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

Tesla on autopilot rear ends motorcycle on freeway, killing rider

ksl.com

Post reply on HN
Print view View original

101–110 of 201 posts

Re: Tesla on autopilot rear ends motorcycle on freeway, killing rider

#101

Post by seanhunter »

As a biker myself, I was always taught that when riding a bike you are responsible for your own safety. This means you should always be taking in as much information as you can in order to make escape plans in the event of something unexpected. Getting rear-ended like this is an absolute nightmare scenario. Even though the relative velocity between the vehicles is small, even the slightest bump to your back wheel will almost certainly cause you to hit the tarmac[1], and often on a freeway you have limited escape routes because you may not be able to change lanes safely and if a car comes up behind you fast you may not be able to accelerate away from them.

Autopilot (as I understand it) is a lane-following dynamic cruise control so should have detected the bike and slowed down on approach. Human beings have a weakness in their stereoscopic vision where because their eyes are not far apart they can’t accurately tell the speed of a narrow object (like a bike) if it’s travelling towards or away from them. It seems like this should not be a problem with a car (because you can put the cameras on either side of the vehicle for example), but I wonder whether Teslas detectors or software have weaknesses with narrow objects.

[1] And even though with good gear you can skid along the road relatively safely, other cars are bound to hit you.

Re: Tesla on autopilot rear ends motorcycle on freeway, killing rider

#102

Post by ActorNightly »

Schroedingersat wrote: post #99
ActorNightly wrote: post #12

Earlier quoted context omitted.

This is possibly the dumbest argument against Tesla one could make. From the origin of autopilot from its aviation history, its pretty clear that autopilot was not designed to prevent you from hitting things.

In aircraft or maritime it's a system you can activate and stop paying attention to the controls or do a task like navigation, eating, or going to the bathroom for timescales on the order of seconds to minutes. Tesla's 'autopilot' system is cruise control but with more mental overhead when the human needs to react because the human is being hypnotised by lack of stimulus and has to keep a mental model of what the car…

To just arbitrarily redefine autopilot for cars as something that can avoid accidents isn't a good argument. You are supposed to still have awareness when using autopilot in an airplane, ship - paying attention to things like radar, radio, ADSB, e.t.c. And in a Tesla, you are supposed to pay attention visually.

Full self driving is still in beta test, with present dangers that you have to accept.

All the accidents that happen with Tesla autopilot use are driver error, full stop. Just like its driver error to get drunk and drive, or be looking down at a phone distracted, and rear end a motorcycle. As someone who rides, Id rather have roads full of Teslas with FSD, because for every time it fucks up there are vastly more times for it to avoid an accident, given that you are not going to get rid of drunk, high, or distracted drivers.

Re: Tesla on autopilot rear ends motorcycle on freeway, killing rider

#103

Post by xodjmk »

theplumber wrote: post #20
skybrian wrote: post #13

Earlier quoted context omitted.

It doesn't seem like all that many people are all that crazy about self-driving cars? It's a frequent topic of conversation because some companies have invested a lot of money in it. But people express fairly negative sentiments about it. Meanwhile, huge amounts of money is spent on improving mass transit. It's fairly conventional wisdom that it's a good thing. (The money doesn't seem to go very far though, compared…

>> It doesn't seem like all that many people are all that crazy about self-driving cars? I think that many people(including me) do not belive in self driving cars. I "know" there will be no level 5 self driving car anytime soon so why would I be excited? All the current and "near future" self driving cars seem to have a steering wheel so they are "fake self driving cars" and I'm not interested in that. Why would I us…

"The car is supposed to drive itself based on solved captcha or an army of people trying to label everything??"

You're learning is a little bit off. Self driving cars are not a big truck. It's a series of tubes.

Re: Tesla on autopilot rear ends motorcycle on freeway, killing rider

#104

Post by woodruffw »

pandaman wrote: post #98
woodruffw wrote: post #95

Earlier quoted context omitted.

> The drunk guy about to T-bone you is more difficult to encounter in the first place than the drunk guy about to stab you on the subway. Statistically, he isn't. If drunks stabbed people on subways at the rate that drunks killed people with their cars, around 46 people would be stabbed on the US's subways each day. And that's killed with cars, not "permanently disabled or disfigured"; that rate is even higher. The p…

Could you explain your statistics a bit? The first link I've found[1] shows 50,930 fatalities in 2019 with estimated 19% involving drunk drivers. So 9676 fatalities per year. Let's assume drunk drivers always kill somebody and survive themselves i.e. no drunk driver ever died in a single car accident. This makes 27 fatalities caused by drunk drivers every day. Did you mean that approximately twice as many people ride…

I got 47 deaths per day from the roughly 17,000 fatalities that are directly attributed to drunk driving each year[1]. Even if we arbitrarily halve that to account for at-fault deaths at drivers' own hands, that leaves ~23 people dying, each day, because of drunks on our roads.

In contrast, around 300 people died in total on public transportation in 2020[2]. In 2019, 34 million people commuted daily by public transportation[3].

By contrast, around 76% of American commuters drive to work[4]. Assuming that "commutes to work" is the same as "employed," that means 76% of 158 million[5], or around 120 million driving commuters.

In other words: 4 times as many people commute by car than by public transport, but at least 20 times as many die each day just via the canonical example of unsafety on public transport. And the actual ratio is likely far higher, since the best number I could find for public transport fatalities (under 1/day) is not filtered by accident, suicide, crime, natural causes, &c. Thus the claim: the things that people claim to fear about public transport are far more real as dangers when commuting by car.

[1]: https://troopers.ny.gov/impaired-driving

[2]: https://www.statista.com/statistics/1295843/number-fatalitie...

[3]: https://www.apta.com/news-publications/public-transportation...

[4]: https://www.statista.com/chart/18208/means-of-transportation...

[5]: https://www.bls.gov/news.release/empsit.t01.htm

Re: Tesla on autopilot rear ends motorcycle on freeway, killing rider

#105

Post by pandaman »

woodruffw wrote: post #104
pandaman wrote: post #98

Earlier quoted context omitted.

Could you explain your statistics a bit? The first link I've found[1] shows 50,930 fatalities in 2019 with estimated 19% involving drunk drivers. So 9676 fatalities per year. Let's assume drunk drivers always kill somebody and survive themselves i.e. no drunk driver ever died in a single car accident. This makes 27 fatalities caused by drunk drivers every day. Did you mean that approximately twice as many people ride…

I got 47 deaths per day from the roughly 17,000 fatalities that are directly attributed to drunk driving each year[1]. Even if we arbitrarily halve that to account for at-fault deaths at drivers' own hands, that leaves ~23 people dying, each day, because of drunks on our roads. In contrast, around 300 people died in total on public transportation in 2020[2]. In 2019, 34 million people commuted daily by public transpo…

NY troopers contradict NHSTA (https://www.nhtsa.gov/risky-driving/drunk-driving) by a lot. I wonder how do they gather their statistics?

But your math does not make sense even with the data you have: you said if the drunks were killing in subway at the same rate as they do on the road then 47 people were killed in subway every day yet you took the 47 number from much larger pool of drivers. Should not it be a quarter of that, 12 people?

Your 300 fatalities in public transit could be interpreted as the fatalities caused by the public transit, are you sure they count people stabbed and/or shot waiting for a bus? The page does not show the source and what counts as the death in public transit (it requires some kind of payment for that). If somebody is stabbed while waiting for a bus, does it count?

Re: Tesla on autopilot rear ends motorcycle on freeway, killing rider

#106

Post by Karrot_Kream »

pandaman wrote: post #94
jgwil2 wrote: post #82

Earlier quoted context omitted.

Such a tedious strawman that somehow keeps getting rolled out every time transit and density come up here... no one is saying you "shouldn't be allowed to have a yard of my own," people are saying that you should have the right to build more densely on your property if you want to , which should have the side effect of creating more efficient land use and less expensive housing. It's about changing zoning requirement…

What good is a yard if nothing grows there because it's surrounded by high-rises blocking sun? And not sharing walls is of little comfort when you are adjunct to a 20 unit apartment where 4 units have a party every night and on the weekends - all 20 do. It's not like we have made up zoning just to piss off hipsters. Zoning came up as a solution to the problem of your externalities not stopping on your property line.…

You buy adjacent property instead of insisting that everyone does the same thing with their property as the homeowner growing things? Invoking "externalities" as if everyone wants to do the same thing (growing things) on their property is exactly how you end up with America's suburban monoculture.

Re: Tesla on autopilot rear ends motorcycle on freeway, killing rider

#107

Post by Karrot_Kream »

cozos wrote: post #87
_hcuq wrote: post #51

Earlier quoted context omitted.

So…… 25 sq miles. Let’s assume a 5 mile x 5 mile grid. Let’s put a bus route every half mile. That’s 20 bus routes. Let’s run a bus on each route every 15 minutes…. The buses run at an average speed of say 20 mph. So each bus takes 20 minutes to traverse the route in one direction. You would need 2 buses per route. That’s 40 buses. Let’s say it costs $100 per hour to run the bus. (Googled it). So that’s $4000/hr to r…

This analysis reeks of confirmation bias. You don't need one personal car for every 3 people if you were to comparing busses with self driving cars. You would have a fleet, just like you do with busses. Here is something posted a couple days ago: https://www.npr.org/2022/07/19/1111765630/on-demand-shuttles... 15 minivans replaced 6 busses, at a cost of 1.6 million to 1.3 million per year - but with much better covera…

Fleets of cars would require additional infrastructure being built (specialized parking structures, staging areas, etc) so would increase costs. At this point we're also talking about the theoretical extension (fleets) of a theoretical concept (self-driving cars) so basing this execution in reality seems fraught.

Re: Tesla on autopilot rear ends motorcycle on freeway, killing rider

#108

Post by Karrot_Kream »

mwint wrote: post #81
_hcuq wrote: post #51

Earlier quoted context omitted.

So…… 25 sq miles. Let’s assume a 5 mile x 5 mile grid. Let’s put a bus route every half mile. That’s 20 bus routes. Let’s run a bus on each route every 15 minutes…. The buses run at an average speed of say 20 mph. So each bus takes 20 minutes to traverse the route in one direction. You would need 2 buses per route. That’s 40 buses. Let’s say it costs $100 per hour to run the bus. (Googled it). So that’s $4000/hr to r…

Where'd this $5000/car(/yr?) number come from? Total cost of ownership for a car is far less than that. You can only really get to $5k/year through very high depreciation, and people buying those kinds of cars won't be riding your busses anyway. Also, $100/hr to run a bus is probably a fine operating cost number, but you do have to buy the bus (capitalized cost) and take the depreciation hit just like cars.

Look at [1] and you see it's the average across the US. It's more expensive than that in many states. Unless you're discounting the purchase price/monthly loan payment/lease price.

[1]: https://www.move.org/average-cost-owning-a-car/#data

Re: Tesla on autopilot rear ends motorcycle on freeway, killing rider

#109

Post by elkos »

shepherdjerred wrote: post #44
coldtea wrote: post #17

What matters is the ratio of autopilot vs human driver accidents. And in that area, the highly cherry-picked and curated test results from Tesla sound absolutely convincing! Actually, scratch that: what matter is that the autopilot sounds cool and high tech. Given that, who cares about the casualties?

Can you explain how Teslas data is cherry picked? Is it unfair to compare crash data for just one particular area? Wouldn’t it make a lot of sense that some roads are inherently more dangerous because of poor designs or conditions?

Is the raw data publicly available for public scrutiny?

Re: Tesla on autopilot rear ends motorcycle on freeway, killing rider

#110

Post by ideamotor »

lawrenceyan wrote: post #91
ideamotor wrote: post #15

Earlier quoted context omitted.

To a programmer every problem looks like a software problem. Information technology tends to free ride on public goods, essentially profiting from commons without contributing to them. That’s the feedback loop that is lacking. The negative externality, if you will, that should be brought in. If, let’s say ubers and taxis paid a tax when they rode around without passengers and that tax specifically went to road upkeep…

Pretty easy to complain in your ivory tower when these engineers are actually trying to do something to fix the problems you’re talking about.

My ivory tower? You are definitely going to have to be a lot more specific.
Post reply on HN