KD5NRH

So, in 9 years, they've done 10%. Only 80 years to go.

Ninety5rpm

No, they did 90% in the first year. They were self-driving all over a number of routes in the bay area in 2010, just like the others like Cruise are doing now. I think Cruise is further along now than Waymo was in 2010, but they don't seem to be nearly as far along as Waymo is today. And that makes sense; they started much later.

tyrion

The old software engineer adage: the first 90% of a project goes quickly, the second 90% goes much slower.

Ninety5rpm

I don't recall this new (Feb 28, 2018) 360-degree video from Waymo being shared here yet. It explains their technology Be sure and move around as you watch it.

robertorolfo

Very interesting, and concerning, article.

These things look like huge red flags to me:

• Disengage for unwanted maneuver of the vehicle
• Disengage for a perception discrepancy
• Disengage for incorrect behavior prediction of other traffic participants
• Traffic light detection
• Unexpected steering due to path change
• Strong unexpected braking

Some of those can only be translated as, 'Nowhere near ready to be used without supervision,' and yet that is exactly what California is going to allow? Are they seriously stupid enough to think that someone monitoring the car remotely, via video, will be nearly as good as someone actually in the car? On what planet are the two things comparable in terms of visibility, vehicle control and, wait for it, the instinct for self-preservation? FFS...

"Unwanted maneuver of vehicle," is insanely dangerous. "Perception discrepancy" is insanely dangerous. "Unexpected steering" is nuts. Ditto the rest.

If someone can't be right there, ready to take over in a split second (as indicated in the article, the humans usually took control in less than one second), then I wouldn't want to be anywhere near those things.

Ninety5rpm

There are 50 companies registered to operate self-driving cars in California. Not all them are actually testing, but those categories of disengagements were culled from the reports of all reporting, regardless of what level they are at in their development, or how far they are from humanless deployment.

What I find most interesting in the article is this point from Waymo:So in each case they can "replay" the situation and find out what would have happened had the driver not chosen to take over. Was it really a problem? If so, they can update the software accordingly.

And, they've been doing this for years, developing sophistication in their simulations as well as advancing the actual driver AI software.

I continue to suspect that the others are not as rigorous... they just haven't been in the game nearly long enough. We'll see.

Ninety5rpm

The misleading nature about the technology by those writing about it is astounding. Here is an article hand-wringing about the effect of solar storms disrupting GPS and thus rendering AVs dependent on GPS useless.

https://www.bloomberg.com/news/artic...s-solar-storms

Although it is mentioned, there is little emphasis on the fact that Waymo discovered years ago that they cannot rely on GPS, and don't.

From 2016:
"That means we don’t have to rely on GPS technology, or a single point of data such as lane markings, to navigate the streets."

https://medium.com/waymo/building-ma...r-723b4d9cd3f4

tyrion

Uber Self-Driving Car Struck and Killed Arizona Woman While in Autonomous Mode


https://gizmodo.com/uber-self-drivin...-au-1823891032

genec

Yup, it's the story all the naysayers have been waiting for...
Uber Suspends Self-Driving Tests After Pedestrian Is Killed In Arizona https://n.pr/2tWQbLm

Uber suspends all testing.

Article has a flaw however... mentions Tesla driver as "the first self driving car fatality..." which is false. Tesla did NOT have and still does not have a full autonomous mode for their vehicles.

trailangel

The news showed a smashed bicycle on the sidewal.k

InOmaha

The car operator must have been on their phone.

Maelochs

Bloomberg Tech article definitely said there was an operator in the vehicle and the lady was jaywalking ... so there is a lot we don't know here. Did she stagger out from between two cars? Was the human driver drunk? Why didn't the backup safety driver take over? I assume from the time of reporting that the incident happened mid-morning Tempe time ... hard to explain.

Luckily Waymo has five million to Uber's two million miles of testing ... no word yet on whether Waymo will go ahead with its AV taxi plan in Phoenix.

robertorolfo

What does that mean, "luckily"?

Hopefully Uber gets sued into oblivion, and everyone else learns their lesson. Also, the AZ governor that signed the executive order allowing this nonsense, should get plenty of heat as well. Who wants to be a human guinea pig for some tech jerks that will play fast and loose with your life?

Maelochs

Lol. yeah. This will end AVs the way the first automobile fatality stopped owner-operated cars.

Rollfast

I think that driver hit a horse or something.

noisebeam

It happened Sunday night at a location on northbound N. Mill Ave between the Red Mountain freeway bridge and Washington/Curry.
https://www.google.com/maps/@33.4367.../data=!3m1!1e3

InOmaha

Based on the map above there was a movie theater there and no place for parked cars on the road. Maybe the person called an Uber to drive them down ....err..... home from the movie.

Maelochs

Thanks. I wonder if she was coming out of the park on the right ... across the way doesn't seem like much to interest a pedestrian. Of course, she could have been walking an established route and habitually cut across right there .... Not enough curve in the road to hide a pedestrian or a car, it looks to me.

Mystery.

squirtdad

article says the person who died was walking a bike which theoretically should have increased detectablity by AV systems.

CB HI

A pedestrian this time, can cyclist be far behind?

https://www.bloomberg.com/news/artic...ash-in-arizona

https://www.cnbc.com/2018/03/19/uber...port-says.html

Or was she a cyclist?


.

rydabent

If Uber persists in using those stupid self driving cars, soon ins companies wont insure them.

noisebeam

It is a musical performance theater and it has a very large parking lot directly south of it. Any pickup/dropoffs should and do occur from the parking lot at the front entrance.

San Rensho 

The woman that was killed was walking a bicycle so I think this qualifies as cyclist getting killed by self driving car.

noisebeam

Zero evidence as to fault.

Maelochs

I assume Waymo won't comment for several days, and will continue with its program as planned.

I am really interested to hear about how this one happened. I wonder if the human driver will be liable?