A Waymo Plowed Through a Construction Zone, Floored It, and Got Chased by the Cops With a Terrified Passenger Along for the Ride

[Zoidz]

Yahoo Autos - A cautionary tale for all users of any level of autonomy - it can and does fail with potentially disasterous results.

"According to Slade, the car initially tried to do the sensible thing and merge out of the closed lanes. When that didn't pan out, it apparently gave up on diplomacy and committed fully to the bit, plowing directly into the active construction zone. There were signs. There were lights. There were cones. The Waymo respected exactly none of them, then accelerated for good measure.

If that weren't enough to ruin an evening, a nearby highway patrol car clocked the situation and gave chase. Slade and his fiance got to enjoy the surreal experience of being passengers in a vehicle being pursued by police, with no steering wheel to grab and no driver to yell at. The car eventually wandered out of the construction zone, took an exit, and went for a little tour of a residential neighborhood before the ordeal ended.

"In that moment it's like, 'oh this technology is not ready. This is 100% not ready,'" Slade said, adding that he genuinely feared someone could have been hit or that the car could have crashed.
"

Sponsored

 

[skyguyscott]

Many engineers find it unimaginable that others do not think systematically like they do. Software development teams are tasked with delivering code that does a specific thing in a specific way within specific parameters. AI works with recognized patterns.

Then Life happens, and the system fails, and the engineers act surprised, or explain that the event in question fell outside the design parameters, and/or that they have now identified areas of improvement and are working on updates that will improve performance. Sorry about the fatalities.

 

[Dark-Fx]

This is the problem with AI driving too much like Human drivers.

 

[loqueid]

Seems plausible to engineer a big red "stop and pull over" button, right?

 

[Zoidz]

Seems plausible to engineer a big red "stop and pull over" button, right?

Good point. I guess I just assumed there was some sort of emergency stop functionality in these driverless vehicles.... No?

 

[LivingInKaos]

This is the problem with AI driving too much like Human drivers.

So true.

 

[LivingInKaos]

Good point. I guess I just assumed there was some sort of emergency stop functionality in these driverless vehicles.... No?

You can stop the ride in the app or on the passenger screen - but that is not quick as it looks for an appropriate place to pull over and stop. The better way is to just open the door- it will break and stop. Then push the button to contact an agent.

 

[Seano]

Many engineers find it unimaginable that others do not think systematically like they do. Software development teams are tasked with delivering code that does a specific thing in a specific way within specific parameters. AI works with recognized patterns.

Then Life happens, and the system fails, and the engineers act surprised, or explain that the event in question fell outside the design parameters, and/or that they have now identified areas of improvement and are working on updates that will improve performance. Sorry about the fatalities.

Just did a couple of days of back to back 8 hour days driving the Rivian and was so appreciative for the driver assist features - both on highway and secondary roads (did about 1/2 & 1/2 on the drives).

I did realize, however, just how many “new” circumstances one encounters when driving - and how quickly we can observe, process and adjust to the condition vs the computer’s “decisions” which aren’t always quite right.

 

[Jeff B.]

I know people want full Autonomy, full self drive, etc.., but the technology will continue to be flawed and plagued by faults. Humans are faulty by nature, some more than others, so complex machines have the ability to eliminate lots of human error, but they will not be perfect 100% of the time. Engineers recognize these limitations and make every attempt to solve them. It's marketing, consumer demand, executives and shareholders that will always push engineers to deliver products to market with limited features or unknown faults. Companies will market things like "significantly better safety records", "85% reduction in injury-causing crashes" or "57% reduction in police-reported crashes compared to human drivers." Again looks great on paper, but that leaves a lot of room for failure.

In this case there most likely was, if not, certainly should have been, an easy to implement failsafe feature that the people in the car could activate for when the car makes a mistake. Same as operating any machine, the user should be able to override the system easily.

 

[UnsungZero_OldTimeAdMan]

Seems plausible to engineer a big red "stop and pull over" button, right?

Sure, but imagine the abuse of it. They'd have to devise it in such a way that it couldn't be abused to hurt Waymo's business. And that would probably mean a process where it couldn't be used as quickly as actual emergencies require.

 

[loqueid]

Sure, but imagine the abuse of it. They'd have to devise it in such a way that it couldn't be abused to hurt Waymo's business. And that would probably mean a process where it couldn't be used as quickly as actual emergencies require.

Use the Uber puke rule. If it happens, you pay (with exceptions for verifiable emergencies).

 

[renderpaz]

"85% reduction in injury-causing crashes" ... Again looks great on paper, but that leaves a lot of room for failure.

I'm not sure what your point is.

It seems you want the user to have an eject button or something? Waymos already have several tools for the rider to use, but those all rely on the passenger. The reason more tools aren't offered? The likelihood of those emergency stop buttons causing harm are greater than them reducing harm.

Over 36,000 people died last year in auto accidents - 4 per hour.

Waymo is already saving lives, of course there will stories like this, what we should all be exclaiming about is how a robotaxi service that does 500,000 rides per week doesn't have far MORE stories like this.

We must all check the availability bias on stories like this and not let perfect be the enemy of really friggin good.

 

[iansriv]

Couple of friends arrived about an hour late for dinner last week. Apparently, they used Waymo and it took a "scenic" route. They could not do anything at all. Just sat there until it finished.

 

[SwampNut]

They could not do anything at all.

Bullshit. When my Waymo was attacked by a crazy homeless person, someone with remote control was available in seconds.

It's interesting that when people do this--at lot--it's often non-news. Because it's so common. But when it's autonomy like this or the misreported Tesla event a few days ago, it's everywhere I look.

 

[iansriv]

Bullshit. When my Waymo was attacked by a crazy homeless person, someone with remote control was available in seconds.

It's interesting that when people do this--at lot--it's often non-news. Because it's so common. But when it's autonomy like this or the misreported Tesla event a few days ago, it's everywhere I look.

What happened mate? Why are your knickers in a twist over what my friend experienced? He certainly wasn't attacked by a homeless person but he unsuccessfully tried to contact Waymo to get to his destination without the detour.

Sponsored