Autonomous Car Progress

[mspohr]

Interesting to see the output from the v9 neural net.
Here’s what Tesla Autopilot can see with the new v9 neural net

 

[cwerdna]

Why People Keep Rear-Ending Self-Driving Cars

 

[R.S]

What I think you meant to say was that if in your example you used 5% interest instead of 3%, the value would be $454k instead of 5k. And if it grew at 7% it would be $37M. Compound interest is a beast!

Right, that's what I meant. And it shows why people are so bad at estimating exponential growth. If something grows 3%, 5%, or 7% per time frame has extreme consequences on the outcome. But since growth isn't always continuous in real life, it's hard to predict the average rate of growth. So if you expect 5%, you can be extremely wrong in both directions, if it happens to be 3% or 7% instead.

 

[cwerdna]

GM's self-driving car reportedly has trouble recognizing pedestrians

 

[strangecosmos]

Reuters article on Cruise:

"Reuters has learned that the driverless Cruise cars still struggle to identify whether objects on the road are moving or stationary, according to one current and three former Cruise employees who have witnessed the problem. The result is that the vehicles hesitate and stop while passing a row of parked motorcycles or bicycles, they said.

At times, the software has failed to recognize pedestrians, and has mistakenly seen phantom bicycles, causing the cars to brake erratically, according to two of the sources. And Cruise does not yet have a data-sharing collaboration with the San Francisco Fire Department, a necessary step to train the cars to respond to fire truck sirens, according to a fire department spokesman."

 

[cwerdna]

California gives Waymo the state’s first permit to test self-driving cars without humans

 

[cwerdna]

GM faces lawsuit after crash between motorcyclist and self-driving Chevy Bolt

Not sure if the outcome was posted already (did a quick search and couldn't find it). Hope this isn't repeat.

GM Settles Suit with Motorcyclist Hit By Autonomous Vehicle in California
First GM Self-Driving Car Crash Suit Settles
GM Settles First-Known Suit Over Self-Driving Car Crash

 

[cwerdna]

Bloomberg: Who’s Winning the Self-Driving Car Race?
A scorecard breaking down everyone from Alphabet’s Waymo to Zoox

 

[cwerdna]

Watch As China's First Electric, Fully Self-Driving Taxis Hit the Road

Not sure if this should go here as it's not exactly progress...
Waymo blames collision with motorcycle on human error, not AI points to The very human challenge of safe driving – Waymo – Medium.

 

[Bladerskb]

Waymo Driverless Test Application

Waymo California Driverless Autonomous Vehicle Tester Program Application.pdf

Interesting point is their car is capable of up to 65 MPH or rather they will operate driverless test on roads with speed limit of up to 65 MPH (not necessarily meaning that 65 is its limit).

Description of their LVL 4 system

Waymo’s vehicles validated for driverless testing include a Level 4
automated driving system under SAE International’s Taxonomy and
Definitions for Terms Related to Driving Automation Systems for On-Road
Motor Vehicles, standard J3016 (SEP 2016). This system is what enables
the capabilities of our fully self-driving vehicles.
Our Level 4 system includes the software and hardware that, when
integrated into the vehicle, perform all driving functions.

Waymo’s self-driving system is designed to perform the entire dynamic driving task
within a defined operational design domain and has the capability to
achieve a minimal risk condition : the ability to bring a vehicle to a safe
stop, without any expectation that a human driver take over. In contrast,
systems at a lower-level of automation, at SAE Levels 1, 2, or 3, are required
to have a human driver take over from the system when necessary.

If the Waymo vehicle can no longer proceed on a planned trip, Waymo’s
vehicles are designed to be capable of performing a safe stop, known as
achieving a “minimal risk condition” without any need for human
intervention, which is a requirement for an SAE Level 4 automated driving
system. This includes situations when Waymo’s fully self-driving vehicle
experiences a problem that prevents the automated driving system from
continuing the driving task or when environmental conditions change in
a way that would affect safe driving within our operational design
domain. Waymo’s system is designed to detect each one of these
scenarios automatically. In addition, our vehicles run thousands of checks
on their systems every second, looking for faults. Our system is equipped
with a series of redundancies for critical systems, such as sensors,
computing, steering and braking. Our vehicle’s response varies with the
type of roadway on which a situation occurs, the current traffic
conditions, and the extent of the technology failure. Depending on these
factors, the system can determine an appropriate response to keep the
vehicle, its passengers, and other road users safe.

During driverless testing, the intended operational design domain of

Waymo’s vehicles will include the following roadway types:
● Freeways, highways, city streets, rural roads, and other roadways.7
● Parking lots

Driverless test map area

 

[malcolm]

Waymo Driverless Test Application

You left out the previous paragraph on page 18.

Waymo’s system is also designed so each vehicle will not operate outside of its approved operational design domain. For example, passengers cannot select a destination outside of our approved geography, and our software will not create a route that travels outside of a “geo-fenced” area, which has been mapped in detail.

So much like the early days of cellphones with initial coverage in main areas.

 

[strangecosmos]

Pessimistic remarks this week from Waymo’s CEO John Krafcik:

“It’s going to be a really long time — I think decades — before you see this technology everywhere in the world.”

“The trucking shortage is now. ... The use case here is fairly straightforward. Moving goods on freeways from hub to hub is a fairly straightforward application of our technology. It’s much easier than the initial problem we’re trying to solve using Waymo technology in a ride-sharing service. So this is something you could anticipate a material contribution to the world from Waymo over the next couple of years.”​

The implication is that it will take a “couple of years” to deploy freeway autonomy for trucking, and scaling up ride-sharing for passengers in cities will take longer.


Source: WSJ Tech D.Live: Are We There Yet? The Future of Driverless Cars

 

[lunitiks]

What a bunch of baloney. Musk says one more year.

Twitter

(«Probably»... At least it will be «technically possible»... Unless the «regulators» screw it up...)

 

[strangecosmos]

It’s really surprising that John Krafcik is talking this way. My impression from Waymo until now was that they were planning to scale up to dozens of cities within a few years. Now it’s not so clear. “Decades”...?! Like, 20+ years before we see robotaxis in every major city in the world? What about every major city in the U.S.? Or the first ten cities?

Sergey Brin and Chris Urmson were much more aggressive in their timelines when they talked about this in years past. They were saying/hoping that Level 4 would happen at scale by 2020.

My hope is that Tesla by applying ~100x scale to the problem in terms of training data collection will be able to accelerate progress. Might not work out that way, but it’s a more inspiring effort than incrementally rolling out the technology over decades

 

[lunitiks]

Actually, I found a compelling clue on tesla.com/autopilot:

Please note also that using a self-driving Tesla for car sharing and ride hailing for friends and family is fine, but doing so for revenue purposes will only be permissible on the Tesla Network, details of which will be released next year.

That fits hand in glove with Elon's tweet!

Woah!

Wait. The website said so in 2017 too?

And in 2016?

..............

 

[electronblue]

It’s really surprising that John Krafcik is talking this way. My impression from Waymo until now was that they were planning to scale up to dozens of cities within a few years. Now it’s not so clear. “Decades”...?! Like, 20+ years before we see robotaxis in every major city in the world? What about every major city in the U.S.? Or the first ten cities?

Sergey Brin and Chris Urmson were much more aggressive in their timelines when they talked about this in years past. They were saying/hoping that Level 4 would happen at scale by 2020.

My hope is that Tesla by applying ~100x scale to the problem in terms of training data collection will be able to accelerate progress. Might not work out that way, but it’s a more inspiring effort than incrementally rolling out the technology over decades

This in my opinion is a case where you are reading Tesla’s competition pessimistically again and Tesla optimistically. A reasonable interpretation is decades for the whole world does not exclude Waymo ramping up in the United States at all. Waymo’s quote was decades for ”everywhere in the world” not every major city on the world which is a big difference.

If this is unintentional on your part please know this is how your messages read. You have made this point in a couple of places now and your reading of Waymo seems excessively pessimistic like the worst interpretation of what they said instead of best or some middle ground. We would be served well by listening carefully of all the players not hyperbolistically any of them.

 

[strangecosmos]

A reasonable interpretation is decades for the whole world does not exclude Waymo ramping up in the United States at all. Waymo’s quote was decades for ”everywhere in the world” not every major city on the world which is a big difference.

Krafcik also said autonomous trucks on freeways could make a material contribution to the world in a couple of years, and mentioned in the same breath that ride-sharing is a much harder problem. This implies that he thinks ride-sharing will take more than a couple of years to deploy at a scale where it will make a "material contribution".

I assume that deploying a full, commercial ride-hailing service in multiple cities would contact as a "material contribution" for Krafcik. So, I'm inferring that he thinks that outcome is more than a couple of years away.

Listen to the full interview if you think my interpretation is too pessimistic. He repeatedly says things like "it's going to take a long time", and talks about how he thinks Level 5 autonomy is a silly goal that might never be achieved. He emphasizes that Waymo's progress has been slow and incremental; it's taken them years to move from one step to the next.

Krafcik doesn't give us an exact timeline or an exact explanation of a "material contribution", but these statements seem to indicate Krafcik believes Waymo is several years away from scaling up autonomous ride-sharing to the point where it will make a "material contribution" to the world.

 

[electronblue]

@strangecosmos This includes heavy interpretation of words like material contribution. Material contribution could also mean a major business for example and it is easy to see how Level 4 trucking could be the first to become a major business while ride sharing would only happen in smaller pilot regions and not be a real business for some time. Does not mean Waymo will be restricted to Phoenix for years necessarily — or at least that is not what is being said in my view.

This is the problem with interpretation. On one hand we have Tesla and Elon Musk who have been saying Tesla Network details coming in 2017, 2018, 2019... full autonomy in two years since how many years now who make outlandish predictions. Then there are companies who are much more conservative in what they say.

It is probably wise not to read the latter pessimistically and the former optimistically.

 

[strangecosmos]

ride sharing would only happen in smaller pilot regions and not be a real business for some time.

Yes, that is how I’m interpreting Krafcik. That he believes moving out of the small pilot stage to a large-scale commercial enterprise will take more than a couple of years. Until today, I thought Waymo was targeting a large-scale commercial enterprise within a couple of years.

Previously, it seemed like Waymo was targeting 2020 for large-scale ride-hailing in multiple cities. Maybe that is still the case. But if so, Krafcik’s comments are very strange. 2020 is only a couple of years away. So why would he say freeway autonomy can be commercialized within a couple of years, but that commercializing autonomous urban ride-hailing is a much harder problem?

Maybe he was playing coy, or maybe he’s trying to underpromise and overdeliver. Maybe he was just sleepy. But I found his comments very surprising and very strange for the CEO of a company targeting 2020 for the large-scale commercialization of autonomous urban ride-hailing.

This isn’t about interpreting anyone “optimistically” or “pessimistically”; it’s about trying to understand what they believe, based on what they say. Regardless of whether you agree or disagree.

 

[electronblue]

This isn’t about interpreting anyone “optimistically” or “pessimistically”; it’s about trying to understand what they believe, based on what they say. Regardless of whether you agree or disagree.

The thing is one can skew that towards optimistic or pessimistic.

For example we could instead start from that presumption that he was being coy or was trying to underpromise and it would change the whole interpretation. Or that his meaning of ”material difference” means large business and has no bearing on Waymo ramping up taxi service pilot areas for example. In my view you just started from a very pessimistic presumption there and made ”hearing the worst” type of assumptions and reported it accordingly over several threads.

Personally I did not get any sense of disconnect between the interview and what Waymo has been doing and ramping up as an industry leader. I do see a much larger disconnect between Elon saying Tesla will solve full self-driving in 2019 and what we know of them otherwise. Yet you are more optimistic there despite this disconnect. So it comes down to interpretation.

Maybe it is not intentional but I do see what some are saying that you seem to read Tesla very optimistically and others very pessimistically so these interpretations don’t tend to match in a balanced fashion. Just my opinion.