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

An AI dashcam app designed to rate every driver

spectrum.ieee.org

Post reply on HN
Print view View original

61–70 of 109 posts

Re: An AI dashcam app designed to rate every driver

#61

Post by FeepingCreature »

lpage wrote: post #19

Assuming that insurance is correctly priced, buying a policy in any form is negative expected value. The companies that write such policies would slowly bleed dry if that wasn't the case. However, that's irrelevant unless you have the bankroll to withstand tail risk. Given that most people don't, as a society, you pretty much have to make things like auto liability insurance compulsory. If you're a "good driver," as…

> negative expected value

This would only be true if the value of money was linear.

The value of money is not linear.

A 10% chance of a 1000$ loss can be far more harmful than a 100% chance of a 100$ loss, especially at low income.

Re: An AI dashcam app designed to rate every driver

#62

Post by fidget »

noja wrote: post #42
sokoloff wrote: post #11

> Nexar will face some ethical dilemmas. For example, should the app inform users when it spots a license plate that’s the subject of an Amber Alert? Absolutely. I'm not one to normally fall victim to the mindless "think of the children", but Amber Alerts are rare and specific enough that they seem perfectly useful and legitimate usage of this tech. (I find most of the rest of the tech between useless and invasive, b…

Absolutely at the moment . That stops being okay once amber alerts start to be misused.

I mean, it's been 20 years, and people seem to be using them sensibly still. I'm not convinced their misuse is a given.

Re: An AI dashcam app designed to rate every driver

#63

Post by icebraining »

golergka wrote: post #49
frgewut wrote: post #48

Earlier quoted context omitted.

I can imagine a lot of applications for these types of systems: identifying undercover traffic police cars, finding whether someone is following you, identifying friend's cars (similar to FB's "Share your location") People would love such systems (and I'm afraid our privacy is gone for good).

This debate happens on HN again and again, but still: what privacy are you talking about in public space? Nobody has a problem with one person with a dashcam — why suddenly connecting these dashcams to a network that shares information is a problem?

Citing the U. S. Privacy Study Commission,

The real danger is the gradual erosion of individual liberties through the automation, integration, and interconnection of many small, separate record-keeping systems, each of which alone may seem innocuous, even benevolent, and wholly justifiable.

The whole report is public, and it specifically addresses insurance in this context: https://aspe.hhs.gov/report/personal-privacy-information-soc...

Re: An AI dashcam app designed to rate every driver

#64

Post by kabouseng »

golergka wrote: post #54
corobo wrote: post #51

Earlier quoted context omitted.

> shares information Probably that bit

OK, so if I record a particular driver being an asshole on the street and then put this recording on Youtube as an individual, I would be wrong?

Absolutely, since it would amount to trial by media, where a person is found guilty / slandered as an asshole with no avenue to defend his position / state his case.

What if he was driving like an asshole because he was speeding to the hospital or similar reasons?

Re: An AI dashcam app designed to rate every driver

#65

Post by noonespecial »

allanb wrote: post #53
golergka wrote: post #49

Earlier quoted context omitted.

This debate happens on HN again and again, but still: what privacy are you talking about in public space? Nobody has a problem with one person with a dashcam — why suddenly connecting these dashcams to a network that shares information is a problem?

"No one has a problem with strangers seing you on the street, but suddenly when some guy follows you around with a camera it's a problem?" looking -> recording -> sharing These are all different things and our behaviour will and should change at each step.

More than that really. If every stranger who sees you on the street can (or worse, automatically does) instantly share that they saw you to a giant database that can then be easily searched it changes the act of "seeing". Now you can be virtually followed at any time by anyone anywhere who can use the database.

No one can be drowned by a raindrop. A flood is another matter.

Re: An AI dashcam app designed to rate every driver

#66

Post by sokoloff »

noja wrote: post #42
sokoloff wrote: post #11

> Nexar will face some ethical dilemmas. For example, should the app inform users when it spots a license plate that’s the subject of an Amber Alert? Absolutely. I'm not one to normally fall victim to the mindless "think of the children", but Amber Alerts are rare and specific enough that they seem perfectly useful and legitimate usage of this tech. (I find most of the rest of the tech between useless and invasive, b…

Absolutely at the moment . That stops being okay once amber alerts start to be misused.

Here's where I believe "think of the children" madness works in favor of people who think like I do. Law enforcement and the DoJ will be very reluctant to start issuing Amber Alerts for drug dealers, bank robbers, bail skippers, and the like. Even missing elderly get Silver Alerts (where they could have just as easily be lumped into Amber Alerts).

Amber Alerts for my area come to my cell phone and I can't recall getting more than one per year. Looking up stats for 2014, they ranged from 9-21 per month (nationwide). I'm pretty comfortable that surveillance over-reaches will be a long time coming to Amber Alerts (if ever).

Re: An AI dashcam app designed to rate every driver

#67

Post by kwikiel »

FeepingCreature wrote: post #61
lpage wrote: post #19

Assuming that insurance is correctly priced, buying a policy in any form is negative expected value. The companies that write such policies would slowly bleed dry if that wasn't the case. However, that's irrelevant unless you have the bankroll to withstand tail risk. Given that most people don't, as a society, you pretty much have to make things like auto liability insurance compulsory. If you're a "good driver," as…

> negative expected value This would only be true if the value of money was linear. The value of money is not linear. A 10% chance of a 1000$ loss can be far more harmful than a 100% chance of a 100$ loss, especially at low income.

That's called expected utility.

Most people on HN knew about expected value but they don't know too much about expected utility. I suggested reading about it and St. Petersburg paradox

Re: An AI dashcam app designed to rate every driver

#69

Post by thaumasiotes »

schiffern wrote: post #22
Xeoncross wrote: post #16

Earlier quoted context omitted.

You realize that this kind of power will be banned and put under the control of the government only right? We can't have companies violating privacy like this. Only the government can do that.

>Only the government can do that. Why would the government want to spend the money and PR capital developing and rolling that out, when they can have a company do it for free, take the heat, chest beat about private enterprise, and then secretly NSL subpoena or crack it and distribute the intelligence to every interested agency, police department, and eventually favored corporate interest (using parallel construction…

I took it more as a comment on public opinion than on government policies.

Re: An AI dashcam app designed to rate every driver

#70

Post by amelius »

Angostura wrote: post #50
syphilis2 wrote: post #32

I wonder how well this will scale with time. I've certainly done things while driving that were reckless to varying degrees, I suspect most drivers have as well. How long until 50% of cars on the road have been tagged doing something reckless? Of course I can imagine a number of ways to handle this. On a more broad note, I wonder how useful it really is to know that a specific vehicle acted recklessly sometime in the…

I presume that the system would simply indicate the proportion of times that a vehicle is seen driving and performing a dodgy manoeuvre. If I'm seen on the roads 1,000 times and seen performing a dodgy lane change once (and I would like to apologise to the Mondeo driver on the A21 yesterday), that's the fact that should be recorded.

Indeed, a proportion tells us something about the probability of making a driving mistake. A boolean tells us little in comparison.
Post reply on HN