Thats an incorrect benchmark. There are millions of human driving cars on the road in any given moment compared to a handful of selfish driving cars. The correct safety comparison should be failure rate, with a statistically significant sample size
Thats an incorrect benchmark. There are millions of human driving cars on the road in any given moment compared to a handful of selfish driving cars. The correct safety comparison should be failure rate, with a statistically significant sample size
The Gallant Pig Man wrote:
DSB15 wrote:
Woman purportedly homeless and pushing bicycle (loaded down with god-knows-what) across a well-traveled street. And we all know how considerate the homeless are when it comes to jaywalking. Notice that runners are never in this situation because they jayRUN. Also, there was a traffic signal/crosswalk less than 100 meters away.
"The penalty for violating The Algorithm is death."
People jaywalk ALL THE TIME, especially in big cities. We do not simply shrug our shoulders and place all of the blame on the pedestrian if a human driver runs down someone who is crossing outside of the crosswalk. If self-driving cars are not equipped to deal with this kind of regular human behavior, then they have some serious flaws. Are you going to be singing the same tune if it's a kid instead of a scary, despicable homeless person next time?
These AV companies are turning the rest of us into guinea pigs while they test out their new, unproven products with little regulatory oversight, and but they want to completely avoid direct liability if someone gets killed. Too bad!
We should blame the jaywalker for not using a crosswalk 50 meters away. We should blame the jaywalker in this instance. The scary despicable homeless person doesn't give a f*ck about themself; they sure as heck don't give a f*ck about you. Notice how many of them don't just jaywalk, they dawdlewalk.
I suspected it would be an uncaring inner city person before even checking the details. Don't know why it became are part of poor/wellfare culture to completely ignore traffic if the message gets out they might start fearing our robot overlords and behaving themselves.
joedirt wrote:
Self driving cars are a bad idea. How many times does your computer freeze up in a given year? Now what happens when it freezes up while it is driving?
End of discussion. The idiots defending this nonsense need to stfu. Shame.
There is good reason they chose Chandler and Tempe to test.. the roads are wide and straight, the intersections signs are huge and easy to read, road surfaces are regularly repaved and have no potholes, the weather is always nice, and every street runs NSEW.
https://jalopnik.com/people-arent-sure-if-theyd-sue-after-an-accident-with-a-1823951013Send in the clowns wrote:
Let's see if the lawyers will kill the concept.
You've got to hope that the robot cars only run over the 50% of people who won't sue.
At first, from all the accounts i'd read, this was a freak accident. But it seems like a human could have swerved/braked/tried to prevent the crash:
yikes.
It looked like she appeared out of thin air in the middle of the road. But even counting reaction time, an alert human would have at least been able to slow down a bit, maybe correctly swerve left.
Could be the decision tree was based on objects appearing at the edge of its sensor range and tracking them from there. That's an important way to tell if something's an object and not part of the background. For example if there were a flash of light that revealed a shiny repair patch on the road ahead, it might decide that's part of the road.
She appeared right when the headlights would have started lighting up the road where she was and making it look brighter.
Bad Wigins wrote:
It looked like she appeared out of thin air in the middle of the road. But even counting reaction time, an alert human would have at least been able to slow down a bit, maybe correctly swerve left.
Could be the decision tree was based on objects appearing at the edge of its sensor range and tracking them from there. That's an important way to tell if something's an object and not part of the background. For example if there were a flash of light that revealed a shiny repair patch on the road ahead, it might decide that's part of the road.
She appeared right when the headlights would have started lighting up the road where she was and making it look brighter.
Why no one blaming the jaywalker? She should have seen the headlights from a mile away.
Maybe the whole "in-car camera" view was re-created with digital video technology after the fact.
Another use of AI wrote:
Maybe the whole "in-car camera" view was re-created with digital video technology after the fact.
What’s the thought behind your speculation? I would assume they would have a built in camera like the patrol cars, no?
Also, why have a person behind the wheel of a self driving car? Are the vehicles not sustainable to be on its own? Sorry for all the questions—just trying to learn how they operate.
I think ALL cars should have in-car dash-cams. Then we will start to suss out the perpetrators of so-called "accidents" on the highway.
Need more Car Regulation wrote:
I think ALL cars should have in-car dash-cams. Then we will start to suss out the perpetrators of so-called "accidents" on the highway.
Most collisions are not on the "highway" per se.
Bad Wigins wrote:
It looked like she appeared out of thin air in the middle of the road. But even counting reaction time, an alert human would have at least been able to slow down a bit, maybe correctly swerve left.
Could be the decision tree was based on objects appearing at the edge of its sensor range and tracking them from there. That's an important way to tell if something's an object and not part of the background. For example if there were a flash of light that revealed a shiny repair patch on the road ahead, it might decide that's part of the road.
She appeared right when the headlights would have started lighting up the road where she was and making it look brighter.
Anyone with night photography or videography experience would recognize that that video does not have the dynamic range of human eyes, meaning it's turning visible shadow details black. In real life, I'm pretty certain that woman would have been visible to the driver if the driver was actually looking up rather than down in her lap before the headlights lit the walker up for the video. The lidar should have definitely have picked up the woman in the dark. That's part of the whole selling point of self-driving vehicles. She was in a visible position for quite a while moving slowly across that left lane first.
If the sensors, when driving at night, are only picking things up in the last second when it's too late to brake then they are overdriving the sensors. If a human driver was doing that, they would be overdriving their headlights. From the articles that I've read, it seems that the sensors didn't brake at all, so it's not like the sensors picked up the walker too late, they might not have picked up the walker at all.
This should also be a wake up call for Tesla, who is trying to do autonomous driving without lidar sensors, claiming that visible light cameras are all that are needed.
City Streeter wrote:
Need more Car Regulation wrote:
I think ALL cars should have in-car dash-cams. Then we will start to suss out the perpetrators of so-called "accidents" on the highway.
Most collisions are not on the "highway" per se.
Despite common usage, "highway" actually refers to roadways in general, not just limited access highways.
DSB15 wrote:
Why no one blaming the jaywalker? She should have seen the headlights from a mile away.
She should have, but a normal driver would likely have seen her and avoided hitting her. It's a clear night, and the car was moving 38 mph, and had pretty bright looking headlights. If you can't see a person walking slowly across the left lane from the median into your path in that situation, you are overdriving your headlights, going too fast for the visibility.
The selling point of autonomous cars has been that they can potentially be safer than humans when they become fully developed. If the real world was a totally controlled situation, that might be possible, but I have my doubts. Primarily with weather: ice and variable ice and the ability of computers to deal with super low friction surfaces, especially when on slopes (where you'd slide through a stop sign at the bottom of a hill even creeping down towards it at 2 mph) or when the ice unpredictable, freezing rain icing up sensors on contact, white-out blizzards, snow, snow obscuring road lines or even road edges/the shape of the road, heavy fog. Also there are proposed vehicles without steering wheels, where you wouldn't be able to take over in difficult situations. You might end up with stopped autonomous cars blocking the roads in an icestorm or blizzard. How would you park a car without a steering wheel in a field, such as a construction site? Or even in your garage, if you are picky about where in a garage your car is parked to leave room to open doors or walk.
Curious Georgina wrote:
Also, why have a person behind the wheel of a self driving car? Are the vehicles not sustainable to be on its own? Sorry for all the questions—just trying to learn how they operate.
The driver is a "safety driver". Autonomous vehicles are in testing and not ready for prime time yet. There are situations where they don't respond correctly, such as this accident, and the developing companies learn from those things and improve the cars. The "safety driver" is supposed to be backup, which is why authorities let these not-ready-yet cars/experiments on the road in the first place. Unfortunately, the "safety driver" wasn't paying attention, looking down in her lap more than looking at the road. The "safety driver" not paying attention kind of predictable. It's an effect that people have predicted for vehicles that do more of the driving for you, such as the autosteering to keep people in lanes, autobraking, or radar cruise control systems that a lot of new cars have.
zzzz wrote:
Anyone with night photography or videography experience would recognize that that video does not have the dynamic range of human eyes, meaning it's turning visible shadow details black. In real life, I'm pretty certain that woman would have been visible to the driver if the driver was actually looking up rather than down in her lap before the headlights lit the walker up for the video. The lidar should have definitely have picked up the woman in the dark.
I know. If the dashcam had exposure set higher, it would have seen much farther up the road, though everything up close in the headlights would burn out.
either
1) something malfunctioned, or failed to function as planned, and it didn't see her
2) it saw her but didn't ID her correctly, or
3) it ID'd her but failed to react as planned
I still think 2 is the most likely
RIP: D3 All-American Frank Csorba - who ran 13:56 in March - dead
RENATO can you talk about the preparation of Emile Cairess 2:06
Running for Bowerman Track Club used to be cool now its embarrassing
Rest in Peace Adrian Lehmann - 2:11 Swiss marathoner. Dies of heart attack.
Hats off to my dad. He just ran a 1:42 Half Marathon and turns 75 in 2 months!
Great interview with Steve Cram - says Jakob has no chance of WRs this year
