Yes, but how happy are you going to be playing your banjo in the back seat, when you know that x number of people a year are killed by their own autonomous cars, randomly and unpredictably?
You might be the next sacrifice to the gods of the marketplace at any time, through no fault of your own, and there’s nothing you can do about it.
On the contrary, I think that might be their best possible current use. You can structure the vehicle in any fashion most efficient for carrying goods without considering the needs or space of a human bean passenger. You can have truck sized ones to carry big items (or many items), you can have tiny ones for tiny items. You can make them electric because range isn’t a factor (they just run out and then back to the distribution point).
The biggest problem is this but that won’t happen often enough to negate the vast savings companies can realize.
I may have said it before in thus thread, but to me the two biggest future uses of autonomous vehicles are 1) exactly what you said, transport of goods, and 2) transport of people in narrow geofenced areas. There are still problems to be solved in the transport of goods, but highway transport of goods is still making pretty good progress.
I attended a presentation a couple of weeks ago on the transport of cargo by autonomous vehicles, and a point the speaker made was that the shortage of truckers has grown by something like 10,000 every year. Now some of the recent growth might have been driven by the pandemic, but at this point there are roughly 100,000 trucking jobs open. This seems like a place where we could benefit the economy without displacing workers… some of whom will still be needed for convoy trucking, and some who will shift to short haul driving.
Don’t know where you might have gotten the idea that I’m for the things being tested on the open road. As I said, I think the tech’s around 20 years away at the very least, and we’ll have a general purpose AI (something I’m not convinced is possible) driving the thing before it’ll stop making mistakes that a human wouldn’t make if they were paying attention. In short, I’m not getting in the back of one with my banjo anytime soon.
If this is a question of how I feel about deadly products, I smoke cigarettes and drink. They both kill their consumers regularly. I attempt to quit smoking periodically, but I don’t lose any sleep over it - and that’s going to kill me. I’m still pretty happy.
Not to derail this interesting and long-running thread but long distance freight transport may also be the best initial usage of a Hyperloop system. Which I can’t make the numbers work for passenger transport no matter how often I run them in my head.
The actual problem is the endpoint dropoff.
Drone vehicles are perfectly suited to the transport part of the final leg of delivery - from the local distribution center you can send out a horde of drones each to a different target address endpoint, or have loaded trucks taking several targets’ packages along an optimized delivery route. For that they work great.
It’s the “get the package off the truck into the hands of the recipient” that they’re not so good at.
About the optimal outcome I can see happening is if you have drones that know exactly where people’s porches are and swoop in and gently drop the package on their door mat, perhaps with a digital “knock on the door” via easily-overlooked text. This is as close as the drone will be able to get to what human drivers can do, and it’s the worst service we accept from them. You’ll have people watching with binoculars to see where there might be exposed packages to snatch.
And that’s with a drone that can find the porch. A robotruck can do little more than park nearby and sit patiently waiting for the customer to come out and unload it. Or send them a text, wait five minutes, and then bail to where the next customer hopefully waits. Which aligns with an even worse delivery method, where drivers knock, check their watch, and leave a note “sorry we missed you; you have to go to the post office yourself now.”
Maybe at some point in the future residences will all have dropboxes that a truck can back up to, drop in packages, and leave, with the package kept in a secure mailbox that the resident can come back and collect from at their leisure. Of course each dropbox will have to be quite large, to accommodate large deliveries, and they’d have to be kept separate from your actual residence, to avoid criminals mailing themselves (or bombs or packages of live rats) directly into your house. Okay, these things are starting to sound pretty large and inefficient.
For warehouse-to-warehouse delivery, though, drone trucks are perfect. Back up to a shipping container flagged with a destination, take it to its destination, back it into a secure loading bay at the warehouse there, then locate the nearest loaded container (perhaps at that very warehouse) and repeat the process. For this, it’s perfect. For residences without loading bays, less so.
I thik this gets to the reason why people will tolerate 30,000 deaths per year caused by people when they wouldn’t tolerate even a fraction of that by machines.
In short, almost everyone thinks they are a better driver than average, and therefore the accident stats don’t apply to them. But if the machine might kill you randomly, there’s nothing you can do about it. So every time someone is killed in their self-driving car, instead of thinking ‘what an idiot for driving that way’, we think ‘that could have been me’.
As an analogy, there are a lot,of people who will attempt a daredevil stunt that has a 1/6 chance of killing them, because they believe they are in control of the odds through their own skill - even though the one in six who died trying it also felt the same way. The same people would never consider playing Russian Roulette, even for a million dollars, with exactly the same odds. One situation gives at least the illusion of control while the other doesn’t.
Every time you get in a self-driving car you are playing Russian Roulette. The odds may be much lower, but if you get unlucky you die, and there’s nothing in your power to stop it.
It also might be the best initial use of a space warp travel technology, which has about as much chance of existing as a working commercial hyperloop.
But mention of the hyperloop reminds me of this point: Surely the easiest thing to make self-driving is a train. And train automation is a thing, but it’s rare and usually limited to very controlled lines, such as airport trains. The longest automated train in North America is the Skytrain in Vancouver, with a total route of 76 km.
Other train automation systems require an engineer to be present at all times to take over when the train automation can’t handle a problem.
Trains operate in a much more controlled environment than cars, and automated trains are still rare. Automated switching removed the need for cabooses and switchmen years ago, but you still need someone to handle the outliers you can’t train into an AI.
And automated switching has been responsible for some horrific train disasters. But then, manual switching errors create far more (132 switching-related accidents in the US last year), but don’t make the news. But when an automated track switch put two trains on the same track causing an accident it made global news a few years ago. That kind of makes my point about our relative tolerance for human vs computer caused accidents.
By the way, the U.S. has the best, most efficient freight rail system in the world. Primarily because A) it’s privatized, and B) freight doesn’t generally have to share tracks with passenger rail.
I know almost nothing about the train industry but is it possible that automation is rare because there’s not much point? Freight trains are massive and probably run by just a few people; automating it wouldn’t cut costs all that much.
Industrial automation is rarely about cutting worker costs. Usually it’s done for A) quality, and B) repeatabiity. This indirectly leads to lower costs, but primarily for reasons like less waste, less downtime, more overall efficiency. In the industry we call it OEE (Oversll Equipment Effectiveness). People create variable inputs that make efficiency hard.
For example, let’s say you want to exquisitely time your trains such that you can use the track for several different trains without any gaps where no trains are on the track. So you time them to enter the track within the minute. But then the human train engineer is five minutes late, and now the entire plan is screwed. Or, the train should be driving at exactly 75 mph, but with a human controlling it, it varies from 70-80, introducing slop in the schedule that forces inefficient use of infrastructure for every train. If it can all be precisely computer controlled, overall efficiency of the entire system goes up. And those types of savings dwarf the salary of the engineer, who they’d probably keep on anyway for exception handling and emergencies.
This is true for almost all factory automation. Automating a McDonalds kiosk is about saving expensive labor. In industrial systems, not so much.
I think if we can get to the point where we have AI good enough to do true unattended self driving, making a robot that can carry a package to a door won’t be a problem. There are already things like the Boston Dynamics Spot that can do the physical part of walking up stairs, navigating obstacles, and such. What’s remaining is the programming to decide what the robot should do.
I imagine something like the standardized 40’ shipping containers, but sized for individual home deliveries (40"?). The truck pulls up, the delivery dog retrieves the correct container, navigates to the door, drops the container, then knocks and triggers other notifications. Maybe it runs over to a neighbor to collect an empty container, gets back on the truck, and does the next delivery.
I also believe that level of self driving is decades away, and I’m somebody who has driven thousands of miles using the current state of the art self driving available to consumers.
I can’t remember which company it was but I read some place about a plan to have semi-tractor-trailer trucks that would be autonomous on the interstates but switch to remotely driven operation once they get into a city.
Remotely driven? Anyone here ever drive in a city and suddenly lose their cell phone signal? I just don’t think wireless communications systems have high enough reliability for this to be safe.
Maybe Dewey_Finn meant manually driven?
No, I meant remotely driven. I may have been thinking of Starsky Robotics, which closed down last year.
Ah. Still, if the plan was to only use the remote control within city limits, the company may as well just use human drivers, if there are tasks like signing forms and unloading cargo involved.
Consider a company like Walmart which ships massive amounts of goods from its distribution centers to the individual stores. The warehouse workers handle the loading and the store workers handle the unloading. Everything in between could be automated.
This strikes me as somewhere between wildly optimistic and entirely unrealistic.
And the first thing that leapt to mind is what will happen when your delivery dog encounters a real dog - or a jerky kid with a baseball bat.
The same thing RoboCop v1.0 did when he/it encountered the dweeb with the pistol in the boardroom.
We’d quickly have a lot fewer real dogs and jerky kids. Sounds like a win to me.
Delivery drivers today have to contend with loose dogs, and the recipients with porch pirates.
In the future the incident will be recorded on multiple cameras from the truck and the delivery dog. Amazon (the only remaining place to buy things) will instantly file interfering with commerce charges against the jerky kid (face ID and all), file a complaint against the dog’s owner (again, face ID for dogs plus microchip scanning), and black list your address as unsafe for delivery.
Like I said, in a world where computers are powerful enough and programing is sophisticated enough for true self driving, lots of things will be easy. The truck can already detect dogs and kids, because it knows to watch for them and prepare for them to run into the street, so no problem picking them out of a video when something goes bad.