That is actually an interesting question. We track how many accidents the automated cars cause, but I'm not sure how many accidents they prevent are tracked.
As a human driver, I'm keenly aware of the "close calls" I've had where I've narrowly avoided a probable collision through good decisions and/or reaction times. I'm probably less aware of the times when I've just gotten lucky.
No doubt self-driving companies have internally available data on this stuff. Can't wait til superior performance to human drivers is a major selling point!
Just a note that this number is hard to calculate accurately with an acceptable degree of certainty.
Anyone claiming that AVs are involved in fewer accidents per distance traveled than human drivers is either extrapolating from incomplete data or baking some unreliable assumptions into their statement.
Welch Labs has a good introductory video to this topic: https://youtu.be/yaYER2M8dcs?si=XEB4aWlYf6gnnTqM
Applying some basic statistics to show to a 95% confidence level that a self driving system causes fewer fatalities than a human you would have to drive 275 million autonomous miles flawlessly.
This would require a fleet of 100 vehicles to drive continuously for 12.56 years.
And in practice self-driving vehicles don't drive perfectly. Best estimates on the actual number of miles driven to validate their safety is around 5 billion autonomously driven miles and that's assuming that they actually are safer than a human driver.
Then you get into the comparison itself. In practice AVs don't drive on all the same roads, at the same times, as human drivers. A disproportionate number of accidents happen at night, in adverse weather, and on roads that AVs don't typically drive on.
Then you have to ask if comparing AVs to all drivers and vehicles is valid comparison. We know for instance that vehicles with automated breaking and lane assist are involved in fewer accidents.
Then of course if minimizing accidents is really what you care about there's something easy we could do right now: just mandate all vehicles must have an breathalyzer ignition. We do this for some people who have been convicted of DUI but doing it for everyone would eliminate a third of fatalities.
In a similar vein, if we put geo-informed speed governors in cars that physically prevented you from exceeding, say, 20% of the speed limit, fatalities would also likely plummet.
But people haaaaate that idea.
If it was 100% accurate and you couldn’t get a speeding ticket if it was active I’d be all for it
I initially read this as "20% of the speed of light", and though you were being sarcastic.
Because what you actually said is true too, and hints at why "it would be safer" is not a good enough reason to implement something.
I'd greatly prefer just automatic enforcement of speeding laws rather than doing things to try and prevent people from speeding.
The toll roads could trivially automatically enforce speed limits. They already precisely know when each car goes through each gantry, they know the distance between each gantry, so they know everyone's average speed.
[1] https://digitalassets.tesla.com/tesla-contents/image/upload/...
Which for Tesla's FSD is obviously not the case.
https://www.motortrend.com/news/tesla-fsd-autopilot-crashes-...
Has FSD had a fatality? Autopilot (the lane-follower) has had a few, but I don't think I've heard about one on FSD, and if their presentations on occupancy networks are to be believed there is a pretty big distinction between the two.
With human drivers -- are we blaming Tesla for those too?
You do you, but I'm here to learn about FSD. It looks like there was a public incident where FSD lunged at a cyclist. See, that's what I'm interested in, and that's why I asked if anyone knew about disengagement stats.
> https://static.nhtsa.gov/odi/ffdd/sgo-2021-01/SGO-2021-01_Da...
The Reporting Entity’s report of the highest- level driving automation system engaged at any time during the period 30 seconds immediately prior to the commencement of the crash through the conclusion of the crash. Possible values: ADAS, ADS, “Unknown, see Narrative.”
Stuff people made up is a bad reason to blame a company.
> The new data set stems from a federal order last summer requiring automakers to report crashes involving driver assistance to assess whether the technology presented safety risks. Tesla‘s vehicles have been found to shut off the advanced driver-assistance system, Autopilot, around one second before impact, according to the regulators.
[1] https://www.washingtonpost.com/technology/2022/06/15/tesla-a...
Especially because the first sentence you quoted strongly suggests they do get counted.
Nah, that would be illegal. Their statement leaves plenty of room for dirty laundry though. I'm sure they won't disclose disengagement data unless forced, but they have plenty of legal battles that might force them to disclose. That's why I'm asking around. I'd love to rummage through. Or, better, to read an article from someone else who spent the time.
Musk has violated many rules regarding investors.
Yes, I'm addressing that point directly, specifically the fact that this "unthinkable number" is misleading regardless of the number's magnitude.
If I had claimed that the 500 million number indicated a certain level of deaths-per-mile safety, that would be invalid -- but I spent 3/4 of my post emphasizing that it did not, even though you keep pretending otherwise.
California companies with real self driving have to count their disconnects and report all accidents, however minor, to DMV. You can read the disconnect reports online.
As an aside, public acceptance of driverless cars is another story. A story filled with warped perspectives due to media outlets stoking the audience emotions which keep outlets in business — outrage, fear, and drama. For every story covering an extreme accident involving human drivers, there might be 100 stories covering a moderate driverless accident. No matter how superhuman autonomous cars perform, they’ll have a massive uphill battle for a positive image.
Further, most AVs explicitly disengage when they encounter situations beyond their understanding. So again, we’re selecting AV statistics based on generally favorable conditions, but we don’t track human-operated miles the same way.
It’s not really a fair comparison.
What do you mean by disengage? Cruise’s AVs don’t have a driver to take over.
But also, there are explicitly times and areas where Cruises don’t operate due to being insufficiently able to operate. And times where they do just pull over and wait for human intervention when they don’t know what else to do. Both of which are safe and reasonable, but which also selects themselves into a safer set of miles actually traveled.
the bigger problem here is that the machine may not realize it isn't fully understanding the situation; I think that's the more common situation : a computer situational model doesn't match reality, but the machine has the confidence in perception to proceed, producing dangerous effects.
[1] It tends to be more conservative than human drivers, and in weird ways. If a pedestrian seems at all to be moving in the direction of traffic, even if they're on a sidewalk and just meandering a bit, the car will try to evade (not violently, but it will brake a bit and pull to the outside of the lane).
I haven’t had a phantom breaking issue in a long time either, I’m not sure if this is because of the FSD package or if the underlying autopilot system has improved.
The question is how many accidents (if any) self driving prevents with respect to average human drivers.
They will not function in high grass, I guess from my experience with different parktronics.
The best "off road" demonstration of self-driving and/or AI assistance I've found is this: https://media.jaguarlandrover.com/news/2016/07/jaguar-land-r...
Note they avoid going into grass. Human can deduce trails from grass profile, AI? I don't think so.
Will you count an inability to reach a lakeside as an accident? I guess, no.
AIs need training before they can do things.*
*They just might be learning to do things without being trained based on the emergency behavior I see in LLMs.
As for "emergency behavior" (emergent behavior, I guess) - we do not know how LLMs are trained. Thus, what you consider "emergency behavior," could very well be a part of training.