Hacker News Reader

Comment by keeptrying | Hacker News Reader

Waymo will have the first self driving cars on the road.

They are so ahead of everyone.

I think everyone including regular car companies buy the tech from Waymo in the short term.

Waymo is going to win because of (1) better sensors and (2) better onboard computing power. Other companies can potentially compete in the former domain but literally no one in the world can go toe-to-toe with Waymo and siblings in the latter department. The ML powering Level 4 driving decisions is not cheap in dollars or power consumption and Google is already way ahead of the curve when it comes to custom fabrication of those type of systems.

The hardware will become commodity. The software is where the action will be.

Imagine what would happen with fighter jets if they only had 3 main sensors. Lidar, Radar and camera.

The thing is: in the long run, yes; in the short run, Waymo is manufacturing its own chipsets that it has a patent on. It's not using COTS kit for on-board processing, which gives it a significant technical advantage that will be hard for competitors to match.

Elon maintains that with just cameras you can achieve super-human (10x) performance with self-driving cars. Fact I nabbed when he spoke about Boring at TED. https://www.ted.com/talks/elon_musk_the_future_we_re_buildin...

I wouldn't ever bet against Musk, but Musk says a lot of things. Yes we should be able to make self driving cars with just cameras because, after all, we do it with just two eyes. But in what world is computer vision perfect? Tesla's Autopilot already has a fatality on its record because it couldn't differentiate the color of the sky with the color of a truck.

LIDAR solves a lot of problems.

There's another aspect of it though. As soon as Waymo makes self-driving cars more popular, and there are less humans on the road - you just need inferior technology to be able to operate. Bootstrapping this is the hardest - after self-driving cars are a thing, normally we'll see accelerated move towards fully automated driving.

In the extreme case, if Waymo cars are 80%, Tesla cars can easily be 20% just based on the fact that Waymo cars will avoid most accidents made by Tesla software errors

You might be right, but I'll devil's advocate a bit, with a few possible predictions (none of which I believe in all THAT strongly, just speculation)

- If there's a period where new self driving cars are popular but aftermarket self driving kits for older cars are unaffordable or nonexistant, then during this period, the price of non-self-driving cars will plummet dramatically. Who knows, you might be able to buy a 2 year old camry for $1000. The market for these cars will exclusively be (other than oddballs/enthusiasts) people whose wages are so low that driving yourself in a non-self-driving car is still cheaper in an overall sense than using self-driving cabs, so that will cap the possible value of such cars at a pretty low point.

- As long as (possibly as a result of the above) the number of human driven cars on the road is at all significant (like, at least 1% or 0.1% or something), the benefits you mention will be hard to attain. In a given trip a car interacts with hundreds of other cars.

- You forgot about pedestrians and cyclists. In many places, there are more of those than there are cars, anyway, so even if all cars are self driving,

- There's going to be weather and construction and falling trees and runaway skateboards and runaway trucks on ice covered hills and wildfires and so on. This kind of software will operate at scale, so improbable events become important to handle.

- Uber is probably dumb enough to hope your argument is true and think they can get away with half-assed self-driving, which is why they're going to fail miserably and go bankrupt.

- There's going to be regulation. As a made up example, if the "Waymo patent tax" is "just" $5000 per car, it's going to be pretty hard for a car manufacturer to justify to regulators that they'll instead develop competing software that is somewhat inferior in safety, but "acceptable", for the sake of saving, say, half of that $5000 (it still costs the thrifty manufacturer money to develop or buy their lower quality software!). All it'll take is one fatality and people will be up in arms. If the "acceptable" technology was the first one out, it'd be different, but it'll be hard to convince anyone it's worth the risk when there's better tech already out there. BTW, Waymo management would obviously be clever about setting a good price point that captures lots of value but still is low enough to discourage this kind of competition.

- My vague software and driving intuitions tell me it'd actually be easier to design a more general system anyway (massive handwaving). It's actually HUMANS (specifically those accustomed to easy suburban western driving conditions) who have a harder time driving when they can't assume everyone will follow the rules, not computers, and I think that's why so many people mistakenly (IMHO) imagine it'd be so easy to make self driving cars if only everyone followed the damned rules. The simplified laws of physics applicable to driving are simpler than the DMV manual anyway.

>Tesla's Autopilot already has a fatality on its record because it couldn't differentiate the color of the sky with the color of a truck.

Do you think in your wildest dream that this exact scenario would ever be allowed to happen again by Tesla engineers? It may have crashed causing a fatality in this instance but it doesn't mean the prospect of camera only vision systems need to be written off. What people often overlook is that these systems get stronger with more and more edge cases and objectively the Tesla team are doing outstanding work with the sensor package that they have.

> But in what world is computer vision perfect?

I'd guess the same world where human vision is perfect...

Sure, but what is the reduction in accidents for just cameras vs all three? Super-human might only be reglatorial approved in the absence of technology being available with high safety.

No doubt turning many car companies into the new Android phone developers.

I'm curious where Tesla fits into this scheme. In terms of road miles, I reckon Tesla would have Waymo beat easily. They have far too many cars on the road and thus, way more data.

Does this factor into the quality of self-driving tech at all?

It's not that simple though. What so you consider "data" when it comes to self driving cars? Waymo collects high resolution data with every car (multiple GB or possibly TB of data per car per day).

Tesla hasn't sent very high resolution data back home yet. Though they are looking to get video from customers now to help make this go faster[0].

So while Tesla has more vehicles on the road, that doesn't mean they are getting better data for self driving car research.

[0] https://electrek.co/2017/05/06/tesla-data-sharing-policy-col...

They may not be sending data back home via the mobile network, but you can bet your boots they make exhaustive logs of Autopilot disengagements and store them locally to be siphoned off at the next service.

I wonder if they have WiFi base stations collecting data at Supercharger points? That's what I'd do.

> exhaustive logs of Autopilot disengagements

That would still be only one slice of the many many situations you need to master, and that Waymo already has. E.g. something as simple as entering and exiting a small roundabout, like that YouTube video showed a Tesla couldn't do. Miles driven by Tesla owners aren't purposefully looking for those situations when they engage autopilot.

Are you sure? Again this is just how I'd do things, but after I had a bunch of disengagement snapshots, the next thing I'd do is start taking snapshots of similar situations. We live in an age of easy and super-accurate image classification, it wouldn't be too hard to start a snapshot every time the front camera view contains a roundabout.

What I'm trying to say is that the number of cars on the road and miles driven aren't what matter to their autopilot learning, when compared to other self-driving companies. Most people use their car to go places comfortably, not to look out for learning experiences for Tesla.

With how long commutes are in the US, a very very small fraction of miles driven by people on autopilot will be other than "the car was already on the highway, autopilot changed lanes a few times, and after a while was manually disengaged to take an exit".

That small fraction is what could be compared with the miles and number of cars driven by the other companies, because in the latter case it's hired drivers actively looking for conditions of interest to the engineers.

Why not use owner's home network?

I definitely wouldn't be happy about Tesla uploading gigabytes of data on my metered connection. Even if it was unmetered, most people don't have very fast up-speeds and that could take days to upload and heavily impact their speeds. Absolutely unacceptable.

What if they had agreement with xfinity/comcast, which opens up a separate wifi hotspot anyway?

Do owners generally connect their cars to their home wifi? (I don't know, is this a thing with Teslas?)

Even if so, uploading gigabytes of data every day is going to be noticed. While I'm sure they have the legal right to do so (via their TOS / EULA) they probably still don't want to emphasize that they're gathering that much data (unless you opt in to their fleet learning stuff).

>Do owners generally connect their cars to their home wifi?

Generally I think they do to access the Tesla App features. Tesla also have a network connection so they may already be uploading the information anyway over cell networks.

Tesla's don't connect to LTE networks by default?

If not, I find it shortsighted and rather cheap that each car doesn't have it's own $50/month data plan (or whatever) to phone home important data.

Teslas do connect to LTE. Why would they pay that much for a data plan? In the US, you can get mobile hotspot plans for $3/month + $10/gbyte. That's the consumer price; wholesale over 100s of thousands of cars has got to be cheaper.

At 80k new cars a year that's a 4mil a month waste of money. Tesla couldn't survive giving each car a 50/mo data plan forever.

Waste of money?

To have a constant high speed IoT-like connection to the web for real-time updating and such?

I am sure some sort of volume discounts could be negotiated as well.

The logs might be pretty big, so it might have a significant impact on people on metered internet connections.

Not if it's only greyscale camera data and some point cloud data which is heavily compressible.

But then you're back to square one : that data is much less detailed than what waymo gets.

Machine vision is usually an exercise in data reduction. Greyscale is typically used in most computer vision applications, so it may be the case that Waymo has more information but that is not necessarily of any benefit.

The video has constant motion though, which makes compression much less effective.

As of 2013 Google's self-driving cars collected about 1 GB per second of data [1], today it must be even more. There is no way Telsa can upload nearly as much data from customer's cars. I also belief that their team is smaller and they started much later.

[1] http://www.kurzweilai.net/googles-self-driving-car-gathers-n...

You shouldn't need to upload the entire 1gb. The central AI can query the cars to get just the bits it has questions about. Part of the AI's job is learning which data to care about and which data to ignore.

The fact that the trees got new leaves again this year and it's dark at 2am and the sign at the Chinese Restaurant is still red don't to be relayed.

More data isn't automatically better. The hard part isn't getting some magical quantity of data the hard part is asking the right questions of it.

A self driving system needs to solve this problem (the frame problem) anyway. So you get intelligence compression for "free".

In a sense, intelligence is compression.

This is naive. You don't know what data you will care about in a year. The ability to have and keep all the data is incredibly valuable.

Why do so many on HN feel the need to structure their responses like this:

> Something Rude. Real argument, which is stronger with first sentence removed.

You may be right (or wrong, I don't know) but opening with rudeness or insults makes it more difficult to have meaningful conversations.

Rudeness? Insults? I said "this is naive", not "you are naive."

Surely we're all adults and can talk about ideas candidly without having to beat around the bush and tell each other little white lies.

A while ago you told someone, "You speak in broad generalizations that miss many points and make many faulty assumptions. "

Any difference between your statement and mine comes down to how individuals respond to specific words. How about we rise above such pettiness and focus on the subject matter?

You've added nothing to this discussion and have wasted people's time.

> You've added nothing to this discussion and have wasted people's time.

What is your response doing?

Historic data is certainly usable. I'm not sure where your idea comes from that it often provides value far beyond what's in the short term cache.

What makes you believe that?

How many things happened today that won't happen next week, somewhere?

And please don't call me naive. That's rude. Disagree substantively; use your words.

You have essentially just described distributed learning. Only transmit the gradients back to the central server, not the data.

That'd be one way of doing it. You could also just send the cars experimental queries to upload, and still do the leaning centrally.

Wow. I reckon this must be even more now.

Google is also a better computing company than Tesla, I imagine, whose strength is hardware, not software.

Very first sentence of your source says "Google’s self-driving car gathers 750 megabytes of sensor data per SECOND!", here goes one fourth of the alleged 1GB.

Then this is about sensor data or raw data, but there is thing known as compression which is quite common and can shrink size by an order of magnitude.

75MB/s seems closer to reality and manageable.

Last week Tesla began collecting short sections of video data from those with their new in-house Autopilot hardware. So really, Tesla's just getting started, and they're doing it with underpowered sensors and compute.

https://www.theverge.com/2017/5/8/15581094/tesla-autopilot-u...

Waymo was ahead of where Tesla is now back in 2012. In a demonstration capacity, Waymo's cars were working 5 years ago, but in the intermittent time Waymo has been validating their software against every possible edge case they might encounter on a real world situation, and Tesla will have to do the same one way or another before their autonomous OS can operate safely without a babysitter.

It is a matter of point of view, how many waymo car have been sold to customers ? how many waymo cars are out of the waymo car factories every months ?

There are several things to tackle and different approaches to things, the DARPA grand challenge is the practical one and waymo did not took part in any edition AFAIK.

From Waymo's wiki page: "Google's self driving car project was formerly led by Sebastian Thrun (...). Thrun's team at Stanford created the robotic vehicle Stanley which won the 2005 DARPA Grand Challenge and its US$2 million prize from the United States Department of Defense. The team developing the system consisted of 15 engineers working for Google, including Chris Urmson, Mike Montemerlo, and Anthony Levandowski who had worked on the DARPA Grand and Urban Challenges."

How many people have died in Waymo cars?

The miles aren't equal, though. Waymo's cars have an amazing array of sensors on them, including 360 degree LiDAR, which Teslas don't have.

teslas don't have lidar, so it's hard to say how useful their data will be in the long term

True, but if it turns out LIDAR is overkill and that self-driving can be done with only cheaper sensors (cameras), Waymo might be behind the curve on the use-case that the market will opt for.

When it comes to your safety, would you rather overkill or underkill? Especially when it comes to something as important as self-driving, I think it's absolutely reckless to try to "maximize profit" and "get it out there ASAP".

It would be an absolutely disaster if the technology as a whole gets a horrible reputation, or even worse, banned in places, just because some company skipped some steps and wanted to be the first or cheapest to the market.

I'm sure that Google gathers so much data from human drivers using Google Maps and Waze that they can build a decent model.

Why do you think Google maps data would be as useful as the camera and steering input data Tesla can collect?

Why do you think Tesla is able to collect all or any of the camera and steering input data? And what proof do you have that the little they are able to get off is actually at all valuable.

Ok, fine, we don't know what either company is collecting, but google maps can ONLY collect GPS and speed data, whereas Tesla has the POTENTIAL to collect gps, speed, camera, steering input, braking input, and acceleration input.

Yes and no. In the bay area and other big cities Teslas are definitely there but there are a lot of other areas in the country where Waymo probably spent the money to drive and collect data. So we probably will not know the reality.

Volvo / zenuity are also doing well. They'll probably have the first level 4 cars in public hands. It remains to see who can get to level 5. https://www.theverge.com/2016/4/27/11518826/volvo-tesla-auto...

Hmm, do they/you have any concrete numbers to back this up?

For example, Uber has a disengagement rate of once per 0.9 mile, whereas Waymo shows numbers higher up at 5000+ miles.

The California DMV report [0] has numbers for a couple other companies that tested in California. The other runners ups are Nissan and BMW with 247 and 638 miles respectively. Still an order of magnitude behind. But Volvo isn't there, and I'm curious to see concrete numbers. Because no amount of press and videos really means anything. Anyone can come up with a video of the car doing a nice ride. Google published one of those in 2009. It's spent the last 8 years perfectly that. That's the hard part, making it work in all cases, not just for one video.

[0] https://www.wired.com/2017/02/california-dmv-autonomous-car-...

I don't have any numbers but they say they are going to set them off driving members of the public this year. Have to see if they deliver.

It's hard to say how comparable those numbers really are. One manufacturer could be only driving routes that they know are the most challenging for their system, and another could be running laps on the freeway.

That's a good point, though it's still the closest to a quantitative number we have. I would love if the government actually forced some more rigorous quantitative testing on these cars, some way we can compare apples to apples. A lot of companies are making all sorts of crazy bold claims but I'd like more concrete evidence before I put my life in their hands.

I believe that Waymo are killing it, but... their desire to not be a car company and just write the software makes me wonder what their plan is to become a serious driver of Alphabet profit.

Unless they have made some real breakthroughs and are managing to somehow keep them under wraps (seems unlikely), optimistically they have a 5 year lead on their competitors.

5 years is a good amount of time to be the only people with self-driving tech, but unless they actually use it to muscle in on some other industry it seems like they will be reduced to just another component supplier when enough competitors eventually catch up.

They seem to have gotten a lot more serious since forming Waymo. In just a few months, they started the actual public trial in Phoenix as well as make this deal with Lyft. Seems like they've finally understood the urgency and have stepped on the pedal.

I don't understand your claim that they have to be a car company though. With this deal, it seems like they are (at least for now) targeting ride sharing, and arguably that's a much better use of the technology for many reasons.

For example, you don't have to worry about parking, you can put limits on where the car drives, you own the cars so you can track them, collect data and transfer it as you wish.

Also, wasn't Uber valued higher than Tesla, at least a few months back? So arguably as a business, it's as good if not better than being a car company, at least for now until the tech gets even better.

It does seem Waymo is going the path of the component supplier (at least right now). While they have the active partnership with Fiat Chrysler, there was a report[0] in December that they were in talks with Honda as well. Then from reddit pointed out a job posted for Waymo looking for people that speak Japanese[1].

[0] https://www.recode.net/2016/12/21/14046496/waymo-alphabet-ho...

[1] https://www.reddit.com/r/SelfDrivingCars/comments/67w27p/fia...

I don't think Waymo necessarily needs to build cars, they could try to be Uber, they could try to be FedEx, they could try to be Seamless. I don't really have a good opinion on what they should be, just that I don't think they will make enough money as a supplier to justify all the risk they've taken.

Maybe they can make some deals where they give a company exclusive access to self-driving tech for a given industry for a significant equity stake in that company, but I feel like no-one would really want to take that deal.

Partnering with others sounds like a much faster route to getting into the hands of consumers (and therefore getting money).

Having to also create and scale up production of a car as well sounds harder and much riskier.

I'm assuming you mean the first ones without human chaperones? Because I've seen plenty of self-driving Uber cars shuttling passengers around the streets of SF over the past year.

I thought Uber stopped that program after the red light thing? I assume they mean the same though because Google have had self driving cars in Mountain View for years - this article claims 2 million miles last October : https://www.wsj.com/articles/googles-self-driving-car-progra... I'm not sure how that stacks against competitors but I imagine they're a long way ahead.

Uber backed down, though I would say I've only seen them once and not "plenty".

http://fortune.com/2017/03/27/uber-self-driving-cars-san-fra...

They did, but they've resumed again recently.

I see them running every day in Tempe, Az.

Ever heard of the DARPA Grand challenge ? The first edition way in 2004 and the project aim is to have one third of US ground military vehicles autonomous by 2015. It seems the US DoD has a quite a headstart on waymo.

> the project aim is to have one third of US ground military vehicles autonomous by 2015

And was it successful?

> Waymo will have the first self driving cars on the road

Which are still way, way off in the future (if ever).

I propose 'self-driving cars' are the 'flying cars' of our generation: Something that seems inevitable and just around the corner, yet here we are.

There is an important difference. "Flying cars" was always a media and popular culture thing - the vast majority of experts thought they were a long way off, if ever. This is not true of self driving cars.

Self driving cars are also a pop culture thing - and the most aggressive deadlines and comprehensive deployments predictions are unlikely.

Let me take the contrarian side and say you will have small 1-2 seat Self-Flying vehicles before Self-Driving Cars.

Except flying cars would require an entirely new transportation infrastructure. Self-driving cars adapt to what's already in place.

5 years.

Reply on news.ycombinator.com