Man Tests If Tesla Autopilot Will Crash Into Wall Painted to Look Like Road https://futurism.com/tesla-wall-autopilot
I see this whole thing is a business viability narrative wherein Tesla would be even further under water if they were forced to admit that LiDAR may possess some degree of technical superiority and could provide a reliability and safety uplift. It must have taken millions of dollars in marketing budget to erase the customer experiences around the prior models of their cars that did have this technology and performed accordingly.
I suspect it would be a major undertaking to add LiDAR at this point because none of their software is written to use it
A simple example. I was coming out of a business driveway, turning left onto a two lane road. It was dark out with no nearby street lights. There was a car approaching from the left. FSD could see that a car was coming. However, from the view of a camera, it was just a ball of light. There was no reasonable way the camera could discern the distance given the brightness of the headlights. I suspected this was the case and was prepared to intervene, but left FSD on to see how it would respond. Predictably, it attempted to pull out in front of the car and risked a collision.
That kind of thing simply can not be allowed to happen with a truly autonomous vehicle and would never happen with lidar.
Hell, just this morning on my way to work FSD was going run a flashing red light. It's probably 95% accurate with flashing reds, but that needs to be 100%. That being said, my understanding is the current model being trained has better temporal understanding such that flashing lights will be more comprehensible to the system. We'll see.
I am willing to experiment in many ways with things in my life, but not WITH my life.
It would be like owning your own bus.
As someone in the industry, I find the LiDAR discussion distracting from meaningful discussions about redundancy and testing
In the water test, Rober has the Tesla driving down the center of the road, straddling the double yellow line. Autopilot will not do this, and the internal shots of the car crop out the screen. He almost certainly manually drove the car through the water and into the dummy.
One person tried to reproduce Rober's Wile E. Coyote test using FSD. FSD v12 failed to stop, but FSD v13 detected the barrier and stopped in time.[2]
Lidar would probably improve safety, but Rober's video doesn't prove anything. He decided on an outcome before he made the video.
Of course, that's why traffic accidents are called 'accidents.' Drivers wouldn't crash their cars, but they do.
¯\_(ツ)_/¯
It's where a bunch of cycling nutters (I'm one of them) post local news stories where a driver has crashed into a building ("It wasn't wearing hi-viz!")
That said, I do think using only visual cues is a stupid self-imposed restriction. We shouldn't be making self-driving cars like humans, because humans suck horse testicles at driving.
Hardly. We drive hundreds of billions of miles every month and trillions every year. In the US alone. You're more likely to die from each of the flu, diabetes or a stroke than a car accident.
If those don't get you, you are either going to get heart disease or cancer, or most likely, involve yourself in a fatal accident; which, will most likely be a fall of a roof or a ladder.
IMHO it kinda is. It's 13x as many people as died in 9/11
No. It's 1.25 per 10,000 per capita. Most people understand the risk ahead of time and yet still choose to drive. They clearly don't think it is.
> It's 13x as many people as died in 9/11
And 50x 9/11 many people die of accidental self inflicted injury. This is an absurd metric.
The majority of those people who had their lives cut short cut it short themselves and didn't take anyone with them.
Likewise, that 40k includes 6k pedestrians and 6k motorcyclists.
You can't just take the 40,000 figure and do _anything_ with it because there are so many peculiar modes of accidents which /dominate/ that data set.
Obligatory “almost nobody in the US chooses to drive” comment.
Driving in the US is a lifeline. It’s closer to food and shelter than a product or action. Remaining economically afloat in the US without a car is extraordinarily difficult. Many people, especially poor people, would much rather lose their job or health insurance than their car.
The only meaningful way to say is to compare it to other countries. Pr vehicle mile it is a lot more than many Western European countries and Canada, and a lot less than Mexico.
If you have to reach that hard to make your point, it's not a great point.
Adding to the sibling's statistic of 40k deaths a year:
> Motor vehicle crashes were the leading cause of death for children and adolescents, representing 20% of all deaths.
> Approximately 1.19 million people die each year as a result of road traffic crashes.
> Road traffic injuries are the leading cause of death for children and young adults aged 5–29 years.
Falls from a ladder/roof do not come close to that as far as I've been able to find. They'd be a subset of falls from a height, which is a small subset of unintentional falls/slips, which is still globally under road accident deaths.
It's true that diabetes, strokes, heart disease, flu, etc. do cause more deaths, but we're really into the absolute biggest causes of death here. Killing fewer than strokes is the lowest of low bars.
I think there's also the argument to be made in terms of years of life lost/saved. If you prevent a road accident fatality, chances are that person will go on to live many more healthy years/decades. If you prevent a death by stroke, flu, or even an at-home fall, there is a greater chance that person is already in poor health (to have potentially died from that cause) and may only be gaining a few extra months.
But, like curing a dread disease, it's often a long, difficult grind and not something that will for sure work by the end of this year for the last 10 years. No pharma company would get away with that hype.
That's not telling you what you think it is. A lot of those deaths are that person in a car on their own. Usually involving drugs or alcohol. It intentionally folds in "deaths caused by others" and "death caused by self" into the same category. It's not an appropriate statistic to base policy on.
> If you prevent a road accident fatality, chances are that person will go on to live many more healthy years/decades.
Chances are that person is going to kill themselves in a vehicle again later as you have failed to examine MODE of accident. Your analysis is entirely wrong.
Sure - going off of NHTSA figures it looks around 35%. There's also a lot of car passenger deaths (~15%), pedestrian deaths (~20%), and deaths of car drivers with passengers (~15%).
Not entirely sure the point of breaking it out like this, though. These are all still deaths that self-driving cars could in theory prevent, and so all seem appropriate to consider and base policy on.
> Chances are that person is going to kill themselves in a vehicle again later [...]
Unsafe drivers (under the influence, distracted, etc.) are disproportionately represented in fatalities, but that neither means most road accident fatalities are unsafe drivers nor that most unsafe drivers will have a fatal car crash. As far as I can tell, even a driver using amphetamines (increasing risk of a fatal crash 5X) still isn't more likely than not to die in a car crash (a very high bar).
Further, if the way the initial fatal crash was prevented was by prevalence of safe autonomous vehicles, the future crashes would also be similarly mitigated.
It does happen on occasion. Seasonally, sublimating snow banks can create fog that intense for hours if conditions are right. Also heavy smoke can create similar conditions.
If you can visually detect the painted wall, what makes you think that cameras on a Tesla can't be developed to do the same?
And are deliberately deceptive road features actually a common enough concern?
HW4 Tesla stopped before the painting of a road https://futurism.com/someone-else-tested-tesla-crash-wall-pa...