Self-Driving Uber Car Kills Arizona Pedestrian (Phoenix: license, transportation, automobile)

[DetroitN8V]

Originally Posted by Cloudy Dayz

This is not a single issue. This is not even the first time someone has been killed by one of these death traps.

The repeated use of absurd hyperbole (i.e., “death traps”), along with posting history says it all.

[Boompa]

What I'm waiting for is for two of those that park themselves to fight it out over a parking spot

[Cloudy Dayz]

Originally Posted by DetroitN8V

The repeated use of absurd hyperbole (i.e., “death traps”), along with posting history says it all.

These cars are killing people. What do you suggest we call it?

[DetroitN8V]

Originally Posted by Cloudy Dayz

These cars are killing people. What do you suggest we call it?

“These cars are killing people.” LOL. That’s all.

[craig11152]

according to the New York Times....

Waymo, formerly the self-driving car project of Google, said that in tests on roads in California last year, its cars went an average of nearly 5,600 miles before the driver had to take control from the computer to steer out of trouble. As of March, Uber was struggling to meet its target of 13 miles per “intervention” in Arizona, according to 100 pages of company documents obtained by The New York Times and two people familiar with the company’s operations in the Phoenix area but not permitted to speak publicly about it.

While Ubers 13 miles per intervention is abysmal I have to say Waymo's "average of nearly 5,600 miles before the driver had to take control from the computer to steer out of trouble." is not exactly the stuff we should be excited about. In fact we should be utterly shocked at those numbers and realize this technology is long way from being truly "driverless"

https://www.nytimes.com/2018/03/23/t...s-arizona.html

[Bitey]

Originally Posted by craig11152

according to the New York Times....



While Ubers 13 miles per intervention is abysmal I have to say Waymo's "average of nearly 5,600 miles before the driver had to take control from the computer to steer out of trouble." is not exactly the stuff we should be excited about. In fact we should be utterly shocked at those numbers and realize this technology is long way from being truly "driverless"

https://www.nytimes.com/2018/03/23/t...s-arizona.html

If a car can drive 5,600 miles without intervention, we're a lot closer than you may think. 5,600 miles is about how much my wife drives per year. She'd be pretty excited at the idea of a car that would only require her to intervene once a year. And I bet that required intervention rate would be a lot lower if all the other vehicles on the road had the same autonomous system.

Now the comparison between the Uber and Waymo would need a lot more fleshing out to determine if an apples-to-apples comparison. Uber has been testing their technology largely in dense urban and suburban environments. How many of Google/Waymo's test miles were in controlled conditions or used pre-determined routes, what was the ratio of test miles in dense urban vs suburban vs low-density exurban/rural environments? If Waymo is testing in the same sort of environments as Uber at about the same ratio of urban vs suburban vs exurban vs rural and their results are that much better, the first obvious takeaway is that Uber needs to stop burning its cash on its autonomous development program and team up with someone that's already way ahead of them.

Plus there are numerous other players in the game such as major auto manufacturers, automotive suppliers, and other tech companies like Intel, Qualcomm, NVDIA, etc. What's their mileage per intervention ratio and in what sort of environments? Pretty soon we'll start seeing a shakeout when some players figure out they're too far behind either in progress or resources and we'll start seeing mergers of companies and data. In fact it's already starting to happen. As the data from these companies starts to merge and compile, they'll get an increasingly clear picture of what works, what doesn't, what data is necessary and what data can be discarded.

[craig11152]

Originally Posted by Bitey

If a car can drive 5,600 miles without intervention, we're a lot closer than you may think.

Are you serious? It means your wife, or anybody else for the moment can't (shouldn't) take their eye off the road or their hands of the wheel. In which case at the moment there is hardly a point to the
autonomous mode. At that pace your wife would be in one accident a year. Is she currently averaging more than that?

[lvmensch]

Originally Posted by craig11152

Are you serious? It means your wife, or anybody else for the moment can't (shouldn't) take their eye off the road or their hands of the wheel. In which case at the moment there is hardly a point to the
autonomous mode. At that pace your wife would be in one accident a year. Is she currently averaging more than that?

Waymo is not good enough yet but getting closer...at 50,000 or so between interventions they can probably launch. The thing you need to understand is that an intervention does not mean an accident. You have to look at the exact nature of the intervention. If, for instance a critical LIDAR fails the machine may simply slow down and stop...or proceed at a reduced speed to a safe place to stop. Or it may introduce a remote operator to navigate the vehicle to a safe place where the repair truck can meet it.

They do need to be sure that critical failures like the UBER accident have very low probability...maybe past one per million or more. Or even with odds similar to being hit by a meteorite.

[Bitey]

Originally Posted by craig11152

Are you serious?

Yeah I'm serious, considering about 10 years ago the miles per intervention ratio was zero. To go from there to 5,600 miles per intervention is some pretty serious progress.

Originally Posted by craig11152

It means your wife, or anybody else for the moment can't (shouldn't) take their eye off the road or their hands of the wheel.

No it doesn't, at least not in the latter case.

Originally Posted by craig11152

In which case at the moment there is hardly a point to the
autonomous mode.

Autonomous features have already been rolled out bit by bit. They still require the driver stay vigilant but they have added tremendous convenience and safety benefits, collision avoidance and reduction of driver fatigue being chief among them.

Originally Posted by craig11152

At that pace your wife would be in one accident a year. Is she currently averaging more than that?

Only if she declined to pay any attention at all, which admittedly is a real danger for a car that can drive itself 99% of the time without the need for intervention. Which is why Waymo hasn't deployed its fully autonomous system yet. But we're closer than I suspect you think we are to seeing consumer-ready, fully autonomous cars on the roads.

[craig11152]

Originally Posted by lvmensch

They do need to be sure that critical failures like the UBER accident have very low probability...maybe past one per million or more. Or even with odds similar to being hit by a meteorite.

One per million what?