My sister and her husband took the train from Colorado to Massachusetts at Christmas and loved it. Didn’t cost more than a plane, and while it took two days it was a pleasant two days seeing the country.
Apparently it only takes 6 gallons of fuel to move one ton of freight coast-to-coast via rail.
Good question. The obvious candidates would be either the car’s human occupant(s), or its manufacturer. I don’t know if this has yet been worked out fully, or whether it’s a work in progress, but clearly there needs to be a framework in place before cars can become fully autonomous.
Even if the unaccountable, proprietary software consistently produced better results (in terms of road mortality and injury) than human drivers?
Ok, if you believe that the issue is that AI will never be good enough to drive a car as safely as a human, fine. Time will tell. But the quote I was referring to and had issue with was specifically talking about the limits of SENSORS being the reason, which sounds to me like it was saying that sensing hardware is a potentially insurmountable problem. As an engineer who lately has been working with self-driving vehicles (in a controlled environment) on a daily basis I just don’t believe that’s the case.
In Germany, there are laws to ensure the train schedules and bus schedules work together, all over the country.
Try 25-30 years.
By which time the roads will have deteriorated into cart tracks
It,s a lot easier to take whatever food and drink you fancy on to a train
The cost of labor rises all the time, as living standards improve. Any form of transport which involves employing people for longer than necessary will be very expensive. This is why surface travel on ships used to be economic, and why supersonic transport will eventually be the best way to fly.
Long distance rail may work if the trains require no operators, but somebody still has to maintain the infrastructure.
So you list a bunch of “fallacies” in my comment, and then just JAQ off for a few paragraphs. Cool. Let me just answer some questions.
They are perfected when they have the same reliability as anything else we rely on regularly. Nothing is perfect, but autonomous vehicles will be appliances. As for the obligations of manufacturers, well, you’re right, we’ve never had a technology enter the market that required the cooperation of manufacturers, insurance companies, and the government to ensure safety for users. Ever. And your continued use of loaded phrases like “vehicular murder” really make me question your biases.
Even if the rise of autonomous vehicles leads to the actual number of vehicles on the road increasing, less roads are required because 1) in any sane model of autonomous vehicles, they operate in sync with each other, reducing if not obviating traffic, 2) autonomous vehicles will likely be smaller and lighter, being on-demand appliances, and 3) insurance companies will likely disincentivize people from owning and driving their own cars.
Because no one has ever figured out how to harden radio signals from interference. Yes, I’m positing a lot of technology infrastructure that has to come along with autonomous vehicles. I’m not expecting this to happen in my lifetime. Also, I know you’re trying to score points, but Elon’s a brilliant idiot. I can think for myself.
Designing an autonomous vehicle doesn’t happen at highway speeds, where split-second decisions and actions are required. Which human brains/bodies aren’t great at. Please point out where I assigned superior morality to computers. I’m talking about action, not judgment. It is possible to design an optimum algorithm. Will that algorithm be responsible for human death? Probably. Will vehicle deaths go down once reaction time is reduced to several milliseconds instead of hundreds? Definitely.
Hey, it’s almost like I’m envisioning a world in which people don’t need to own their own vehicles, where all vehicles are autonomous, and you can just step into one when you need it. It’s almost like you have no idea what my vision of public transportation is like, and are happy beating up a bunch of strawmen. I mean, whatever floats your boat.
I guess none of the issues you raise will ever be solved. We will never have an infrastructure that relies on autonomous vehicles, because a couple of people said something “may never change”. Oh, well.
And that’s assuming the music doesn’t keep you awake.
No, you don’t understand, we can never solve these problems. All software ever created is unaccountable and proprietary. Regulation and open source doesn’t exist. Anyone who thinks a computer can react faster than a human is crazy, and deserves spite and derision. In the Comic Book Guy voice, which is what I keep hearing.
I’ve taken the London-Scotland sleeper a few times before, and it is a decent alternative to an expensive hotel room for the additional night. The thing is, the distance isn’t really far enough to require a sleeper. Edinburgh to London is only 4 hours, and there are over 30 trains a day in both directions, so there will be one going when you need it.
https://www.seat61.com/london-to-edinburgh-by-train.htm
The slow overnight service makes more sense on the Highland routes to Inverness and Aberdeen, which take around 7 hours anyway.
Also, there was supposed to be a sleeper service through the channel tunnel to link mainland Europe to destinations throughout the UK, but a combination of the rise of cheap flights and a lack of government support killed it, and the proposed direct Eurostar services to not-London.
Sadly, even if this new service from Malmö starts, you’re still not going to be able to go to sleep in Sweden and wake up in London, as the proposed service only goes to Cologne, and you’ll have to change trains twice to get to London later that morning via high speed rail (Cologne - Brussels, then Brussels - London)
This illustrates that what Europe needs is more integrated cross border services to make high speed rail work across the continent, and better interchange with ferries as well.
DH and I like Amtrak, although the experience was somewhat the worse for wear last time we got a sleeper.
We took the Coast Starlight from Seattle to LA. Spectacular scenery and we never noticed ‘stale’ air; you also had the option of stepping off the train briefly at certain designated stops. We indulged and got the self-contained suite with a built-in shower, sink and toilet, which also had a wider bottom berth that could conceivably fit 2. In comparison, the smaller roomette was more comfortable IMHO, and it wasn’t worth the extra for your own bathroom. Our steward asked if we wanted to leave the bunks down, which was nice because first I had back trouble, then DH got sick; we alternated laying down. It was probably less work for them, so a win-win.
Meals and the Viewliner cars were fun; we met all types of people by random assignments, or wherever vacant seats popped up. There was always more freedom of movement and bigger, more comfortable seating no matter where you were on the train, compared to a non-first-class plane seat.
I do agree that Amtrak made a devil’s bargain with the freight train companies that own the railways, which does result in delays when you have to get out of the way of a freight train. If I had the time, I’d still rather take a train over sardine-class airlines for medium-long hauls.
Yep, still not OK arrogating lethal agency to algorithms even if they produce better outcomes so long as those algorithms are private and proprietary. Open auditing, transparency standards, and strict testing regimes would be improvements that I’d be more OK with.
Enh, probably even then. There’s a problem with the way we think about the success rate of learning machines. We don’t distinguish between a 99% success rate that has a 99% chance of success for everyone from one that has a 100% chance of success for 99% of people and 0% chance of success for 1% of people.
All things with senses are capable of misperceiving and hallucinating. The trouble with algorithms is that 1000 cars that run the same algorithms may be all vulnerable to misperceiving and hallucinating in exactly the same way under exactly the same circumstances.
So our assessments of threats usually build in our confidence in how success and failure of a system is naturally distributed. You take a road where there is a 1% chance of a serious accident on a given day. You put in a blizzard condition and we expect the chance of an accident to go up, but we see it as a bunch of people having their skill tested under harder conditions and thus more fail.
Human skill is black line, computer skill is purple. We test and test under green conditions and fallaciously extrapolate from that to red conditions. We try to test under red conditions (the blizzard) but we don’t know what they really are because the things that might mess with the computers’ sensors aren’t the same things that mess with human sensors.
I think there is a deep, fundamental problem with applying our current ideas around learning algorithms to the real world that we haven’t gotten our heads around yet. I read an article by a person with schizophrenia once about working. How when they see something strange the first thing they ask themselves is whether it is possible, the next is whether anyone else is reacting to it. By doing that they are able to tell, for example, that the cloud of fireflies in the middle of the boardroom is just a hallucination (it’s possible but surely someone else would react to it). Without similar checks I don’t trust computers to make important decisions, but those checks cost us in terms of accuracy, and we are currently measuring for accuracy only.
Cool, more fallacies! You have a lot more faith in technology companies acting in good faith and regulators acting empirically and proactively than I do.
What term would you prefer besides “vehicular murder”? Do you prefer the cleanliness of “collateral damage” or “negative externalities”? Giving lethal autonomy to algorithms without human oversight is kind of a Big Deal. Handwaving technological fixes for it doesn’t diminish its import and computers being faster doesn’t solve everything.
Cool, more strawmen! Because I totally advocated for just handing over control to half-baked software with no human oversight. And I totally said all vehicular problems will be solved by computers.
I know it will be hard. Which is why I said so. And so you know, I’m bored, and we’ve derailed things enough already, so I’m muting you. Sorry.
One imagine you’re not as constrained about what you can take on board a train though, so surely packing sufficiently portable food to account for two meals wouldn’t be the hugest hardship.
My honeymoon was the same… only we were in a Holiday Inn.
I recently looked into taking my dog with me on the train from Seattle, Wa. to Salem, Or. There was a weight limit of… I think 20lbs. Also the dog needed to fit into a carrier that fit under my seat.
My 50lb dog was a clear no go for that train route. I’m not sure if that applies to all Amtrak routes or if the rules vary.
ETA: I love riding the train and at one point I was living in Portland and working in Seattle. I had hoped to ride the train on Friday and Sunday but the speed of the train and the price made it impractical. I ended up doing that commute for 5 years by car.
Cyclops was a one-off vehicle that had to look unusual and maybe didn’t need a functional interior
but there were real buses like this around the same time
