Slashdot Mirror
A Sleeping Driver's Tesla Led Police On A 7-Minute Chase (sfchronicle.com)
"When a pair of California Highway Patrol officers pulled alongside a car cruising down Highway 101 in Redwood City before dawn Friday, they reported a shocking sight: a man fast asleep behind the wheel," reports the San Francisco Chronicle:
The car was a Tesla, the man was a Los Altos planning commissioner, and the ensuing freeway stop turned into a complex, seven-minute operation in which the officers had to outsmart the vehicle's autopilot system because the driver was unresponsive, according to the CHP...
Officers observed Samek's gray Tesla Model S around 3:30 a.m. as it sped south at 70 mph on Highway 101 near Whipple Avenue, said Art Montiel, a CHP spokesman. When officers pulled up next to the car, they allegedly saw Samek asleep, but the car was moving straight, leading them to believe it was in autopilot mode. The officers slowed the car down after running a traffic break, with an officer behind Samek turning on emergency lights before driving across all lanes of the highway, in an S-shaped path, to slow traffic down behind the Tesla, Montiel said. He said another officer drove a patrol car directly in front of Samek before gradually slowing down, prompting the Tesla to slow down as well and eventually come to a stop in the middle of the highway, north of the Embarcadero exit in Palo Alto -- about 7 miles from where the stop was initiated.
Tesla declined to comment on the incident, but John Simpson, privacy/technology project director for Consumer Watchdog, calls this proof that Tesla has wrongly convinced drivers their cars' "autopilot" function really could perform fully autonomous driving...
"They've really unconscionably led people to believe, I think, that the car is far more capable of self-driving than actually is the case. That's a huge problem."
Officers observed Samek's gray Tesla Model S around 3:30 a.m. as it sped south at 70 mph on Highway 101 near Whipple Avenue, said Art Montiel, a CHP spokesman. When officers pulled up next to the car, they allegedly saw Samek asleep, but the car was moving straight, leading them to believe it was in autopilot mode. The officers slowed the car down after running a traffic break, with an officer behind Samek turning on emergency lights before driving across all lanes of the highway, in an S-shaped path, to slow traffic down behind the Tesla, Montiel said. He said another officer drove a patrol car directly in front of Samek before gradually slowing down, prompting the Tesla to slow down as well and eventually come to a stop in the middle of the highway, north of the Embarcadero exit in Palo Alto -- about 7 miles from where the stop was initiated.
Tesla declined to comment on the incident, but John Simpson, privacy/technology project director for Consumer Watchdog, calls this proof that Tesla has wrongly convinced drivers their cars' "autopilot" function really could perform fully autonomous driving...
"They've really unconscionably led people to believe, I think, that the car is far more capable of self-driving than actually is the case. That's a huge problem."
"I just had a hell of a dream. What the?!..."
Table-ized A.I.
Thank you, Elon
The CEO of Waymo!
John McAfee 'It was like that time I hired that Bangkok prostitute; to do my taxes, while I fucked my accountant'
"They've really unconscionably led people to believe, I think, that the car is far more capable of self-driving than actually is the case. That's a huge problem."
And yet...nobody was hurt, no cars were wrecked...so it did much better than any other car with a sleeping driver.
Driver fell asleep at the wheel, and instead of crashing into things as in a conventional car, semi-autonomous vehicle came to complete stop with no loss to life or property.
The California Highway Patrol on Friday pulled over a Tesla Model S that was traveling down the road—but whose driver appeared to be asleep at the wheel. The vehicle was traveling southbound on Highway 101 in Palo Alto.
Officers said that they were unable to get the man's attention.
"One of the officers basically ended up going in front of the vehicle and basically tried to slow it down," a California Highway Patrol spokesman told KCBS radio. The process took about seven minutes, and the car traveled for about seven miles before coming to a stop.
The driver was Alexander Samek, who serves on the Los Altos Planning Commission. He was arrested for driving under the influence.
So how was the vehicle able to travel for more than seven minutes with an apparently sleeping driver? The obvious theory is that the Model S had its Autopilot system turned on, but officials said on Friday that they hadn't confirmed that yet. It's quite possible that Autopilot saved Samek's life.
The situation is a bit of a puzzle because Autopilot is supposed to detect if a driver's hands are on the wheel and disengage if they're not. Tesla has steadily tightened up these rules, with recent revisions of the software warning drivers in as little as 30 seconds. So if the driver did fall asleep at the wheel the car should have started slowing down on its own within a few minutes.
In a similar case back in January, police encountered a man asleep behind the wheel of a Tesla car on the San Francisco–Oakland Bay Bridge. When police woke him up, he insisted that everything was fine because his vehicle was "on autopilot." Unfortunately for him, there's no autopilot exception to drunk-driving laws.
Top comment: onkeljonas
"It's entirely plausible that he was asleep with his hands on the wheel."
I'm not going to lambaste Tesla over this.
The guy was drunk. Has he driven drunk before, in the Tesla or in another car (whether he's gotten caught or not)? Did he intend to have the Telsa drive him home, or did he start driving himself and just fell asleep?
It does seem obvious that the driver made some very bad decisions, regardless.
#DeleteChrome
is that the driver was too exhausted to be driving in the first place. And, that had he not be driving a car equipped with as an intelligent a safety system, there would have been a substantially higher probability of injury, loss of life or property.
" calls this proof that Tesla has wrongly convinced drivers their cars' "autopilot" function really could perform fully autonomous driving... "
Those of us who know better call this FUD.
09 F9 11 02 9D 74 E3 5B - D8 41 56 C5 63 56 88 C0 45 5F E1 04 22 CA 29 C4 93 3F 95 05 2B 79 2A B2
Did he pass the FST ? The title of the article read he was a drunk driver but no where in the article does it state he was. Did he just decide to take a nap ? Why should Tesla be held to truth in advertising while other car manufacturers can show their cars doing rail slides on bridges and many other behaviors that a car can not accomplish ? Missing far too many basic facts to really render any sort of judgement. The Chronicle should fire whomever wrote this trash and hire a qualified writer/reporter, say your average 5th grader.
errr....umm...*whooosh* *whoosh* Is this thing on ?
Whatever the case, Alexander Samek needs to be fired by the city of Los Altos and given the same punishment as any "commoner" would receive for a DUI (ie. suspended licence, large fines and mandatory DUI classes).
Stupid fucks like him are responsible for murdering people every day.
Tesla declined to comment on the incident, but John Simpson, privacy/technology project director for Consumer Watchdog, calls this proof that Tesla has wrongly convinced drivers their cars' "autopilot" function really could perform fully autonomous driving...
"They've really unconscionably led people to believe, I think, that the car is far more capable of self-driving than actually is the case. That's a huge problem."
So let me see if I have this straight. 10,000 people a year die in DUI crashes, yet all these drunk drivers are not the fault of Ford, Toyota, Chevy, Nissan, etc. The liability is totally on the driver that decided to operate a vehicle while intoxicated.
However when someone drives a Tesla drunk, it is Tesla's fault. Yes, that makes perfect sense, Mr. Simpleton. I mean Simpson.
Better known as 318230.
Nor anyone else for that matter. I'm not sure what this idiot's plan was, but except for getting pulled over, the car was going it's job. What I'm not sure of is why they didn't just merge in front of it and hit the brakes? The Drive On Nav feature still requires driver input to merge around slow traffic (by activating the indicators).
By the way, plenty of people fall asleep behind the wheel, and drive under terrible circumstances (drunk, high, exhausted). Whatever this driver did, he is responsible for what he is doing, not Tesla.
Plenty of people survive falling asleep behind the wheel.
Just not usually the ones who are in the path of the car.
The driver? Pretty much they stand a good chance of living in a modern car no matter what they hit.
Probably they didn't want to create a chain reaction rear end accident. By running a traffic break behind the Tesla before then slowing down and bringing the Tesla to a halt by slowing a patrol car down in front if it, they cars behind were already slowed down a bit, there was more of a gap between following cars and the Tesla, and drivers behind anticipated that something was going on ahead. This reduced the chances of someone crashing into the Tesla when it stopped in the middle of a traffic lane for no apparent reason in the early morning hours on a freeway.
At least that's my guess.
Why is there an "insightful" mod and why isn't it "-1"? If I wanted insight, I wouldn't be reading
Emergency services should have an override on automated vehicles, specifically for situations like this.
They've got one, the same one they use with human drivers. They used it. Problem solved.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
If he vas elected, still no. And more so not if he was chosen for the job by HR, as this has nothing to do with his job. What you do in your free time, is none of your bosses business. But I wonder, how come the autopilot did not react on the tools police use - lights and sound, which is same as red light by autopilot understanding. I doubt Tesla just forgot such functionality when implementing traffic rules adherence.
... the opposite of John Simpson's statement. Tesla cars can actually drive better than their owners and furthermore, conventional car manufacturers are shitting their pants after this 'incident' that also proves that Tesla cars aren't easy to carjack on a freeway. The highway police had a such a hard job at stopping the car I think you can assume that it could easily defend itself.
Who is John Simpson really serving?:
Tesla declined to comment on the incident, but John Simpson, privacy/technology project director for Consumer Watchdog, calls this proof that Tesla has wrongly convinced drivers their cars' "autopilot" function really could perform fully autonomous driving...
This is just fakenews:
"They've really unconscionably led people to believe, I think, that the car is far more capable of self-driving than actually is the case. That's a huge problem."
Bach says it all.
One of the main points of an autonomy system is to improve safety via pairing the vehicle's constant attentiveness with a human's decision-making ability. The more annoying you make your system with its nagging, the less people will use it, defeating any safety advantages.
Tesla actually has spent money on the hardware that would be needed. Look. You see that? That's a driver-facing camera. Every Model 3 has one. So what "cost savings", exactly, do you think they're getting?
Tesla has experimented endlessly over the years with nag frequencies, types of nag, and types of driver monitoring. This is what they've arrived at as the best balance between "encouraging people to actually use it" and "discouraging inattentive driving". And by and large, it works very well - even if some drunk happened to pass out at the wheel. Which, while we're on that subject... what's the alternative? Have drunks ever been prone to not driving? When a drunk passes out at the wheel, would you rather the car just crash? It's still DUI either way, but in the former case, everyone walks away unscathed, while in the latter case some random person has a drunk crash into and possibly kill them.
That's not to say that the current approach is perfect - far from it. There's a difference between a naggy, "oh my god you looked away from the windshield" system, and a system that can detect if a person has passed out (but still had their hands on the wheel), for example. Implementing the latter would very much be a good thing. But with the former, if you drive people off of using it, you lose out on any potential for improving safety.
You people make me envy the deaf and the blind!
This is why we have a legal system with standards of evidence and sentencing. We do not need to implement extra-judicial systems in addition to what we have.
There's a difference between a naggy, "oh my god you looked away from the windshield" system, and a system that can detect if a person has passed out (but still had their hands on the wheel), for example. Implementing the latter would very much be a good thing. But with the former, if you drive people off of using it, you lose out on any potential for improving safety.
There's also a sensible middle ground. If someone looks away for the windshield for long enough, then they ought to be reminded to keep their eyes on the road. Just like the steering wheel nags, there's a sweet spot for eyes-on-road nags as well.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Nor anyone else for that matter.
Yep.
"They've really unconscionably led people to believe, I think, that the car is far more capable of self-driving than actually is the case."
Apparently the car was quite capable of not hitting anything and of coming to a complete stop when something was put in its way.
"That's a huge problem."
Given that people fall asleep at the wheel every single day and crash, I'd call it a huge win.
No sig today...
Besides, the police DO have an override- they did exactly that. They boxed in the car and forced it to stop.
They didn't even have to box it in. They just put one car in front of it and slowed down to a halt. If you were driving the car, you would have just changed lanes, but this car was on autopilot and doesn't change lanes on its own.
The reason it took so long is that you can't just stop a car on the motorway, because of the huge risk that others drive into it. One driver had to slow down traffic behind the car to create a huge gap for safety, that's what took the time.
so can specified overrides
On a limited-access highway, "coming to a stop" is usually WORSE than "staying in the lane & maintaining a normal cruising speed". Autopilot lanekeeping & collision-avoidance is now better on average at avoiding accidents than most human drivers. It's UNUSUAL situations that create the danger.
A car staying in its lane & moving appropriately is the norm on a freeway. A car stopped on the shoulder creates an active road hazard for everyone else. Remember, his autonomous vehicle wasn't the only one on the road. It's not a safety improvement if your "safety feature" creates a situation that's MORE dangerous than letting the car just do what it does best -- follow the lane, avoid collisions, and behave in a predictable manner.
If you really need to punish drivers, add warning lights to visually communicate to other drivers that a car is operating without active human control. On a limited-access highway not under construction in good weather, inattentiveness with autopilot is a statistical non-issue... and in the real world, it's probably a net improvement over human drivers who are semi-distracted. On a non-freeway that has at-grade cross traffic, it's a genuine hazard. In bumper-to-bumper city-street gridlock, it's a non-issue (the car can stop within inches if necessary). On city streets at 30mph, it's likely to be dangerous.
The point is, there is no blanket "one size fits all" rule. Autonomous systems have known constraints. Stay within them, and you're fine. Deviate, and you create problems. Ignore the constraints entirely, and you're in uncharted territory risk-wise. Following lines & not colliding is easy. Finding a safe place to pull over is enormously harder.
Autopilot lanekeeping & collision-avoidance is now better on average at avoiding accidents than most human drivers.
This (along with the assumption that autopilot is always going to be operational and operate as intended) is clearly the assumption underlying your entire post, and I'd be surprised if you have any real data to back that. If you'd like to share some, I'd be happy to look at it. (Remember, it has to span a reasonable sample of all potential weather conditions, road conditions, and routes, not just a cherry-picked sandbox.)
A car stopped on the shoulder creates an active road hazard for everyone else.
Limited-access highways are designed with emergency breakdown lanes that put the car completely out of the travel lanes. The only reason a car outside the travel lanes would cause a hazard to people inside the travel lanes is if the people inside the travel lanes are driving unsafely.
Again, where's the data to show that keeping a 2018 semi-autonomous vehicle rolling down the highway without a conscious driver is safer in the aggregate than pulling that car to the side of the road?
Following lines & not colliding is easy.
Yeah. Just ask Joshua Brown.