making assumptions about which ones you're alluding to) more successful when their success is predicated upon corporate espionage and IP theft from the aforementioned American companies.

People are browsing in browsers created by american companies, compiled with compilers created by american companies, running on Operating Systems by american companies. The easy surface problems (making websites and apps) has indeed started to be tackled by the rest of the world, but the US is clearly the birthplace of all the computing stack. It's not surprising they are the first to tackle self driving cars.

These companies are successful inside China but if actual free market competition was allowed in the communist country, they would have hard time competing with US counterparts imho.

And these companies are?

Can you provide their names? I’m curious if I’ve heard of them.

Was it not the DARPA challenge in the USA that honed the skills of urmson, CMU, Stanford, and other engineers? What role did Japan have to play in FSD vehicles?

Why? We lead the world in software by a long margin. It's one of the reasons english is the lingua franca of software development. As for electric cars, it's only the US and the chinese who are heavily invested in it. The japanese have a presence in EVs but they are more heavily invested in hydrogen cars.

European manufacturers are heavily investing in electric cars. You're perhaps blinded by local news -- then again I'm surely too, as I don't know of US companies other than Tesla that are doing something with electric.

GM has the Bolt and Volt, and a Cadillac hybrid. Ford has the C-Max plug-in and a few other hybrids, and is going to build a hybrid F-150 and Mustang in the next few years. And for some reason Chrysler managed to beat Toyota and Honda to market with the first hybrid minivan.

Given that the F-150 is the bestselling vehicle in the United States, Ford might become one of the world's top sellers of hybrid cars when they release it. And the hybrid Mustang is supposed to be one of the fastest Mustang models.

>for some reason Chrysler managed to beat Toyota and Honda to market with the first hybrid minivan

Chrysler pretty much invented the minivan and is actually very competitive in the minivan market. Here's[0] a review of Chrysler vs Honda & Toyota.


I'm skeptical that a hybrid F150 would be much of a success, outside of people that probably don't need to have a full-size truck anyway.

It might be mildly interesting if any batteries were stashed in the rear - in the winter I have to put about 1000 pounds of sandbags to make it safe to drive on snow and ice.

The hybrid F150 will sell well based on one feature alone: using the vehicle’s engine and electric motors to function as a powerful worksite generator. Of course most people won’t use it for that but it gives them the excuse to buy it.

That and fleet sales are huge for f150 iirc. A more efficient, even if more boring to drive option exists, they will procure it en masse.

GM has been shipping the Chevy Bolt for a while though sales are down. Of course there are lots of hybrids.

The Nissan Leaf seems more than "a presence".

In what way other than wages does the US lead the world at all, never mind "by a large margin"?

off the top of my head, the three most valuable tech companies in the world are based in the US.

who do you think is the world leader in tech?

Toyota production system heavily favors autonomation vs automation. That combined with incremental process improvements and heavy bureaucracy has been a huge impediment

Theee is a reason why TPS is the current gold standard in production, regardlesz of industry. Automation is fine in theory, but no silber bullet solution. Large hardware manufacturing is something different than small electronics.

If it weren't for the threat of Google and Uber American car makers would still probably be uh... asleep at the wheel :)

Automakers use each others parts all the time. Dodge used Mitsubishi engines in the K cars and early minivans. Rolls Royce use Mercedes engines today. Nissan used Toyota's hybrid transmission in early-2010s Altimas.

This is a good sign for Criuse being a competitor in the self-driving space, and a sign of self-driving market health in general. Much better than Toyota investing $500 billion in Uber for self driving.[∆]


Rolls Royce is owned by BMW, they definitely don't use Mercedes engines today

Perhaps they are thinking of Aston Martin, who have fairly recently entered into an engine deal with Mercedes.

The Honda Passport and Acura SLX were both built by Isuzu, which is partially owned by GM.

Only the 1st gen of each iirc, needed for a head start in the blossoming SUV market at the time. A smart move, if a bit risky based on their relative QC at the time. And you left out the 1st gen Odyssey, also an Isuzu rebadge. You can tell them apart easy, as they are 4-door rather than sliding door models.

I think you mean Toyota invested $500 million.

Honestly this makes sense, and I hope the rest of the industry follows suit. I'd rather all car companies be working towards one, strong implementation of self-driving, than for every individual company to work on their own.

It would also mean liability could be diffused to the industry rather than companies.

So GM bought Cruise for $500MM in 2016 and Honda just invested in Cruise, valuing Cruse at $14.6B. Therefore the value of GM should be an extra $14B, and yet GM's stock has been flat since the purchase. What's going on here?

It could be that the stock market doesn't agree with that private valuation. Also, that valuation is a little odd because some of the investment doesn't kick in until they launch a product, and in the case of Honda, some of it is in machines and other non-monetary things.

Or it could be that as Cruise's value goes up, GMs goes down. Look at what was Yahoo for an example -- for a while the stock was worth less than just the value of their Alibaba shares, meaning the stock market attributed a negative value to what was left of Yahoo.

> Look at what was Yahoo for an example -- for a while the stock was worth less than just the value of their Alibaba shares, meaning the stock market attributed a negative value to what was left of Yahoo.

That is not what the situation meant.

Yahoo was trading at less than their book value per share. The vast majority of their book value was the share of Alibaba the company owned. This asset was not generating revenue for Yahoo, nor was it a liquid investment that Yahoo could convert into cash whenever it wanted to. There may have also been other conditions on sale of the stake that meant, in at last some circumstances, the liquidation value of the stake was significantly less than the paper value on the books.

Therefore, it gets discounted by the market. Since it comprised such a large portion of Yahoo's book value, the size of the discount was greater than the value of Yahoo's operations, leading to a price/book ratio less than 1. Yahoo's other operations still had positive value.

The latter seems unlikely given the stock of Ford, Toyota and other related companies have mostly followed the same peaks and troughs of GM (for quick comparison, look at GM vs CARZ etf). Unless GM has managed to bleed the exact value of Cruise over that period.

How does any autonomous driving startup compete against that pile of cash?

I don't think that you can do autonomous driving without a pile of cash. The problem is too complex.

It’s early days. If you have a better idea or can make your tech significantly more reliable/cheaper you could still jump out front of the pack.

But it’s not going to be easy. Lots of companies are working hard so the chance of a come-from-nowhere great idea seems lower and lower. If it happens I’d assume it comes from someone with experience in the domain... i.e. an existing player with a new angle.

Autonomous driving is a very capital heavy sector. You need real-world driving to get data and to test the software which requires capital for vehicles, equipment and salaries for people to be in the car. Uber's self-driving division is costing the company $200mm per quarter.

You figure out something you can sell them.

Lol, cash does not equal innovation. Comma.ai has only raised 8~ million and already has a product you can buy today on par ( or better ) with GM's supercruise.

Do you work for comma.ai? Because I can't imagine any other reason for the absurdity of your cheerleading. On par with Super Cruise? Gimme a break.

Lol no just a big fan and someone who ported one of their cars. Just most of the self driving car news lately seems scammy to me (by that I mean a lot of hype and no product), and I wish more people knew about the actual state of the industry.

Have you tried both systems? SuperCruise is really good at where it's able to drive, maybe better than openpilot but not by much. Openpilot is also getting really good and can be enabled anywhere, tho most of it's usage is still on highways.

This is interesting... GM bought Cruise for what, $2 billion? And now Honda buys a minuscule stake in Cruise company for the same price GM bought the whole thing for.

> And now Honda buys a minuscule stake in Cruise company for the same price GM bought the whole thing for.

No. The Honda investment is $750M. The rest of $2B commitment are for Honda to implement/integrate with Cruise's hardware/software technologies - not really payments to Cruise.

That is normal. A small amount of money buys a large stake in the early days. As the technology gets closer and closer to reality the amount it costs to buy goes up.

However don't read the above as invest in the early days - there are a lot of seemingly great ideas that will turn out to fail for some reason. Honda's bet is much less risky because it seems much more likely that cruise will have something to sell in the future. GM's bet was more risky, and [I forget who they bought cruise from] made an even more risky bet. This time it looks like it will pay off, but not all do.

$500 million cash and another $500 mill in performance bonus, contract buyouts etc.

I figure that by the time this self driving tech makes it to Honda's vehicles, few people will be buying them, because Honda has failed to get on the EV bandwagon. Honda and Mazda may be victims of the surge of EVs soon to come from China, Europe and Tesla.

Other Asian carmakers such as Toyota and Hyundai are doing much better getting onboard the EV/Hybrid train though.

Honda is doing fine, and was early into the hybrid market. Whatever are you on about?

The only missteps was their zev hydrogen tech which doesn't make engineering sense (to me), and a black eye from bad v6 transmissions for way too long in the 90s thru 00s, and a departure from small, low-displacement, high efficiency, well handling cars with sla suspension for pigfat appliances with cheap mcstrut suspension and more wings.

Lucky for them, the world is a bit pigfat and risk adverse themselves, and still buy CRVs like crazy. While the 88-01 civic, and s2000 still somehow support their reputation in enthusiast circles thanks to those now purely nostalgic qualities.

But I digress. I think ICE cars will continue to serve a purpose. Mazda just announced a new rotary ICE hybrid engine design this week! If self-driving tech takes off, it will be applied just as easily to either tech. I see Honda at no major positional disadvantage here.

MacPherson struts are fine. Even the most ardent enthusiast has difficulty telling the difference between a MP setup and a double wishbone.

BMW uses them and Porsche uses them on their GT# 911s. You'd think if there were a case to be made for the performance demerits of a MP setup, Porsche wouldn't be using them on their halo cars (or their class-winning race cars).

I'm on about solely electric vehicles: EVs. Honda is more interested in Hydrogen than EVs. They have some underpowered hybrids, but few, overpriced models on the global market.

Lots of EVs are on their way and Honda and Mazda aren't announcing any future plans to enter that market.

Initial cars will have a reasonably high power consumption requirement. Vehicles with adequate voltage supply will probably have an advantage compared to less sophisticated energy distribution.

A couple of years back I made the same comment about the European car industry. I got wiser since then. The main lesson I drew from Tesla's struggles with manufacturing was 1) efficient mass manufacturing is making the difference between life and death for EVs too and 2) an EV is easier to build than an ICE powered car.

I don't see the early mover advantage, just a realistic risk for really late companies. And rigjt now it is not too late for anyone at a long stretch. It is important to remember thst the EV game now moved to car makers core competencies in mass production. There a stll a ton of challenges for them but none as existential as it seemed a couple of years back.

I thought GM bought Cruise. Is that not the case?

GM did buy Cruise. Then GM sold stakes to SoftBank and now Honda. Cruise is structured as an independent entity with, as of now, three owners. GM remains the majority owner.

IMMO, this is a good strategy from GM. The more big autos to adopt / invest in Cruise, the better.

Especially since it’s on track to being a commodity technology.

What does these kind of investment mean for a company like Cruise? Does it provide them with more financial liquidity or does this only fatten up GM's numbers since Cruise was an investment after all? I've been stalking their NYC career opportunities but nothing yet even after the Softbank investment :/

I saw an interview with the GM president Dan Annman (sp?). $750mill in cash... Who doesn't like cash?

$2 billion in other resources. Honda has a good European and Japanese market to sell cars.

He said timetable for initial launch is same... 2019, all depends on safety. But the investment is more for after that.

Makes sense since Honda went in with GM on On-Star. I guess they view it as needed, but better as a partnership with a company they have dealt with before.

Anyone know whether it's too late to join Cruise?

The great boondoggle continues

Masayoshi Son strikes again.

2030 for full autonomous driving? That is much later than all the hype. So is Honda being pragmatic, and most everyone else is blowing smoke? Or has self driving had a big setback in the last year?

In my experience, American companies tend to be overly optimistic, and Japanese companies tend to be overly cautious about timelines.

See uncertainty avoidance and long-term orientation: https://www.hofstede-insights.com/country-comparison/japan,t...

Waymo is about 5 years ahead of everyone else, so I think it's likely a combination of the two.


Could you please stop with the repetitive pro-Tesla/Musk comments? They aren't teaching us much and that's why we're here.

I didn't know there is a rule against posting something factual that happens to be pro-tesla. If there is anything inaccurate about my post let me know.

First, Tesla is not the tech leader in autonomous driving. So that is factually wrong. Second, whether this tech has to be kept in-house or not is, without a detailed view of the particular situation (which is different for every car maker out there) just an opinion. We are talking about an industry where outsourcing has proven to be a viable business strategy.

Perhaps you should finish reading the rest of that sentence..tesla is the most advanced "ON THE MARKET". People keep comparing it to non-existent market competitors like waymo. Tesla has close to 1 million cars on the market. Waymo has zero. It's really easy to compare your power point presentation to a company with a giant fleet out in the wild and claim you have the best tech... And as to your second point, yes the fact that companies should bring their tech in house is my own opinion. What's your point?

You asked for reason why you are down voted, I named two possible ones. For my first point, there are studies out there other people cited already that were done by people much more qualified to judge such things than I am. And for the second one, opinion is fine, most of the things I post are, in a sense, an opionion too. It does help to explain the reasoning behind that opinion so others can learn from it. Just saying.

EDIT: Typos

According to Navigant, Tesla is in last place. https://www.theverge.com/2018/1/16/16893452/detroit-auto-sho...

Can you buy any other self driving systems? As far as I know there are only 3 self driving products. Tesla Autopilot, GM Supercruise, and Comma.ai Openpilot. Hard to be in last place vs a product that doesn't yet exist.

Mercedes Drive Pilot, though is in its infancy.


Navigant study is a complete joke I don't know why people keep citing it. They look at factors "beyond autonomous driving"...to see who's leading autonomous driving. They look at things like who has the most "potential" and "market strategy", etc...things that have nothing to do with the actual tech. They basically go off of whoever has the best power point presentation. The fact that Tesla is last on the list should make it really obvious that the study shouldn't be taken seriously but people keep citing this

But selling 80k vehicles a month with autopilot hardware onboard. Who will reach the finish line first? That's the question.

It’s the most widely deployed, I think that’s quite fair.

But that’s VERY different from ‘most advanced’.

While I agree with the sentiment, to be fair to the original poster, they did say "on the market," which I take to mean "available for sale to members of the general public," which most more-advanced entrants like Waymo can't claim.

Exactly. Tesla has close to 1 million cars on the road. Waymo has power point presentations.

True. But the market is so small (IS anyone else selling something besides GM Super Cruise) that’s not necessarily saying much.

I get the impression that Tesla says their hardware is more capable than it actually is and they ok doing things that would give other companies pause for concern with regards to safety.

Cadillac Super Cruise seems to be just as capable as Tesla's Autopilot but kept on a much tighter leash because GM is more tempered and pragmatic about it.

I have aspirations to own a Model 3 eventually but I have absolutely no interest in Autopilot or any Lvl 3 autonomy and will not purchase a vehicle with it. You should either be 100% engaged or 100% disengaged from the driving process.

Just to be clear, SuperCruise and Cruise are two different entities. Supercruise is an ADAS system that competes with Autopilot, and is developed by GM's in-house autonomy team. GM Cruise (formerly Cruise Automation) is a subsidiary acquired by GM in early 2016 and is focused on fully autonomous robotaxis.

Also, to be clear, despite what Elon Musk may have told you, Autopilot has no chance of ever being fully autonomous on the current hardware, and it never did.

"Also, to be clear, despite what Elon Musk may have told you, Autopilot has no chance of ever being fully autonomous on the current hardware, and it never did."

Why is that?

Tesla lacks LIDAR which most people agree is required for full autonomy. Maybe Musk will disprove everyone. Waymo, Cruise and everyone else has incorporated LIDAR.

The computer Tesla is using isn't underpowered by a factor of 40.

I remember Mercedes and particularly Volvo started on self-driving very early. In addition, Google has probably passed 10 million self-driving miles.

What makes you think Tesla is the most advanced, rather than simply the most willing to toss what they have out into the public?

as did Fiat back in the 90s if I remeber well. Clumsy so with hardware taking much of the cars space except the two front seats but still. The only issue they had, anesdote, was the use of black and white cameras that had issues in construction areas identifying the yellow lanes to follow. No idea they ever tried it outside of highways, so.

You could do both. Which is what Tesla did. They licensed MobilEye for AP1 while developing their own AP2 which was only deployed once they had a falling out with MobilEye (and IIRC the initial rollout of AP2 was a bit rocky, at the time MobilEye tech was still superior)

Tesla fanboys are some of the most delusional people.

With cars that work.

This is what I'm talking about when I say they are delusional. Tesla's flagship vehicle, the Model S, is frequently labeled as the least reliable car of their division. The X is similarly shoddy. We shall see about the 3.

The only brand besides Uber that has killed anyone in autopilot is Tesla and they've killed three people.

https://www.businessinsider.com/survey-uk-drivers-say-tesla-... https://www.cnet.com/roadshow/news/consumer-reports-survey-m...

From what I have seen it seems like Waymo is first followed by Cruise. Tesla has had some pretty spectacular failures resulting in the deaths of 2 drivers.


Humans have had some pretty spectacular failures resulting in 1.3 million deaths in 2017 alone.

Why doesn't anyone do a deal with comma.ai? They actually have a product that kind of works with just handful of people.

Edit: Why all the down votes. Go check out some of the videos. It works pretty well. Its a level 2 system so the user still needs to pay attention to the roads which is true with any lkas system out now days.

They have built a single, forward looking, camera system on an Android platform and are using python for the control system. No serious auto manufacturer would trust a system like that for vehicle safety.

It's a less reliable version of the lane keeping systems that the automotive manufacturers already have and nothing close to the autonomous system that George Hotz was promising would be easy a few years back.

I also doubt automotive manufacturers want to deal with the drama around comma.ai and its publicity given they have a reputation to maintain as serious, safety conscious, organizations.

For all current LKAS systems you need to pay attention to the road including the Tesla. The LKAS most manufactures use aren't that good.

Comma.ai has the same constraints. No monocular camera system works without careful user supervision. Even lidar based systems are not that reliable yet that you could trust them to the point of not paying attention.

As far as quality goes it's way more reliable. Most of the "advanced" driver assistance systems would kill you within a minute if you left the car to their control.

What they have now is a Devkit that is level 2, which they are using to iterate on, gather data, gather feedback, and will ultimately ship a much more complete consumer friendly product.

It's really the only option for hobbyists and enthusiasts to get self driving capabilities today without 50k+.

> As far as quality goes it's way more reliable. Most of the "advanced" driver assistance systems would kill you within a minute if you left the car to their control.

Do you have any evidence/studies for either of these statements?

> It's really the only option for hobbyists and enthusiasts to get self driving capabilities today without 50k+.

Right. It's not a serious automotive grade safety system.

Sure. Go on a highway with your stock LKAS/ACC on your Toyota/Honda/Mercedes/mostSystems and don't touch the wheel or gas/brakes, and see how long before you feel the need to intervene.

Now do the same with Comma's Openpilot.

rjeli 6 months ago [flagged]

lol, don’t bother. HNers have no idea of the sorry state of the “””industry”””

Maybe not, but please don't post unsubstantive disses here. That only makes this place worse.

Instead, if you know more, share some of what you know so we can all learn something.


I work in the industry and am scptical of even the big players ability to deliver on the promises they made 3-5 years ago including waymo and cruise.

yep, I agree.

Right now everyone has a proof of concept demo and people are pouring money in. It's going to take a while and I really can't see a company like comma.ai being a serious player given their approach but maybe I'm wrong. Good luck to them, I hope nobody gets hurt.

You want "kind of works" with a self driving system?

Tesla's Autopilot and GM's Supercruise only "kind of works"

I don't know much about GM's Supercruise, but I'm still in the camp that Tesla's Autopilot is dangerous and irresponsible.

More specifically, I'm against anything less than L4 autonomous self driving cars.

Have you considered people using L2 systems as like an advanced cruise control?

For example, if people are going to text and drive, would you rather they have an intelligent L2 system that reliably keeps you in the lane and not crashing into the car in front of you, or just do as they are doing now?

I think there is much value in considering these systems as a better safety feature rather than fully autonomous.

It's not just about safety, but I think that the possibility of not paying attention when you need to is higher with an L2 system than standard cruise control.

Beyond safety, I think L2 accidents (which are already happening) will be VERY bad for the perception of autonomous driving in general, and will make it VERY hard on a regulatory level to get L4+ autonomous cars out on the road when they're finally ready.

Anecdotal, but still: I used to drive an Audi with a ton of advanced L2 systems (adaptive cruise control + lane assist). Both systems work from standstill to ~180km/h (well above highway speeds). While not perfect, I did use them both all the time, and the way you drive changes. You learn 'how' the systems work really fast.

The thing that changed for me while driving with both systems engaged is that my role changed from driving and continuously making small adjustments to stay centered in the lane and keeping the right distance. With the systems engaged, I was mostly monitoring that they work accurately. I don't have to bother continuously with most things, I can spend more time watching all surroundings to watch for anomalies / things the systems can't handle.

The end result for me at least felt a lot safer, and more relaxed. The systems felt very, very reliable.

That's because you were using them the right way. But a lot of people engage those systems and then pick up their phone and start texting or watching videos. Look at the guy who literally lost his head in his Tesla when it went under a truck. Records show that he was watching YouTube at the time.

The question is whether or not cars with L2 systems have a lower accident rate than cars without them. As with L4 systems, they don't need to be perfect - just better than we are.

That would be a good number to start to refute my safety argument, but it does no good with my non-safety argument.

Are you insane? It's literally running on an android smartphone...

Uber’s system “kinda worked” too...

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact