Remix.run Logo
defrost 9 hours ago

It's a bit of a non story, even with the fake image.

From the article:

  Trains were halted after a suspected AI-generated picture that seemed to show major damage to a bridge appeared on social media following an earthquake.
...

  Railway expert Tony Miles said due to the timing of the incident, very few passengers will have been impacted by the hoax as the services passing through at that time were primarily freight and sleeper trains.

  "They generally go slow so as not to disturb the passengers trying to sleep - this means they have a bit of leeway to go faster and make up time if they encounter a delay," he said.

  "It's more the fact that Network Rail will have had to mobilise a team to go and check the bridge which could impact their work for days."
Standard responsible rail maintainance is to investigate rail integrity following heavy rains, earthquakes, etc.

A fake image of a stone bridge with fallen parapets prompts the same response as a phone call about a fallen stone from a bridge or (ideally !!) just the earthquake itself - send out a hi-railer for a track inspection.

The larger story here (be it the UK, the US, or AU) is track inspections .. manned or unmanned?

Currently on HN: Railroads will be allowed to reduce inspections and rely more on technology (US) https://news.ycombinator.com/item?id=46177550

https://apnews.com/article/automated-railroad-track-inspecti...

on the decision to veer toward unmanned inspections that rely upon lidar, gauge measures, crack vibration sensing etc.

Personally I veer toward manned patrols with state of the art instrumentation - for the rail I'm familiar with there are things that can happen with ballast that are best picked up by a human, for now.

hedora 9 hours ago | parent | next [-]

They should already be able to detect line breaks using old technology. They send current pulses down the line to detect stuck switches, since stuck switches can cause collisions. Also, the pulses are conducted through the wheels and axles of any trains, so they can use resistance and/or timing to figure out where the trains are.

Having said that, if it was 2020 and you told me that making photorealistic pictures of broken bridges was harder than spoofing the signals I just described, I’d say you were crazy.

The idea that a kid could do this would have seen even less plausible (that’s not to say a kid did it, just that they could have).

Anyway, since recently-intractable things are now trivial, runbooks for hoax responses need to be updated, apparently.

defrost 9 hours ago | parent | next [-]

> They should already be able to detect line breaks using old technology.

Yes. That doesn't do much to detect a stone from a parapet rolling onto the line though.

Hence the need for inspection.

> runbooks for hoax responses need to be updated, apparently.

I'd argue not - whether it's an image of a damaged bridge, a phone call from a concerned person about an obstruction on the line, or just heavy rains or an earthquake .. the line should be inspected.

If anything urban rail is in a better position today as ideally camera networks should hopefully rapidly resolve whether a bridge is really damaged as per a fake image or not.

ceejayoz 8 hours ago | parent [-]

> I'd argue not - whether it's an image of a damaged bridge, a phone call from a concerned person about an obstruction on the line, or just heavy rains or an earthquake .. the line should be inspected.

Ideally? Sure.

But when someone can generate plausible disaster photos of every inch of every line of a country's rail network in mere minutes? And as soon as your inspection finishes, they do it again?

hedora 6 hours ago | parent | next [-]

Yeah; it’s completely a matter of frequencies and probabilities. Also, technology keeps improving.

If I were working for the train line, and bridges kept “blowing up” like this, I’d probably install a bunch of cameras and try to arrange the shots to be aesthetically pleasing, then open the network to the public.

The runbook would involve checking continuity sensors in the rail, and issuing random pan/tilt commands to the camera.

defrost 7 hours ago | parent | prev [-]

plausibly correlated with what?

This correlated with an earthquake - this is the event that should have triggered an inspection regardless.

> But when someone can generate plausible disaster photos of every inch of every line of a country's rail network in mere minutes?

In the UK (and elsewhere) a large percentage of track is covered by cameras - inspection of over the top claims can be rapidly dismissed.

> And as soon as your inspection finishes, they do it again?

Sounds like a case for cyber crimes and public nuisance.

It's also no different to endless prank calls via phone, not a new thing.

ceejayoz 7 hours ago | parent | next [-]

> This correlated with an earthquake…

Plenty of disasters don't. "No earthquake, no incident" obviously can't be the logic tree.

> In the UK (and elsewhere) a large percentage of track is covered by cameras - inspection of over the top claims can be rapidly dismissed.

"Yes. That doesn't do much to detect a stone from a parapet rolling onto the line though. Hence the need for inspection."

Sounds like you now agree it's less a need?

> Sounds like a case for cyber crimes and public nuisance.

"Sorry, not much we can do." As is the case when elderly folks get their accounts drained over the phone today.

array_key_first 5 hours ago | parent | prev [-]

> It's also no different to endless prank calls via phone, not a new thing.

Of course it's different. If I do 5 prank calls, that takes, say, 15 minutes.

In 15 minutes how many hoaxes can I generate with AI? Hundreds, maybe thousands?

This is like saying nukes are basically swords because they both kill people. We've always been able to kill people, who cares about nuclear weapons?

mschuster91 23 minutes ago | parent | prev | next [-]

> They send current pulses down the line to detect stuck switches, since stuck switches can cause collisions.

That's not done in any European rail network I am aware of. The switches have, well, switches that confirm if the mechanical end positions have been reached, but there is no confirmation by current pulses on the actual rails themselves.

> Also, the pulses are conducted through the wheels and axles of any trains, so they can use resistance and/or timing to figure out where the trains are.

That technology is, at least in Germany, being phased out in favor of axle counters at the begin and end of each section, partially because axle counters allow speed and direction feedback, partially because it can be unsafe - a single locomotive braking with sand may yield a false-free signal when sand or leaves prevent the current passing from one rail to the other.

energy123 8 hours ago | parent | prev | next [-]

If whatever technology they installed said everything was fine, I would still want them to do what they did because the costs of being wrong are so much higher than the costs of what they did.

The point of that technology needs to be to alert you when something is wrong not to assure you that everything is fine whenever some other telemetry indicates otherwise.

qingcharles 8 hours ago | parent | prev [-]

Any idea how the road barriers in the USA detect a train to lower themselves? I assume it's something to do with current passed from one rail to the other through the axle?

When I stuck train wheels on my DeLorean and rode it down the tracks it lowered the barriers automatically which caused a bit of a traffic incident in Oxnard.

doctorwho42 6 hours ago | parent [-]

There are sensor sections on both sides. If you short the tracks together with a large enough wire, it triggers the signal box. Actually learned this at the MIT swap fest when manning the back gate a decade ago. Got some cheap alligator clips and strung to them together, no luck.... Larger gauge copper did trigger it, and confused a ton of people when no train came by lol

lambdas 8 hours ago | parent | prev | next [-]

90% manned. A lot of money and time goes into getting track access.

And collecting unmanned data is still such a pain. At the moment, you stick calibration gear to a train and hope it gets as much noise free data as it can. All whilst going at least 40mph over the area you want - you’re fighting vibrations, vehicle grease, too much sunlight, not enough sunlight, rain, ballast covering things, equipment not calibrated before going out etc etc.

tiew9Vii 8 hours ago | parent | prev [-]

Regardless of how many people it disrupted or not, it’s not a non story.

It’s highlighted a weakness. It’s easy to disrupt national infrastructure by generating realistic hoax photos/videos with very little effort from anywhere in the world.

defrost 7 hours ago | parent [-]

It's not a new story, nor has it highlighted a new weakness - people have had the ability to claim tracks are covered in stone or by a dead cow for a good many years now.

Tracks have cameras to rapidly discount big claims, in this specific case there was an actual earthquake which should (and likely did, the story doesn't drill down very deep) have triggered a manual track inspection for blockages and ballast shifts in of itself.

silversmith 3 hours ago | parent [-]

If I do a prank call, it's easy to see the intent to disrupt.

If I post AI generated images to twitter, and those get amplified by my followers (that might or might not be real people) enough to surface on some rail engineers feed, well, that's just me showcasing my art, no harm intended, right?

defrost 2 hours ago | parent [-]

If I if enough hypothetical if's that's just a giant empty if, right?

It'd be useful if commenters view this from the pragmatic real world track maintainance PoV.

Verifiable calls from the public about blocked lines made to official numbers with traceback etc. carry more weight than social media buzz.

In urban rail the bulk of AI generated images can be discounted via camera feeds and sensors (eg: there's no indication of a line break so that image is BS).

There are already procedures to sift prank calls from things that need checking, to catch serial offenders and numbnuts that push bricks from overpasses.

In the specific instance of you hypothetically "just me showcasing my art, no harm intended" .. in a UK jurisdiction that would fall to the estimation of the opinion held by a man on the Clapham omnibus as channeled by a world weary judge with an arse sore from decades of having such stories paraded before them by indolent smirking cocksures.

YMMV.