[tachyonbeam]

> Then what? You have 20 palettes of dog chew toys or asian pears....? Some of the goods have serial/tracking numbers.

Your comment made me laugh. Very true. Besides the fact that there is no black market for 20 palettes of dog chew and asian pears, the other problem is that unless you successfully steal a few highly valuable items, you would need one or multiple trucks to carry what you've stolen from that semi. You probably don't want to use the semi you just stopped, because it could have multiple GPS trackers onboard.

If you knew that you were stopping a truck full of laptops, then maybe it would be worth it somehow, but as you pointed out, laptops have serial numbers... And then what, you load a few hundred laptops into a van, but your van is now "burned". It's been photographed and you need to dump it somewhere, further complicating your operation.

Then, I don't know, it seems to me like people just love to come up with imaginary reasons why self-driving cars/trucks can't work. They seem to fail to realize that, well, we can come up with even better safety measures. You could make your automated truck very hard to open. It doesn't need to have a lock that can be opened by a human with a physical key. It can have an electronic lock inside the door, shielded behind a 5mm thick steel plate, that's completely invisible from the outside.

You could also install a remote-controlled drone on top of the truck (value < $1000). The truck has cameras all around that record continuously, and as soon as the truck gets stopped (or even slightly before), it phones home. A remote operator sees everything the cameras saw. The thieves waste precious time opening and unloading the truck, and just when they're about to take off, the remote controlled drone starts up and follows them around for as long as its batteries will allow (~10-20 minutes), informing the authorities as to their position.

If robberies of automated trucks became rampant, I'm pretty sure we could come up with many ways to mitigate the problem. I mean, heck, we could even install pepper sprayers around the truck. But the robbers can just wear full-face masks, you say. Sure they can, but those masks aren't foolproof, and it's an additional piece of logistics they need to deal with.

[gwern]

> Then, I don't know, it seems to me like people just love to come up with imaginary reasons why self-driving cars/trucks can't work.

The way I like to put it (https://www.gwern.net/Complexity-vs-AI#technology-forecastin...) is: 'The critic asks “can I think of any reason this system might not work?” and stops as soon as they find one excuse, but the forecaster needs to ask, “can I think of any system like this which could ever work?” and keep going.'

[riversflow]

I don’t think this is a fair characterization of adversarial attacks on AI. People on hacker news are simply forecasting the future of hacking. People aren’t really saying ‘it won’t work, forget about it!’ they are simply pointing out that historically technology and criminal enterprise have followed a similar pattern to the garter snake and the rough skinned newt[1]. The more sophisticated deterrents evolve, the more clever adversarial strategies become.

[1] https://www.atlasobscura.com/articles/how-a-deadly-camping-t...

[gwern]

> The more sophisticated deterrents evolve, the more clever adversarial strategies become.

And yet, not clever enough to be so successful as to stop automation, or economic development, or increasing global wealth, or...

[starlust2]

You'd also have to dodge every other vehicle in the fleet, which are all technically drones.

[0x8BADF00D]

It’s relatively trivial to attach a shock baton or flash bang ordnance to the drone, which would incapacitate the thieves as well.

[CPLX]

> seems to me like people just love to come up with imaginary reasons why self-driving cars/trucks can't work

You do have a point. The core reality that self-driving is equivalent to full artificial intelligence and passing the Turing test should be sufficient I'd think.

[tachyonbeam]

I believe in incremental improvement. We can design vehicles that are increasingly more autonomous. They might never become completely autonomous (case in point, they will likely always need someone to come and repair them when damaged), but they can become autonomous enough to be useful. We're not far from the point where autonomous vehicles can become useful in specific scenarios, such as always following the same path on the same stretches of highway.

[pault]

The thing I've always wondered about is: if governments can put street signs for humans on every road, why can't they do the same for autonomous vehicles? Sure, you could vandalize the signs and wreak havoc, but you can also remove a stop sign or hack a traffic light and do the same now. If a standards committee was formed to develop a spec for autonomous vehicle guides it seems like we could get to full autonomy far faster than waiting for AGI. Maybe you still have to drive on backcountry dirt roads, but wouldn't automating 90% of traffic be an enormous win for society?

[lmm]

Partial automation might be worse than no automation at all - if the human is forced to take over they'll be less prepared for it.