Comment by necovek | Hacker News Reader

Comment by necovek | Hacker News Reader

Parent Full thread

As I explained in my another comment on this thread, sometimes they can behave unpredictably — eg. car brakes without a need — causing someone else to crash into you (these are most common crashes I see in the city streets). Perhaps they are not useful for me because I am a very attentive driver (eg. I prefer not to talk on the phone even with the BT hands-free system while driving because it takes my attention away from the road), but they are far away from being all positives.

And how many more cars are going to be produced and bought because of "planned obsolescence" when electronic systems in modern cars start to give up (eg. electric cars and their engines should be able to last for a really long time)? Are they going to keep getting security updates for 20, 30 years? Or will we start seeing people doing something to fool their sensors so they crash?

While I like all the technology, it's short-sighted to think that the alternative is desired simply because people don't care about potentially hurting someone with their cars. Like everything else in life, this issue is multifaceted too.

> I am a very attentive driver

Everyone thinks they're a great driver.

I did not say I am a great driver, but an attentive one. That means that when I drive, I focus on driving. I am sure there are people who can notice the same things I do while paying less attention on the road: that would make them "better" drivers.

The most attentive driver in the world cannot prevent an inattentive driver from T-boning them. Automatic braking can.

There is nothing automatic braking will do to help you from getting T-boned. In fact only an attentive driver can prevent that one.

Automatic braking will prevent you from rear-ending the car in front, yes. Although an attentive driver should also be doing that.

> Although an attentive driver should also be doing that.

Society doesn't allow people to say that they're good enough to opt out of safety mechanisms designed to protect us all.

This is why 'I don't wear a seat-belt I can brace myself' also doesn't cut it.

"safety mechanism"

A seatbelt is a safety mechanism.

Driving assist functionalities encourage unsafe behaviors of drivers more than safety. Why should I pay attention to the road if my car will automatically brake for me? Well, until the one time it doesn't and someone dies.

> Why should I pay attention to the road if my car will automatically brake for me?

Was literally an argument people made against seat belts.

Absolutely nobody made an argument that they don't want to use seatbelts because then they'll just start crashing all the time.

(I'm old enough that seatbelts weren't a thing when I was a kid, so I remember well the arguments when they became common and later required.)

Tons of people thought seatbelts would do more harm than good.

https://www.pennlive.com/opinion/2020/08/people-once-complai...

> trapping people in burning cars or causing drowning or preventing rescue personnel from being able to free those who are injured

> belts cause internal damage when an accident causes the wearer to slam forward

> folks wear them incorrectly, across their stomachs rather than their waists, where they will do more harm than good

After reading the Superfreaconomics chapter on child seats I can believe seat belts do quite some harm in accidents.

There certainly can be, and have been, surprising drawbacks from what would've seemed like a clear win safety technologies.

For example ABS increased crash rates for a good while when it was new, contrary to all expectations.

From this NHTSA report: https://crashstats.nhtsa.dot.gov/Api/Public/ViewPublication/...

"Analyses of data from the early 1990s showed significant increases in fatal run-off-road crashes with ABS, on the order of 28 percent. The increase was baffling, given the success of ABS on the test track. However, at that time, many drivers did not yet know how to use ABS correctly."

I observed this effect first hand, because from the late 90s to the late 00s I was frequently a car control clinic instructor. One of the exercises of these clinics was to teach panic braking. In the earlier years most people showed up with non-ABS cars so this was a great exercise in teaching proper braking control.

As ABS cars became more and more prevalent we (the instructors and the school) thought that this exercise can go away since there is no skill to learn braking if you have ABS, just mash the pedal.

Alas, no! It is difficult to believe, but approximately nobody that came to these car control clinics with an ABS car was actually able to initially brake hard enough to even activate the ABS. Sometimes it took many many runs through the braking exercise to finally get it: just press hard on the brake pedal.

Yes, it will. If someone isn't paying attention then they will T-bone you when they drive a car without automatic braking and will not when they drive a car with automatic braking.

I see, in that sense yes. If the tboners car has autobraking, it can help. I read it initially as the tbonees car having autobraking, which won't help at all (since there is nothing in front of the tbonees car at any moment for it to sense).

Although, I'm dubious there would be enough time in that scenario to help much. The typical case is someone running a red light and intersecting a car from the other direction, both going full speed. By the time the tbonee car enters the field of the tboner car front sensors they are very close and less than a second from impact. So it'd reduce speed a bit which is good, but won't prevent the hit.

An attentive driver on the other hand, can see far wider than the sensors and has more opportunity to avoid the incident entirely. If paying attention, of course.

I doubt automatic braking would help a t-bone situation. Do you have evidence of this?

Meanwhile, any driver worth his salt should be checking cross traffic when nearing any intersection. I've avoided several accidents by examining potential cross traffic at intersections.

Accident avoidance should be drivers ed 101, but for some reason, especially in these conversations, no one seems to know how to look for and avoid potential accidents.

Autobreaking typically only cares about what is in front of the car correct? There are possible scenarios where an attentive driver may opt for a frontal collision to avoid a T-bone that could kill them.

There are arguments to be made both ways, but there definitely isn't a clear solution.

I am guessing you mean automatic braking in inattentive driver's car, because automatic braking in attentive driver's car works with mostly the same data as the driver themselves (it needs to predict collision, and to predict it, it needs to see the approaching car, estimate the speed, likelihood it will stop...).

Automated system would have advantage there only where it can better "see" things (eg. LIDAR in pitch dark conditions, or maybe positioned at the roof of a vehicle and seeing over other vehicles that a driver can't).

My point was that for attentive drivers, "safety features" are more a nuisance than help in their own car (the worst you can do is have frustrated drivers behind the wheel). If that results in drivers disabling them, we are not better off at all: technology does not get developed and never matures enough not to be a nuisance. I keep them enabled, but I can understand that not everybody wants to go through these rough edges.

I recommend watching TeslaCam on YouTube - there are a few videos where the AI reacts faster than any human could, even a very attentive one.

It's by no means perfect (and the city beta is kind of a mess), but the highway safety bits are very functional and useful.

FWIW, I've checked a few near-misses, and I can't find anything that looks super confidence inspiring. Sure, it might be better than those not paying attention at all, but while that's a win on average, it should beat the most attentive and the most capable drivers before we consider the tech top-notch.

Reaction time in all the car-braking studies is usually measured at 0.2-1s.

That's indeed the time a computer can use to their advantage, and maybe Teslas do, but many in-car systems today are not that fast, and esp not the one in my Volvo. All the things it does it seems to do them slightly later than I'd like to for a seamless, defensive drive.

It's all a matter of cost of per-unit hardware (processing power), development time and state of the art.

We'll get there, but we are not there today.

Related, every car company will think it’s software is perfect. In fact they’ll insist on it, because otherwise they’d be treated as responsible should something go wrong.

Reply on news.ycombinator.com