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

MTA's A.I. bus cameras issue mistaken parking violations

nbcnewyork.com

Post reply on HN
Print view View original

11–20 of 114 posts

Re: MTA's A.I. bus cameras issue mistaken parking violations

#11

Post by freejazz »

SoftTalker wrote: post #7
meta_x_ai wrote: post #2

If you prevent government from launching technologies for every software bug / mistake, you will get a bloated government which will do everything manually increasing it's cost to taxpayers and decrease in quality of services. The key is still to launch fast and iterate over bugs especially in the realm of traffic / parking violations which at the most causes annoyances to people ( I mean the same people are willing…

Maybe government tries to do too much.

Yeah, perhaps we should just get rid of the roads in Manhattan.

Re: MTA's A.I. bus cameras issue mistaken parking violations

#12

Post by ta1243 »

doctorpangloss wrote: post #10
SoftTalker wrote: post #7

Earlier quoted context omitted.

Maybe government tries to do too much.

Okay. Do you drive? Would you support bounties for ordinary people to enforce parking?

I suspect they want the roads to be private and then have a major corporation with no accountability issue the charges

You can always vote with your wallet and not use that corporation's roads

Re: MTA's A.I. bus cameras issue mistaken parking violations

#13

Post by JumpCrisscross »

doctorpangloss wrote: post #10
SoftTalker wrote: post #7

Earlier quoted context omitted.

Maybe government tries to do too much.

Okay. Do you drive? Would you support bounties for ordinary people to enforce parking?

> Would you support bounties for ordinary people to enforce parking?

Honestly, why not? Let people take a photo of the offending vehicle, picture of the license plate and VIN number and tag the violation. Submit with a copy of your New York ID, and use that to filter out spam and nonsense reporters.

Hell, pay a bounty for submissions that result in the city collecting a fine. (You can deduct bad submissions from said bounties.)

Re: MTA's A.I. bus cameras issue mistaken parking violations

#14

Post by bobthepanda »

devmor wrote: post #3
meta_x_ai wrote: post #2

If you prevent government from launching technologies for every software bug / mistake, you will get a bloated government which will do everything manually increasing it's cost to taxpayers and decrease in quality of services. The key is still to launch fast and iterate over bugs especially in the realm of traffic / parking violations which at the most causes annoyances to people ( I mean the same people are willing…

Seems like the premise you’re going with here is a false dichotomy. There’s no reason this project had to issue parking violations from launch - it could issue soft warnings and get the same feedback from impacted civilians to fix the issues before causing them undue distress.

> The cameras also failed to realize that both the M79 and Bx35 bus routes were still in the “warning” phase of a new enforcement pattern – which means even legitimate infractions should not have resulted in monetary penalties.

Re: MTA's A.I. bus cameras issue mistaken parking violations

#15

Post by JohnFen »

jejfjfjfjsj wrote: post #9
xnx wrote: post #5

Despite the "A.I." in the headline, the source of the errors seems to be bad categorical data input and nothing AI related. The same problem likely could've happened with human traffic enforcement if they were given the wrong instructions.

In some German cities, bus driver simply hit a camera button when actually being blocked. Sounds like a good filter for this use case?

That seems like a far superior solution.

Re: MTA's A.I. bus cameras issue mistaken parking violations

#16

Post by myhf »

meta_x_ai wrote: post #2

If you prevent government from launching technologies for every software bug / mistake, you will get a bloated government which will do everything manually increasing it's cost to taxpayers and decrease in quality of services. The key is still to launch fast and iterate over bugs especially in the realm of traffic / parking violations which at the most causes annoyances to people ( I mean the same people are willing…

No amount of automation will have any significant effect on the size or efficiency of a bureaucracy.

http://www.berglas.org/Articles/ImportantThatSoftwareFails/I...

Re: MTA's A.I. bus cameras issue mistaken parking violations

#17

Post by xvedejas »

jejfjfjfjsj wrote: post #9
xnx wrote: post #5

Despite the "A.I." in the headline, the source of the errors seems to be bad categorical data input and nothing AI related. The same problem likely could've happened with human traffic enforcement if they were given the wrong instructions.

In some German cities, bus driver simply hit a camera button when actually being blocked. Sounds like a good filter for this use case?

This is also how it works in San Francisco (SFMTA Muni buses)

Re: MTA's A.I. bus cameras issue mistaken parking violations

#18

Post by croes »

xnx wrote: post #5

Despite the "A.I." in the headline, the source of the errors seems to be bad categorical data input and nothing AI related. The same problem likely could've happened with human traffic enforcement if they were given the wrong instructions.

Bad training data is pretty much AI related.

People outside AI research expect good answers no matter what you give to AI.

Or to put it that way, wrong instructions are still natural intelligence related

Re: MTA's A.I. bus cameras issue mistaken parking violations

#19

Post by lofatdairy »

I mean I still think this is a good idea. Seems like a combination of poor communication and probably an incomplete understanding of edge cases but illegal parking is a huge problem in NYC and even though there's a snitching program I think the scale can't be addressed without some level of automation. Obviously there are concerns with these sorts of things like misuse of data and whatnot but I don't think we're living in 1984 just because red light cameras exist.

Re: MTA's A.I. bus cameras issue mistaken parking violations

#20

Post by rgovostes »

When I lived in California years ago I had a custom 2-letter plate. I gave the car to a relative who switched its registration to another state.

Years later, I got a toll demand from an express lane around LA that had a false positive scan of my plate number—someone else’s partially obscured maybe?

I argued the car was registered elsewhere and wasn’t driven in CA, and they denied the appeal. I asked for evidence but they refused. I tried a FOIA request but it was denied because there’s an exemption for data that reveals individuals’ driving patterns—to get the photo I would have had to show up at an office in LA to show my ID and registration to prove it was my car, which it wasn’t!

CA law allows the plate scanning and dictates the toll violation letter must show the make and model of the vehicle in question. They just look up the DMV records for the scanned plate. They’re not doing any kind of verification that the vehicle matches. They have 100% faith in the infallibility of these scanners with no appeals process.

Post reply on HN