« Your John McCain open thread | Main | Re-re-re-re-reading »

August 04, 2017

Comments

Trump is pretty well an irrelevance when it comes to electric vehicles.

The major players are already investing very large amounts of money
(for example):
http://media.daimler.com/marsMediaSite/en/instance/ko/Daimler-builds-battery-factory-in-Beijing.xhtml?oid=22832397
https://electrek.co/2017/04/27/panasonic-electric-vehicle-battery-factory-china/

While fuel regulations and subsidies for electric vehicles have been useful in priming the market, I think we are just about at the tipping point when the cost advantage of fully electric vehicles will see them take over the market almost completely.
Most interesting is the likelihood of fully autonomous self driving vehicles, which if realised will likely end the majority of individual car ownership, and lower the cost of motoring by between 50 and 75%. (Think always available, driverless Uber...)

We are way passed the tipping point, all electric fleets are going to bridge to autonomous driving the way flip phones bridged to smart phones.

It is possible I could live to see a predominantly autonomous driven world. That and I would like to die on Barsoom, or send my ashes to the next earth like planet if I don't make it. I am still able to just be in awe of some things.

Oh, and getting rid of orphan diseases genetically. Imagine the human suffering and money saved.

I still, in the back of my mind, believe we can create Heinlein's immortality before I die. I want to have lots of choices for my next body.

"It is possible I could live to see a predominantly autonomous driven world...."

I very much hope so - it's entirely conceivable within about a decade and a half.

The problem I have with autonomous vehicles is that I have too much experience with how buggy software can be. And frequently is.

It's one thing to have your finances messed up. That can be fixed. But a vehicle that crashes because something got coded wrong? Not saying that we won't get there. But I'm going to take a fair amount of convincing before I trust myself to one.

I'm with you, wj.

I bought a new car in 2013 -- first (and probably last) time I ever bought a brand new one. I test drove ten cars before making my choice, which was shaped by a bunch of factors, including the fact that I have back problems that make all but about 3 chairs on the planet uncomfortable for me.

Great car, I've had no trouble with it, knock on wood, and hope to keep it for a good many more years.

But -- the clock gains time. Over the course of several months it gets ahead of the real time, until when it's five or so minutes fast (once I waited til seven), I reset it to the correct time. In 67 years of life I've never had a clock, in a car or on the wall or on my wrist, that did that.

They can't even make the damned clock keep the time correctly, and I'm supposed to trust them to keep me out of an accident? Bah.

A favorite buggy software story.

I suppose it should be noted that cars today include dozens of computer chips. And an auto "mechanic" needs at least as much aptitude with a computer as with a wrench. So in some ways we are already quite a ways down the road.

I like driving. Especially cross country. Will I get to have a dummy steering wheel placed in front of me?

When I'm not driving over distance, I get bored and fall asleep and miss all of the sights.

Or will the annoying GPS lady give me a poke when we reach the Grand Canyon. If she says it in French, I might respond.

Without Neil Cassady, Jack Kerouac in an autonomously-driven vehicle wouldn't have had much of a story.

Will there still be car chases in movies? Or will the good guy chase the bad guy, both going the same speed, until the bad guy has to recharge?

There will be a spate of movies at first featuring rogue autonomous vehicles that connive with your appliances and run off with the girl. I'd like to see a Buster Keaton or Harold Arlen silent short featuring their interactions with autonomous cars.

Will there be autonomously-driven motorcycles?

I can see the concept working in urban areas, but without some sort of much improved long distance public ground transportation in the States, I don't see the attraction.

Who is going to enjoy the Daytona 500 when all of the hot cars are driverless. Maybe the racetracks will be spectatorless as well.

Or maybe our sex robots can go together on junkets to watch the cars go round and round. When the former aren't faking their orgasms, which seems like a technological innovation that leads us right back the reality we were trying to escape.

What about driverless bicycles. That would be a work saver.

How bout driverless bumper cars at the amusement park. Think of the fun in that.


Count,

Cars? Way too dangerous, think of the savings in human lives. You want to see the sights? Get an iPad. Honestly, I love driving cross country, but soon enough I will have to admit that you will be safer if I'm not driving.

So Heads Carolina, Tails California is on Spotify right now. Whats on there?

There is that.

I'm on an IPad as we speak.

I hear the Washington Monument is awesome on an iPad. I only have a Surface so I had to go see it.

I don't think the colors were as realistic live.

But -- the clock gains time. Over the course of several months it gets ahead of the real time, until when it's five or so minutes fast...

I have one of those in my 2008 Honda Fit. Standard cheap crystals are +/- 20 parts per million. Maximum error is about 50 seconds per month. But it's built into a gods-damned radio, and there are a number of ways to get a time tick to within a second over the air, and determine the proper correction. This is routine stuff that we assume entry-level real-time programmers know how to do. Or at least we used to.

I have an alarm clock that I've never had to set the time in the 15 or so years I've had it. It picks up a timing signal from, I think, Fort Collins, CO.

Use to have a watch that did the same.

The problem I have with autonomous vehicles is that I have too much experience with how buggy software can be. And frequently is.

seconded.

there are industries where the software quality is really really good. aircraft, medical devices. truly life and death systems.

they are that good because those industries willingly submit to regulations and best practice disciplines that make them that good.

even if self-driving vehicle systems met that standard, i'm not sure any of those domains are as straight-up real-time chaotic as navigating in traffic, with variable road conditions, and pedestrians, intermixed with live human drivers for at least another generation.

a kid runs into the roadway. you can hit the kid, or crash into the car next to you, or deliberately exit the roadway at some risk to your own life.

some decisions i would rather not have a machine make for me.

if we want to re-engineer the physical infrastructure currently used for driving to better support autonomous vehicles, i could see a cleaner path forward, but that is a much bigger lift than i think anyone wants to take on.

it'd be easier and probably more sensible to build out a usable rail system.

as far as trump getting on boatd, his priority appears to be extracting fossil fuels, period. the world will just work around him.

Just like coal fired power plants have to/had to go...

Wherever there is individual/familial motorized transportation as a primary mode of social transportation, it probably has to go.

We should be insisting/promoting/attentive to electric busses and walkable urbanism.

Cars = Isolated Ambition = Dead End

a kid runs into the roadway. you can hit the kid, or crash into the car next to you, or deliberately exit the roadway at some risk to your own life.

some decisions i would rather not have a machine make for me....

But what if that machine were better at detecting that kid, and faster reacting than any human driver ?

I think that ethical squeamishness might disappear should it become possible to demonstrate that the 30,000 road deaths a year in the US would drop to 3,000.

Indeed at some point I think the costs of auto insurance for human driven cars will make driving yourself a luxury available only to the very wealthy.

Wherever there is individual/familial motorized transportation as a primary mode of social transportation, it probably has to go.

We should be insisting/promoting/attentive to electric busses and walkable urbanism.

An ecosystem of driverless vehicles would probably see the current sharp distinction between personal and public transport disappear.

wj, look up ISO 26262.
We are a ways off from fully autonomous vehicles, but I think we'll see them a lot quicker than you expect.

Nigel,

your point about car insurance is valid only if you assume that vehicle insurance is fully cartellised. I don't assume that current insurance prices are causing insurance companies losses. They are as high as market will bear. If there would be driverless cars on the streets, the accident rates for driven cars would probably stay the same, so the cost of insuring drivers for accidents would not change. Raising vehicle insurance prices so high that they become unaffordable is possible only if all vehicle insurance providers on the market decide to do so and no one breaks the cartel.

However, I must say that the vehicle insurance in the US is really bad. In Europe, it typically covers all medical and other personal damages of all affected parties without ceiling, and the property damages of the non-guilty parties without ceiling. (Or the ceilings are so high as to be inconsequential, e.g. typically 100 million euros in Germany.) Law prohibits worse insurances. I was shocked by the fact that in many US states, you can have an insurance with liability ceiling around 20 grand, which is really nothing.

Nigel,

The current common wisdom in Europe seems to be that self-driving cars are a way ahead. Instead, we are going to see more autonomy in vehicles, and perhaps autonomous driving in optimal conditions like on motorways. However, autonomous driving in bad weather conditions and on rural roads is not going to happen soon.

This is probably partly because the traffic environment in Europe is much more difficult than in most of US. The intersections are not, as a rule, rectangular, and especially country roads off the motorways are both prevalent and difficult to perceive with machine vision. So, it is unfeasible to market "self-driving" here, as an "autonomous" vehicle is going to need human help in most routes. It is better to market "driver aids" that allow the car to drive most of the time.

The problem I have with autonomous vehicles is that I have too much experience with how buggy software can be. And frequently is.

this.

also, despite what many of its boosters claim, auto-driving cars can not be simply reactive; they can't just sense and respond quickly. at the speeds that make cars useful, momentum is the real problem. and to deal with momentum, the driver has to recognize, evaluate and anticipate what other cars (and animals and random trash on the highway) are doing. and that's AI, and we're not even close.

as russell mentioned, letting software decide who to kill, you or the pedestrian, is unsettling.

soon, computer science departments will require courses in ethics - not just business ethics, but the fundamental stuff: ten credit-hours on .

hmmm.

... ten credit-hours on the Trolley Problem.

"the driver has to recognize, evaluate and anticipate what other cars (and animals and random trash on the highway) are doing"

You don't need "AI" for that. Just Petawatt Death Lasers to clear that stuff out of the way.

Why, you don't even have to have a self-driving car for that after-market accessory to come in handy.

Just Petawatt Death Lasers to clear that stuff out of the way.

old-school 'cow-catchers' on the front of every car might work, too. plow your way to work, Mad-Max style!

formerly known as cleek,

again, in Northern Europe and - I think - in rural parts of North America - we have moose. They are large animals, and the bulk of the carcass is above the engine hood. If you collide with a moose head-on, a cow-catcher won't save you. It comes through the windshield right at you.

A moose jumps on the road, usually directly from the forest's shade, in bad lighting, because it moves in twilight. In such case, you must not only brake but swerve so that you anticipate the moose's movement, and you'd better drive to the ditch than collide. In fact, I'd be pretty happy to have an autonomous assistant taking care of the action. I drive on a road that has a lot of moose accidents regularly, and moose encounters are usually the most demanding driving situations I get into.

You can also get interesting situations in Australia, involving the wombat -- otherwise known as the "mobile land mine", due to its effect on any vehicle which hits one. (The wombat is typically uninjured.) Not to mention that vehicles driven outside the cities routinely come with a "'roo bar" attached to the front.

again, in Northern Europe and - I think - in rural parts of North America - we have moose. They are large animals, and the bulk of the carcass is above the engine hood. If you collide with a moose head-on, a cow-catcher won't save you. It comes through the windshield right at you.

Moose range in North America and high car density don't overlap a lot. F*cking deer, OTOH, are everywhere. Per this article, in the US there are about a million deer-vehicle collisions per year, killing about 200 people and resulting in over $4B in property damage. Canada has similar statistics for deaths when scaled for population.

A cowcatcher would eliminate some but not all deaths due to deer-vehicle collisions. The ones where the carcass catches in a wheel well and causes severe loss of control are avoided, assuming the cow catcher is wide enough. The ones where the carcass comes through the front windshield are not.

I've only been involved in a collision with a deer once. In broad daylight, several came running up out of the brush in the right-hand ditch. I had gotten almost stopped when one of the last ones tried to leap over the hood of the small truck I was driving. It failed to clear, slid across the hood on its belly, landed on its feet and kept running.

y'all have much greater faith in the capabilities of software and machine intelligence than i do.

notably, the folks who seem most skeptical here - cleek, wj, myself - are all hands-on software people.

driving is really complex. you aren't talking about getting one vehicle o navigate a complex environment, you're talking about getting thousands and thousands of them to do so, at the same time.

another point not yet mentioned is the susceptibility to mischief, i.e., hacking.

to follow on b9n10nt's comments (welcome!) IMO we'd do better to invest in credible public transportation. if you want to sit back and watch the world go by, take the bus.

gadgets are cool in a gee-whiz way, but there are solutions to reducing the risks of getting from point a to point b that are much simpler, and less expensive, and lower risk, than teaching a car or truck to drive itself.

to be honest, IMO the real impetus behind self-driving stuff is shedding the labor cost of drivers. load up the semi, code in the destination, and off it goes.

guess what the most common reasonably-paying job is in a lot of good old middle america.

cui bono. always a good question to ask.

there are industries where the software quality is really really good. aircraft, medical devices. truly life and death systems.

I posted this above A favorite buggy software story and now I'll point out that it's about the F22 and the international date line.

IMO the real impetus behind self-driving stuff is shedding the labor cost of drivers.

Also to repeat (but from longer ago, and probably not for the last time): as Jared Diamond says, invention is the mother of necessity.

The move to driverless vehicles is surely driven more by $ and ego (Elon Musk anyone?) than any altruistic considerations, after-the-fact arguments aside. The smartest guys in the room are dreaming this up...quite literally. So we will have it.

Public trans? We tried that, and look what happened. (Three links, that last from a skeptic, just in the name of equal time. Even there, there's debate in the comments.)

Cui bono indeed.

i am a bit surprised that car companies are pushing this driverless car thing so hard, given the liability issues that they'll face.

On the upside of vehicle/animal collisions...

Harvesting Roadkill Now Legal in at Least Half of U.S. States: Oregon is the latest. Let's welcome this tasty trend.

i live about 2 miles from a big cat rescue park. they'll happily take all the fresh roadkill you can bring them.

If I were going to develop an autonomous car, the market I would focus on would be aging Boomers in the suburbs (and their grown, worrying children), looking to extend the time they could spend in their own home. Small, electric, limited range (50 miles should be plenty), limited top speed (45 should be enough). Trips to the doctor, to the grocery, out to eat, etc. Probably stay in on bad weather days ("Sorry, doc, I need to reschedule because it's snowing, and my car doesn't like snow").

y'all have much greater faith in the capabilities of software and machine intelligence than i do.

I have limited faith in the practice of, year after year, programming new driver units saturated in alcohol and hormones...

Streetcar funds also come at the expense of more effective transit solutions. "Many cities are spending money on streetcars at the expense of their bus system," notes Feigenbaum, even though buses are cheaper, faster, and more flexible. To Feigenbaum, this underlines the fact that these projects are "really not about transportation—they're about economic development. But even the economic development is not really there, so what's the point?"

America's Cities Double Down on Trolley Follies: Streetcars continue to see cost overruns while failing to produce promised economic development.

If I were going to develop an autonomous car, the market I would focus on would be aging Boomers in the suburbs...

Soccer moms could also benefit. They could get their kids to/from soccer practice and all of the other places kids too young to drive need to be without having to do the driving themselves.

who do you think is writing code? :)

Speaking of software:

I own a computer, a smart-ass phone, and a "smart" TV at home. I work with digital oscilloscopes and other software-driven instruments, on a software-controlled medical "robot". When taking a break, I get my coffee out of a software-controlled Keurig machine.

One of these days, I will buy a cheap stopwatch. I will use it faithfully for a week (Sunday midnight to Sunday midnight) to measure the cumulative time I spend, per week, waiting for software to load or initialize or restart or whatever the hell it's doing between whenever I flip a switch and the device in question has finished "thinking".

Any bets on how many hours that will add up to?

If I ever buy a self-driving car, will I have to start it an hour or two before I need to leave, to give the software time to wake up?

--TP

Tony,

I own all those things and my advice is don't turn them off. Find a good low light screen saver for the electronics including the tv and just leave it on all the time. Besides the vast majority of electronics failures are when you hit the on switch.

I have also saved a ton of time using biometrics everywhere it's offered to login.

"will I have to start it an hour or two before I need to leave, to give the software time to wake up?"

The tubes have to warm up, same as always.

As for the moose hazard, Trump has appointed Boris and Natasha to a task force to deal with the problem.

No comment from the squirrel that lives on Trump's head.

If I ever buy a self-driving car, will I have to start it an hour or two before I need to leave, to give the software time to wake up?

imagine the joy that will ensue when you're rushing to make it to a job interview on time, jump in your smart-car, tell it to burn rubber and get you cross-town, now! but it's busy installing a new OS patch. please wait. 5%..... 6%.....

CharlesWT,

a tram or another type of rail transport becomes necessary when you get buses moving every ten minutes or so. After that, the buses will simply stack and the quality of transport no longer gets better. Second, trams become necessary when the passanger loads are around 150-200 per unit, as buses that large or larger are about as expensive than trams and require at least as much maintenance.

In Helsinki, we tried to have a rail-like bus system with partly dedicated roads (line 550, the "Jokeri"), but it has been popular enough that these constraints are now preventing further improvement of service, so it will be changed to a tram. Similarly, the city of Tampere is now replacing its most crowded bus line with a tram simply because bus technology does not allow further improvement. The same situation has been experienced in numerous European cities.

As far as I know, rolling stock on rails is built so that it is, despite initial capital cost, quite durable. The service life is 50-60 years, while buses serve about 10-15 years maximum. Of course, this assumes that maintenance is done well.

When it comes to US cities, I have very limited experience in their mass transit. However, for example, San Diego trams I have ridden on don't seem to have passanger loads that would really justify the system. I doubt that rather few American urban areas have population densities justifying rail systems.

Marty,

Don't take "flip a switch" too literally. Even a permanently-on computer can't be running all its apps all the time. Launching a program amounts to "flipping a switch" in my parlance.

In the work environment, unplugging things like scopes or "robots" so you can move them from one place in the lab to another is unavoidable.

And, what cleek said about "updates" happening at inopportune moments.

--TP

While automakers focus on defending the systems in their cars against hackers, there may be other ways for the malicious to mess with self-driving cars. Security researchers at the University of Washington have shown they can get computer vision systems to misidentify road signs using nothing more than stickers made on a home printer.
Researchers Find a Malicious Way to Meddle with Autonomous Cars

Ohio’s banking on a $218 million “driverless car” highway to get techies out of sunny California and test their autonomous tech alongside cornfields, cows and Cedar Point.
Ohio Is Building The Country's Longest Driverless Car Ready Highway

Michael Cain,

in Finland, moose are an everywhere species, as the whole Southern part of the country is rather well-suited for them. The current mode of forestry creates an environment that is very favourable for moose and roe deer, so we have a winter population of roughly 60,000 moose and 100,00 deer. However, in European perspective, we are a far-off and rural country, about as central as Oregon.

If I ever buy a self-driving car, will I have to start it an hour or two before I need to leave, to give the software time to wake up?

The typical luxury car today has >20 networked processors executing code compiled from >10 million lines of source. How long does it take that software to wake up?

I don't know, Michael -- I drive a 2005 Hyundai Elantra :)

Even so old an econobox as mine surely relies on one or more computers, but aside from controlling the engine I don't think they have very much to do.

And now for something completely different: what I want is a modular car. For commuting and daily running about, 2 seats are plenty. Occasionaly, lugging around a rear module for extra passengers or cargo would be convenient. This "modular car" fantasy is probably easier to implement with electric drive and drive-by-wire, though of course at the cost of some additional software :)

--TP

the computers in cars come up fast because they don't have to copy their software from disk into a place where the processor can see it (RAM). in embedded devices, the software is always where the CPU can see it.

it's similar to the very low-level OS that personal computers use: the BIOS. that code lives on a chip and is available to the processor as soon as it's powered on. that's the first boot screen you see on a PC. that software then loads the OS, which then loads all your startup apps and device drivers, etc.. that's what takes up all the time.

also, the computers in your car might not actually shut down when you turn off the car - they might only need to boot when you first connect the battery.

Some years ago, there were rumors that Microsoft was planning to extend the BIOS into a low-level operating system. What ever happen to that?

There's a big difference in reliability between embedded systems and general purpose computers crawling with third-party software.

Any bets on how many hours that will add up to?

everything is on the network now.

you're not just turning on a system, you're turning on a system, which reaches out to check for updates, and connects to and synchronizes with any other systems it interoperates with, and sends usage an other instrumentation information to a server, and so on, and so on, and.....

eventually, returns to your control.

Note that, if you just have a few dozen chips in your car talking to each other (i.e. over hard wires), you are reasonably secure. But if you routinely go out over the Internet for updates, information, etc., you are wide open for hacking.

Why would anyone want to hack your car? No idea. But in a world with nut cases who are willing to become suicide bombers, someone who will hack a car in order to create a traffic mess doesn't seem like a stretch.

Random remarks...

The latest iPhone designs suggest that updates over an open network can be acceptably secure. Apple is confident enough that they publish white papers explaining exactly how they do it.

The heavy lifting for vision/ultrasound/radar processing is typically done on highly parallel special purpose hardware. The Nvidia platform used by Tesla and others doesn't normally reboot -- it's put into a special low-power state where no processing occurs, but state is preserved.

In my experience, both first hand and observing, people who specify and code high-reliability high-availability software and who are good at it are all textbook paranoids while they are working. The one big job I did in that area was the only time in my career when I woke up at 2:00 AM just knowing that I'd missed some failure mode. 75% of the processing in the key decision-making component was answering the questions, "Am I in a valid state?" and "Does this instruction I've received make sense in my current state?"

There is a growing body of evidence that the people who do the control software for our electric grids are not nearly paranoid enough.

A study has found that it would be fairly simple to program autonomous vehicles to make similar moral decisions as human drivers. In light of this, the question becomes whether we want driverless cars to emulate us or behave differently.
Study Finds That Human Ethics Could Be Easily Programmed Into Driverless Cars

A study has found that it would be fairly simple to program autonomous vehicles to make similar moral decisions as human drivers

I submit that the moral decisions humans are required to make, while driving or otherwise, are not simple. They are everything other than simple. It will therefore never, ever be simple to program them.

Code doesn't make hard things simple, it just makes them automatic. Fast, accurate, not (or less) prone to factors like fatigue and boredom; all machine virtues. Making inherently difficult things simple; out of scope, whether for machines or people.

The only moral decisions that will be "fairly simple" to program are trivial ones. Maybe that's sufficient for driving, maybe not.

What makes programming hard is the fact that reality is hard. There is no programming language, technique, or discipline that makes that go away.

Lane departure, automatic braking, collision avoidance... the future is now, sort of. I still rely on my 2008 Teutonic tech, so I don't have any experience with it. And none of this is fully autonomous, to my knowledge. But we're handing over the reins slowly but surely.

I wonder how much damage a mischievous soul might wreak with a can of white spray paint on an open road?

I wonder how much damage a mischievous soul might wreak with a can of white spray paint on an open road?

It's an arms race between AIs and "mischievous souls." Which the AIs will eventually win. They never forget what they learn and instantly tell all their copies. Pranks by humans are just learning opportunities.

But then there are the occasions when someone makes an emergency road sign. Which a person would give at least provisional notice, depending on circumstances. But the well educated AI "knows better" than to pay any attention to.

AIs can do lots of impressive things. But it will be quite a while before they reach the point of having anything resembling judgement.

Experience might substitute for some judgment. Especially if that experience is being shared among 100s of thousands of autonomous vehicles from all over the world. Although I suppose, it would take an advanced AI to intergrade all that experience.

"You’re probably safer in a self-driving car than with a 16-year-old, or a 90-year-old," says Schoettle. "But you’re probably significantly safer with an alert, experienced, middle-aged driver than in a self-driving car."

Puny Humans Still See the World Better Than Self-Driving Cars

The guy who played Jar Jar Binks:
https://www.wired.com/2017/07/ahmed-best-jar-jar-binks-new-podcast/

I submit that the moral decisions humans are required to make, while driving or otherwise, are not simple. They are everything other than simple. It will therefore never, ever be simple to program them...

I'd entirely agree.

A couple of observations, though...
In the case of situations which require an instant reaction, you have the luxury of all the time you need, if coding a moral calculus - whereas the human actor could be paralysed by indecision, or simply make a decision they deeply regret with the benefit of hindsight.

Following on from that, though, I'd argue that far too little attention is paid by society at large to the underlying moral assumptions, norms etc which are necessarily hard wired into our electronic ecosystem.
Lessig's Code and Other Laws of Cyberspace was one attempt to open the discussion to society at large, but it seems to me a discussion which never really took off.

Which is to say I think we're having the wrong argument.
It shouldn't be if such systems will be implemented, as it seems entirely probable that they will, rather how they should be implemented, and what kind of public oversight should there be.

i don't see any of the big auto or tech companies opening up their code to public scrutiny. if they're going to spend a decade developing it, they aren't going to give it away.

i don't see any of the big auto or tech companies opening up their code to public scrutiny. if they're going to spend a decade developing it, they aren't going to give it away.

Which is precisely my point.

We accept a situation where algorithms make decisions on anything from creditworthiness to criminal justice bail and sentencing while allowing those who write them to claim commercial confidentiality and thereby evade any assessment of the criteria on which such decisions are taken.
Unacceptable in a civilised society.

A case in point:
https://www.nytimes.com/2017/05/01/us/politics/sent-to-prison-by-a-software-programs-secret-algorithms.html

A lot of AI is black boxes. Not even the software's creators can know, much less, explain why the software made a particular decision.

JanieM's very apt point about the F22 aside, the aviation industry is one that (a) is highly dependent on software and (b) has an outstanding safety record.

Avionics software development is ruled by DO-178, a document that captures standards for all aspects of systems development for avionics. That document in turn provides a standard that is used by the FAA and other national and international authorities to certify software systems intended for use in aircraft.

I don't know if there is a legal requirement for systems used in aircraft to be FAA certified, but legal requirement or not it almost certainly is, if only for reasons of manufacturer and operator liability.

There is an analogous standard for software systems intended for use in ground transportation - ISO 26262, cited by Nigel upthread. There is also an authority responsible for ground transportation safety, the National Highway Transportation Safety Administration.

So, the elements for an industrial culture with the same level of safety focus that exists for aviation are in place for ground transportation. I don't know how the scope and rigor of ISO 26262 compares to that of DO-178, nor do I know what level of compliance with ISO 26262 is required for automotive software systems, nor do I know what degree of authority the NHTSA has for requiring that compliance. But, the basic elements are there, should we actually decide to employ them, which is a sketchy proposition in the era of (R) de-regulatory mania.

My understanding, from a very limited amount of general industry anecdotal information, is that a very skilled SW engineer working in avionics can expect to produce about 1,000 lines of deliverable code.

In a year.

Take that with several grains of salt, because it ain't my domain, but that's my order-of-magnitude understanding.

That's a very, very, very, very laborious development process.

Aircraft are a very high-margin item. They cost many, many millions of dollars, they have very long design-to-delivery timeframes, the development process includes extensive safety and real-life test cycles.

Ground transportation is not exactly the same development lifecycle. Similar, maybe, in some ways, but not exactly the same.

Can the auto industry absorb the cost of something like avionics safety standards? Maybe they already are in place, I don't know. Maybe somebody reading this knows the answer.

In any case, there are often gaps between what is possible, and what is actually feasible within a given business context. Planes are expensive, high-margin, limited run items. Cars are, at this point, cranked out like toasters. I don't know how that would or would not affect the economics of the development process.

And all of that to the side, ground transportation operates in an utterly different kind of environment than aircraft.

Imagine a world where the entry and exit of every automobile on to and off of a major highway was orchestrated by representatives of a federal agency. Where every car stayed in its lane, always, and maintained a strict distance from the car in front of it. Where the path that every car was going to take, from point of departure to destination, was negotiated and recorded with a federal agency before the car ever entered the highway.

That's the operating environment of aircraft.

I have no problem with automated assistance for stuff like braking, collision detection and avoidance, staying in your lane, etc. Those things are all great.

There are over 250 million cars in the US. There are millions of cars on the road in the US every day. There are millions of cars on the road in some individual major cities, in the US, every day.

It's a really f***ing hard problem. Like all really hard problems, solving it is going to create other problems, and we won't know what most of them are until they pop up. Because that's the nature of hard problems.

What makes building good software hard is the fact that reality is hard. Solve one problem, you simply surface - or create, or exacerbate - another one. Or, another five. Or ten.

Will the problems solved by autonomous ground transportation create more value than the problems it creates? Who knows. We won't know until we get there.

We love our gadgets, and the only thing we love more than our gadgets is improving return on investment by making humans redundant, so we will surely try this on.

But don't kid yourself that it's gonna be an easy lift, or that we'll like where we end up any better than where we are now.

Nigel: We accept a situation where algorithms make decisions on anything from creditworthiness to criminal justice bail and sentencing while allowing those who write them to claim commercial confidentiality and thereby evade any assessment of the criteria on which such decisions are taken.
Unacceptable in a civilised society.

Russell: Avionics software development is ruled by DO-178, a document that captures standards for all aspects of systems development for avionics. That document in turn provides a standard that is used by the FAA and other national and international authorities to certify software systems intended for use in aircraft.

Note the difference here. The avionics software is written to a set of standards. The code may be different, even its quality may differ. But it doesn't depend on "secret algorithms".

I would have little problem with using legal software which was written to published, and reviewable, design guidelines. But one where it's a black box that's not subject to review? That's a problem.

Similarly with "ethics software", whether in cars or anywhere else. You can, maybe, keep your actual code secret. But the design, the ethical standards, have to be open.

Thanks for all this; I'm learning something here.

russell, your 11:02 could have been a front page piece.

The F22 thing struck my fancy because here we have a $Gazillion aircraft, presumably designed by the very best people, and a stupid little "oops" creeps in that could happen to anyone (it could!), but that in an unlucky context might have cost many lives.

Do voting machines have published, reviewable standards?

The most basic question I have about all of this is:

Do we really want to make even greater investments in a transportation model that requires every person who wants to get from point A to point B to own their very own personal ton-and-a-half or two-ton (or more) machine to get there?

Consider the resources and infrastructure we spend on personal automobiles - basically, an entire society and culture organized around the ability of every person to jump in a really large, resource-intensive, expensive machine, and drive it wherever they want, whenever they want.

It's extraordinarily convenient, and extraordinarily expensive, in about ten thousand different ways.

Why perpetuate that?

If we're going to invest the resources needed to do something like let tens of millions of robots with people in them zoom around autonomously in environments full of people and property, why not spend those resources on building something better.

Where, for "better", I mean scaleable and sustainable.

Everybody in the US has cars. Lots of people in the EU have cars. "Having a car" is the model of having arrived, having achieved first world status.

A billion people in China, a billion three in India. Those places are moving toward first-world status. At least for some of their folks.

They are going to want cars. Can the world afford to support billions of cars?

Just asking.

Maybe there are better choices out there.

China: I spent a month there with my son in 2010. At the end of that year (IIRC), Beijing was about to put a cap on the # of cars that could be sold in the city per day. Leading up to that cap, 3,000 cars were being sold every day in the city. The essay I linked a few days ago in a different context, a review of Peter Hessler's "Country Driving," has a lot of material about cars and China. Indeed, as a pedestrian in 2010 I was taking my life into my own hands ever time I crossed a street. ;-)

Quibble about population: Google gives China's as 1.379 billion. On the street estimates when I was there were that it was already up around 1.5 billion.

Otherwise, about cars, wrs.

But that notion of being able to hop in the car and go anywhere you want at any moment is deeply ingrained by now. Just try not being able to drive for a while, and see what it does to your psyche. My mom voluntarily gave up her car keys almost four years ago (she's almost 94), and she still mourns and moans about it.

an entire society and culture organized around the ability of every person to jump in a really large . . . machine, and drive it wherever they want, whenever they want.
.
That's really the core. And the piece that, if we want to address the "resource-intensive, expensive" part, will have to be maintained. At least, if we want to get any significant level of voluntary acceptance. There's just too much of the population who lives where anything else just isn't viable -- certainly not as an exclusive approach.

It's not just the society and culture. It's also the infrastructure and (geographic) distribution of people and businesses.

wj -- the distribution of people and businesses rearranged itself over time as cars came to be widely used. At least in theory, is there any reason that the distribution can't/won't re-rearrange itself in response to a contrary stimulus? The problem is, as usual: $. Who gets it?

Just try not being able to drive for a while

It's not just the society and culture. It's also the infrastructure and (geographic) distribution of people and businesses.

From about 1981 to about 1985, I did not have car. I lived in Philadelphia, then in Salem MA. Most places I needed to get to on a daily basis I walked to. If it was more than 20 or so minutes, I took a subway, train, or bus. Anything long distance, train, or public transportation to airport and then plane.

Beginning about 1984, I lived in Salem and worked in Boston. I took the train. I walked to the train station. When it rained, I carried an umbrella.

Not an uncommon way of life, for a lot of people.

I drive more or less daily now, but I drive less than 10K miles a year. If I got off my lazy @ss and rode my bike more I could probably cut that in half.

If there was reasonable bus service from home to work, I'd use that.

What I'd say about the geographic distribution of people and businesses is that a lot of that distribution has occurred with the assumption that people would drive everywhere. So, society and culture grows up around basic ground facts. Often, as for us at present, those are so ingrained as to be second nature - you don't even think of it as a choice,

But it is, nearly always, a choice.

Yes, in very rural areas, public transportation isn't practical. Drive cars in those places.

Most people don't live in those places.

What we're discussing in this thread is, literally, this:


  • deploy millions and millions of robots
  • weighing anywhere from one ton to (for freight transportation) 40 tons
  • carrying people and valuable (and sometimes dangerous) goods and materials
  • through environments full of people and valuable property
  • to operate autonomously or nearly so
  • on surfaces of widely varying quality, in all kinds of weather conditions
  • in compliance with rules that vary from location to location, and from context to context
  • interoperating with human-driven versions of the same, for at least another generation
  • at least as safely as humans do this now, if not significantly more so

Please note the "millions and millions" part. Millions and millions of multi-ton autonomous robots, carrying people, zooming around the country.

Nothing I'm advocating for seems any nuttier than that, to me. It would just require people to change some aspects of how they live, and arrange their lives.

People do that all the time.

At least in theory, is there any reason that the distribution can't/won't re-rearrange itself in response to a contrary stimulus?

I think it comes down to sunk costs. When we first built out suburbs (and that's a big part of what we are talking about), those moving there could sell their city residence, if they owned it. There were plenty of potential buyers. But if we try to obsolete them, where's the market for those suburban houses? And it's not like we can convert them to the kind of high density areas that make cities workable.

russell, a couple of rebuttals to your thoughts on self driving cars...

First, if they do work, then everyone won't own one. The single biggest plus in their favour is that they will enable a car owning lifestyle without actually having to own a car - at a cost maybe a third of car ownership.

Secondly, it's quite likely they won't be adopted first in the US.
A small state like Singapore would much more easily be able to legislate for them - and the driving envelope is quite likely much narrower there. Once the rather large cost and convenience advantages become apparent, then it will be a matter of having to adopt them so as not to lose competitive advantage.

a lot of jobs can be done 'at home' - programming, certainly.

my last manager was happy about letting work from home whenever i needed to. but, my current manager thinks being able to talk with his people in person is more important.

he also drives a Prius. which is fine. but if saving gas is important...

i haven't been exasperated enough to tell him that i could 'save' more gas than his Prius does by just working from home a few days every week. i use two gallons of gas a day. making that trip to/from work, just on the off-chance that he needs to pop in to my office and tell me something that he just can't tell me via email.

my wife used to work for IBM, and they were fie with employees working from home, too. then some bean counter figured out that they were maintaining and cooling/heating a giant office building that was mostly empty because everyone worked from home. solution: make everyone come in.

we should do better.

From autonomous public buses, driverless taxis to truck platooning, Singapore is looking to catalyse the development and deployment of autonomous vehicles (AVs) to enhance its transportation system. The Land Transport Authority (LTA) is exploring how self-driving technology can be applied to bring in new forms of shared mobility, and also address constraints in land and manpower.
Singapore gears up for autonomous vehicle technology

The Japanese manufacturer of autonomous buses, SB Drive (a unit of SoftBank), has not only just raised venture capital from Yahoo! Japan in the amount of over 4.6 mio dollars, but also showcased how the solution can be used in rural areas.
SB Drive Shows Autonomous Bus in Rural Areas

wj and Nigel, all good points.

charles, thank you for always finding interesting and relevant data points

Last night I finally had a chance to read the Michael Lewis article in Vanity Fair that the Count flagged up in the other (obsolete?) open thread. I strongly urge anybody who hasn't read it yet to do so. The implications (when you look at the programs and situations where oversight has withered and stalled) are truly mind-blowingly horrendous, even if the failure of the Trump administration to make a proper transition was foreseeable and is unsurprising. This is the first thing I have read that makes me think that, all in all, Pence would be better. I have taken the view that, because he would be more competent, he would probably get more horrendous stuff done than Trump, but perhaps a competent (if ideologicallyy awful) administration would at least be able to staff such a vital department with effective people. What think you all?

Should have said for anybody who didn't read the Count's original post: the article was about the appalling, dangerous situation in the DOE.

then some bean counter figured out that they were maintaining and cooling/heating a giant office building that was mostly empty because everyone worked from home. solution: make everyone come in.

Didn't someone mention sunk cost upthread? I thought bean counters were supposed to understand such things at least as well as anyone would.

Unless office space is in serious oversupply locally (which historically comes and goes in most urban areas) a big office building that you already have could be made into a serious cash cow. At least by any halfway competent finance department.

What think you all?

I think there is tremendous value in competence, especially regarding things of such consequence.

I don't know if, net/net, we'd be better off with POTUS Pence or not.

Funny times, these.

Meanwhile - news that may be of great interest to the ObWi community!!!

Looks like we may have started something...

Well, we always knew the folks here were ahead of our time.

"In related news, the cat owner said that she hadn't siezed the radio station..."

So it turns out our kitty was a forerunner of a desperate (although moderate) legion of cats, armed to the teeth and heading to join the resistance. After all, I assume the 2nd Amendment doesn't specify that the militia has to be human....

anyone know where that kitty pic came from?

anyone know where that kitty pic came from?

The question inspired me to go back to the first few posts to see if it was mentioned.

It was not.

I first came here at the time of Andy's death, when the blog was (in my mind at least) overwhelmingly associated with hilzoy. Going back to see that Moe Lane, Katherine, and von were the ones here at the beginning was quite interesting.

Anyone know offhand when hilzoy joined?

There are so many things about blogs that it seems like some Ph.D. student would/should be studying: like, how commenters come and go, frequency of posts and comments, how moderation works in practice, the flow of political opinions, etc. But this tiny dip back to the beginning of ObWi reminds me that blogowners come and go too. There was a lot of evolving just in the first few years I was reading, and now look where we've gotten ourselves to. ;-)

For some reason, Google image search associates the kitty pic with Schloss Burg(Burg Castle).

i've seen people using it as their Disqus avatar.

On a totally different note:
http://www.eastbaytimes.com/2017/08/07/is-vladimir-putin-trolling-donald-trump-with-his-he-man-vacation-photos/

Then I find this from the Economist:

Britain started shaping the legal environment for self-driving cars, issuing new guidelines on mandatory protection against hackers, allowing drivers the right to delete their personal data, and telling manufacturers to adopt more stringent design standards. They must plan to protect security over a vehicle’s lifetime. The government also heralded legislation governing insurance for self-driving vehicles.
Looks like a start on addressing the hacking issue at least.

JanieM's 10.31 inspired me to go back and read a few posts from the beginning (ObWi was all about hilzoy for me too when I started lurking), and I found this wonderful Camus quotation in Katherine's first post:

"In this way the Right abandoned the monopoly of the moral reflex to the Left, which yielded to it the monopoly of the patriotic reflex. The country suffered doubly. We could have used moralists less joyfully resigned to their country's misfortune and patriots less ready to allow torturers to claim they were acting in the name of France....

If I annoy anyone by writing this, I ask him merely to think for a moment about the divergence between the ideological reflexes. Some what [I assume "want"] their country to identify itself wholly with justice, and they are right. But is it possible to be just and free in a dead or subjugated nation? And does not absolute purity for a nation coincide with historical death? Others want the very body of their country to be defended against the whole universe if need be, and they are not wrong. But is it possible to survive as a people without doing reasonable justice to other peoples? France is dying through inability to solve this dilemma."

She was talking mainly about Iraq (it was 2003), but it seems to me that this is a very interesting and significant area for thought and discussion, no matter the time and place.

"Britain started shaping the legal environment for self-driving cars..."

leading the way into the automobile future, as always!

The "Locomotive Acts" passed in 1865 by the British Parliament set out a series of legal restrictions ... one provision set the speed limit at 2 miles per hour within towns; another required a person to walk in front of the vehicle waving a red flag...at all times.

That could work. Even russell might find it acceptable.

Snarki, I seem to remember that those rules included one which said that if livestock were being bothered by a car waiting for them to pass, the driver had to then disassemble....the whole car.

Even russell might find it acceptable.

LOL.

Only if every possible downstream consequence is carefully considered before moving forward.

Think of the effect of waving red flags on the local bull population, for example.

GFNTC, thanks for the Camus quote. Haven't read him in a long, long time - his thinking always seemed profoundly humane, to me.

Decent goalkeeper, too.

The comments to this entry are closed.

Blog powered by Typepad