Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Driverless cars are stuck in a jam

economist.com

Post reply on HN
Print view View original

91–100 of 556 posts

Re: Driverless cars are stuck in a jam

#91

Post by _xnmw »

sliken wrote: post #85

Tesla claims that autopilot is involved in one accident for every 4.34 million miles driven. Seems like a useful number, of course the media reports that the average is 1 accident per 500,000 miles. So 9 times safer... right? Question is how many accidents do similar people (read that as rich enough to buy a $50k+ car) in similar conditions (highways) with similar cars (newish luxury cars) without autopilot? No fair…

I think that is an incredibly misleading statistic. For it to be comparable to human driving, autopilot needs to be turned on all the time, no exceptions. I think most people only turn on autopilot under ideal road conditions, so you would need to compare the human driver accident rate under perfect highway road conditions.

Re: Driverless cars are stuck in a jam

#92

Post by mr__y »

sliken wrote: post #89
davidgould wrote: post #88

Earlier quoted context omitted.

Surely some of those autopilot miles involved drunks, so they shouldn't be excluded entirely.

Indeed, but likely at a lower rate than the general public. In any case the general idea is to find as fair a comparison as possible with the Tesla published numbers. I suspect $60k BMW owners driving on the highways in good weather are much safer per mile than random folks in random cars on random roads in random weather.

while I generally agree with your point, that new cars with great braking, enhanced safety and in perfect technical conditions are safer than average, I'm not entirely convinced if BMW is the right make for this kind of examples

edit: typo

Re: Driverless cars are stuck in a jam

#93

Post by _xnmw »

zrm wrote: post #64
goatinaboat wrote: post #34

Earlier quoted context omitted.

The driverless bar for safety is way higher than very clever driver assistance or even rocket science. One bad driver kills themselves and maybe a handful of others. One bad software update kills... everyone. So, the demand that the software be near-perfect is not at all unreasonable. It’s strange not to acknowledge this fundamental difference.

Problems like that have technical solutions. You phase in updates over time. A hundred vehicles a day for the first week after release, then a thousand a day, then ten thousand a day and so on. That way if there is a serious bug, you find it when it's running on hundreds of cars rather than hundreds of millions.

Aside from the suggestion of beta testing on human lives, you are assuming that your phase-deployment mechanism always works correctly. Even Google has previously botched phased deployments and slow rollouts.

Re: Driverless cars are stuck in a jam

#94

Post by ghaff »

macspoofing wrote: post #73
crocal wrote: post #8

The article misses an important point. It’s not only a problem of technology, it’s also a problem of risk acceptance. While we accept to be killed by our own mistakes at the wheel, society (we) will not accept that a machine makes the same, even if the failure rate is on average /better/. This is the fundamental issue Elon Musk has. The driverless bar for safety is way higher than very clever driver assistance or eve…

>even if the failure rate is on average /better/ Which has not been true thus far. That's the problem. > The driverless bar for safety is way higher than very clever driver assistance or even rocket science. Why shouldn't it be higher? Why shouldn't it be on the level of airline travel?

>Why shouldn't it be on the level of airline travel?

In all fairness, commercial airlines operate in an environment that's mostly highly regulated with respect to maintenance, training, traffic management, etc. But I agree with your general point. After all--except arguably with respect to some drug side effects--there isn't really any consumer-facing product that regularly kills people and we shrug out shoulders and say "Hey, stuff happens" and get on with our day.

Re: Driverless cars are stuck in a jam

#95

Post by _xnmw »

listenallyall wrote: post #65
theieindjri wrote: post #38

Earlier quoted context omitted.

I don't understand why so many otherwise smart people fell for FSD hype. I'm not talking about the many con artists - I'm talking about the tech-savvy people who genuinely bought into FSD hype. How is it not obvious that FSD most likely requires artificial general intelligence and is, thus, probably somewhere between "many decades away" and "humans will never achieve it"?

I agree and think a lot of smart people still don't really comprehend how non-linear the task is. In other words, they believe if a self-driving car handles itself properly in 95% of situations, the final 5% is right around the corner, when in fact the final 5% (often the difference between life and death) may, in all likelihood, take longer to get right than the previous 95%. It's just that much harder to be perfect…

Furthermore, even 99.99999% reliability is not good enough, and that's already incredibly hard to achieve. Five 9s of reliability still means 5 minutes of unreliable behaviour per year, which is unacceptable in the context of driving.

Re: Driverless cars are stuck in a jam

#96

Post by petilon »

Tesla is already selling "Full Self-Driving Capability" for $6000.

Excerpts:

Coming later this year:

Recognize and respond to traffic lights and stop signs. Automatic driving on city streets.

Fine print includes this:

The activation and use of these features are dependent on achieving reliability far in excess of human drivers as demonstrated by billions of miles of experience, as well as regulatory approval, which may take longer in some jurisdictions.

https://www.tesla.com/model3/design#autopilot

Re: Driverless cars are stuck in a jam

#97

Post by goatinaboat »

_xnmw wrote: post #93
zrm wrote: post #64

Earlier quoted context omitted.

Problems like that have technical solutions. You phase in updates over time. A hundred vehicles a day for the first week after release, then a thousand a day, then ten thousand a day and so on. That way if there is a serious bug, you find it when it's running on hundreds of cars rather than hundreds of millions.

Aside from the suggestion of beta testing on human lives, you are assuming that your phase-deployment mechanism always works correctly. Even Google has previously botched phased deployments and slow rollouts.

Or that a nice phased rollout over the summer doesn’t fail as soon as leaves fall or it rains

Re: Driverless cars are stuck in a jam

#98

Post by sliken »

_xnmw wrote: post #91
sliken wrote: post #85

Tesla claims that autopilot is involved in one accident for every 4.34 million miles driven. Seems like a useful number, of course the media reports that the average is 1 accident per 500,000 miles. So 9 times safer... right? Question is how many accidents do similar people (read that as rich enough to buy a $50k+ car) in similar conditions (highways) with similar cars (newish luxury cars) without autopilot? No fair…

I think that is an incredibly misleading statistic. For it to be comparable to human driving, autopilot needs to be turned on all the time, no exceptions. I think most people only turn on autopilot under ideal road conditions, so you would need to compare the human driver accident rate under perfect highway road conditions.

Well that's not realistic... yet.

But it would be interesting to see how safe people are without autopilot in similar close to ideal conditions with a close to ideal car.

Re: Driverless cars are stuck in a jam

#99

Post by 2sk21 »

deepnotderp wrote: post #4

The most pertinent point here being that although collecting lots of self driving car data is easy, the tail and unexpected events that you care about are surprisingly rare. And many of these events can also be entirely new (never seen before). Current AI SOTA techniques simply require too much data and cannot reason well enough to solve this. Leaving Musk aside, amongst those who are serious about self driving cars,…

If only there was some way to compute entailments from unlikely combination of events :-) But seriously, Gary Marcus and Ernie Davis make a good case in their new book Rebooting AI that logic has to play a role in any real AI.

Re: Driverless cars are stuck in a jam

#100

Post by sliken »

mr__y wrote: post #92
sliken wrote: post #89

Earlier quoted context omitted.

Indeed, but likely at a lower rate than the general public. In any case the general idea is to find as fair a comparison as possible with the Tesla published numbers. I suspect $60k BMW owners driving on the highways in good weather are much safer per mile than random folks in random cars on random roads in random weather.

while I generally agree with your point, that new cars with great braking, enhanced safety and in perfect technical conditions are safer than average, I'm not entirely convinced if BMW is the right make for this kind of examples edit: typo

Heh, ok, pick any luxury car priced similarly to the Tesla.
Post reply on HN