Tesla Is A Zero - Karl Denninger

[RE]

I'm actually quite surprised it happened, and I don't rule out some kind of monkey wrenching. By posting the youtube vids bragging about the car, he could have put himself squarely in the crosshairs of a target.

Eddie, I'm surprised you're surprised.  I've thought carefully about what it would entail to have a machine do the driving -- ever, to any degree, from "keep your hands on the wheel" to "sit in the passenger or back seat and read your newspaper."

Given your response, I suspect you have not thought carefully thorough what it means to allow a car to drive itself in a real world environment.  It should be clear to anyone smart and informed that we humans simply don't have that kind of technology, and won't anytime soon (if ever).  It would require an AI apparatus at least as good as a human driver.  Driving a car is a very, very sophisticated thing.  (Obviously many humans are not up to the task!)

At least according to the press, Google's self-driving cars have logged millions of miles around LA.

Like you, I am pretty skeptical of this technology in real life situations.  I can certainly see it working in controlled environments like say at Disney World, but the open road simply has too many Random Factors and Random Events that occur to see how it could all be programmed in advance.

I myself certainly cannot imagine riding in an auto pilot taxi cab in NYC traffic.  You would need to have a Death Wish for that.

RE

• Tweet

[RE]

Most humans aren't particularly good drivers. Even with an occasional failure, I'd bet driverless tech reduces fatal and non-fatal accidents substantially. But at this stage of the game, one big failure might be enough to put it on hold for years.

Elon Musk is the poster boy for divorcing cars from fossil fuels. (Whether that's an accurate assessment is doubtful, but it doesn't matter).

Who benefits from taking Tesla down?  Just the oil companies....and maybe Detroit automakers. Would they murder someone just for profit?

Even if you accept this theory as true, it presents a whole other problem for driverless carz.

All anybody needs to do to kill somebody they don't like is to Hack the car's OP system.

RE

• Tweet

[Eddie]

Like this?

https://www.wired.com/2015/07/hackers-remotely-kill-jeep-highway/

or maybe this?

http://whowhatwhy.org/2015/02/20/car-hacking-report-refuels-concerns-michael-hastings-crash/

It's the Achilles heel of all modern cars, not just driverless ones.

• Tweet

[RE]

Like this?

https://www.wired.com/2015/07/hackers-remotely-kill-jeep-highway/

or maybe this?

https://www.wired.com/2015/07/hackers-remotely-kill-jeep-highway/

It's the Achilles heel of all modern cars, not just driverless ones.

One of the reasons I drive Old Carz. 

RE

• Tweet

[Eddie]

The second link was supposed to be about the murder of Michael Hastings.

• Tweet

[jdwheeler42]

Given your response, I suspect you have not thought carefully thorough what it means to allow a car to drive itself in a real world environment.  It should be clear to anyone smart and informed that we humans simply don't have that kind of technology, and won't anytime soon (if ever).  It would require an AI apparatus at least as good as a human driver.  Driving a car is a very, very sophisticated thing.  (Obviously many humans are not up to the task!)

It seems to me like these car companies have been going about it all wrong with the "driverless taxi" approach, both from an AI and a fiscal standpoint.  An "enhanced cruise control" approach seems like it would be much more productive, for use on limited-access highways, combining collision-avoidance radar with passive RFID chips embedded in road markings that would provide information on route, mile marker, lane, and speed limits.  Active RFID chips could broadcast overrides for lane closures and reduced speeds.

This could be sufficiently less expensive to become a standard feature on new cars, especially since the cost of embedding the RFID chips would be passed off to the taxpayers.

• Tweet

[JRM]

Most humans aren't particularly good drivers. Even with an occasional failure, I'd bet driverless tech reduces fatal and non-fatal accidents substantially.

Oh, believe me, I'm poignantly aware of the fact that too many human drivers are very ill-suited for the task of not killing pedestrians, bicyclists, other drivers and babies in baby carriages.  I figured this into my response! The problem with these human drivers is that they seem to think they are asleep and dreaming (many actually are), or that they are playing some kind of video game (again, man of them are). What they lack is attention, which is diverted at a moment's notice by any shiny thing (or thought, memory....  I get that! Many drivers don't even know what they are doing is putting everyone around them at risk. They are, in short, really dumb. Don't give them firearms!

My point is not relevant to all of this. My point is that the technology does not exist at present to have driverless cars that will not on some occasions do just as poorly as a poor human driver, or even much worse.  Which brings us to ...

But at this stage of the game, one big failure might be enough to put it on hold for years.

There will not be one big failure as long as the roads, streets and highways of our society are the test site for this failed and inadequate technology. There will be just enough failures to insure that the public relations catastrophe will inevitably come, again and again and again. The reason is simple: Machines are terrible drivers in real world complexity (actual driving conditions). Nothing can done to change this fact ... until these damned contraptions can ubiquitously pass the Turing Test, passing as humans.  Even if that were to occur, the cost per unit would be prohibitive in your lifetime, my lifetime and the lifetimes of our grandchildren (where applicable).  Let's leave the more philosophical questions to those grandchildren to contend with. We, on the other hand, should know better.

• Tweet

[Eddie]

That's a feeling you have. The data don't support it.

Flown in a plane in the last ten years or so? They all fly themselves, most of the time. Driving a plane is more complicated than driving a car.

• Tweet

[RE]

Driving a plane is more complicated than driving a car.

No it is not.

Planes have defined levels at which they fly East or West, North or South.  There are many fewer of them in the air at any given time than there are cars on the road in your average city.  Planes do not have bad road conditions to deal with or for the most part drunk pilots in the air.

The roads are a very complex place with a large number of vehicles, many different parameters and constantly changing environment.  You can go from dry to a t-storm in seconds, you can go from decent traction to black ice in nano-seconds.  Programming all those possibilities in advance is about impossible, and the AI is not there to make split second decisions that can save your life.  I know, because I made those decisions, and when I went AIRBORNE off the Glenn Highway, no AI could have done what I did.  Only a well trained Homo Sap driver could have done that.

RE

• Tweet

[Surly1]

Driving a plane is more complicated than driving a car.

No it is not.

Planes have defined levels at which they fly East or West, North or South.  There are many fewer of them in the air at any given time than there are cars on the road in your average city.  Planes do not have bad road conditions to deal with or for the most part drunk pilots in the air.

The roads are a very complex place with a large number of vehicles, many different parameters and constantly changing environment.  You can go from dry to a t-storm in seconds, you can go from decent traction to black ice in nano-seconds.  Programming all those possibilities in advance is about impossible, and the AI is not there to make split second decisions that can save your life.  I know, because I made those decisions, and when I went AIRBORNE off the Glenn Highway, no AI could have done what I did.  Only a well trained Homo Sap driver could have done that.

RE

Jesus, RE, you have GOT to stop snorting up lines of Ajax.

Learning to fly a plane is considerably more difficult than learning to drive. The issue here is that you are making two points.

The statement you made above, regarding split second decisions and what years of OTR training did for you on the Glenn Highway is unarguable. Humans are still for more capable in unpredictable situations requiring split-second judgment than any AI. You were also an experienced driver with many years of rig driving; I'd suggest that a 19 year old girl driving the same vehicle in the same situation would not have fared as well.

But learning to fly a plane is a difficult motherfucker. The testing alone is time consuming, expensive and will screen out the incapable, whereas any neck tattoo who can fog a mirror can get a standard operator's license. You yourself have written about your time in the cab, and what it took to get your CDL, which is entirely different kettle of fish.

• Tweet

[JRM]

Given your response, I suspect you have not thought carefully thorough what it means to allow a car to drive itself in a real world environment.  It should be clear to anyone smart and informed that we humans simply don't have that kind of technology, and won't anytime soon (if ever).  It would require an AI apparatus at least as good as a human driver.  Driving a car is a very, very sophisticated thing.  (Obviously many humans are not up to the task!)

It seems to me like these car companies have been going about it all wrong with the "driverless taxi" approach, both from an AI and a fiscal standpoint.  An "enhanced cruise control" approach seems like it would be much more productive, for use on limited-access highways, combining collision-avoidance radar with passive RFID chips embedded in road markings that would provide information on route, mile marker, lane, and speed limits.  Active RFID chips could broadcast overrides for lane closures and reduced speeds.

This could be sufficiently less expensive to become a standard feature on new cars, especially since the cost of embedding the RFID chips would be passed off to the taxpayers.

These are the most intelligent words on this topic I've seen in a long while.  What is being proposed here is entirely doable, and would likely succeed. But still it contributes to the perpetuation of the general thrust / trend which our familiar approach to technology has been asserting upon the world, and since that trend / thrust has a shelf life of five to ten more years at most I'd rather not invest (concern) even in such fantasies as this one. 

The near term future of transportation is not difficult to imagine. It will be intermodal. This means you can take your bicycle or electric scooter on trains and buses. It will be rail-centric, not car-centric -- which means so very much (like no need for a bazillion parking spaces covered in asphalt everywhere). Few private automobiles will exist in the near term future. They will be relics of a bygone era, like pay phones and 8-track tape players.

The future will be both much more socialistic AND much more anarchist than the present -- which is to say it will be a democracy. Finally. Thank heavens.

And it will again be possible to speak the truth and be heard.

Finally.

Thank heavens.

• Tweet

[azozeo]

I picked up my private pilot ticket at age 18.
Took me 3 trys with an FAA examiner, & that was back before AI was on board.
I'd be lost in a cockpit today. If your into flying, get a ticket to fly ultralights.
Seat of the pants flying will cure your lofty dreams.
Better yet go for a joy ride in a rusky vintage fighter. That'll get you to squirt
chocolate milk outta' yer' nose

• Tweet

[JRM]

That's a feeling you have. The data don't support it.

Flown in a plane in the last ten years or so? They all fly themselves, most of the time. Driving a plane is more complicated than driving a car.

Actually, that's very, very, very far from the case.  You're so very, very, very far from understanding here that if you were as bad a dentist as you are a complexity scientist / philosopher you'd be filling teeth with wads of cotton and sawdust.

Do I really need to explain why? Sheesh. Maybe somebody else here will type it for me. 

Eddy, you're a very, very bright man. But you don't know a think about this tech talk we're having. Not a thing.

• Tweet

[JRM]

Autopilot on a modern airplane is a dream of tranquil simplicity for the algorithms and computer programming involved ... as contrasted with the boisterous jangle and jungle of wild-assed complexity of a typical city or suburban street.  These are not only not in the same ball park they are not even both ball parks. One is not an apple and the other an orange. One is a golf ball and the other is an ecosystem. 

I LIVE in complexity. I know it like one who lives in it. It's crazy difficult unto impossible to include all of the crucial things into a map of it. This I know. 

• Tweet

[jdwheeler42]

Driving a plane is more complicated than driving a car.

No it is not.

Planes have defined levels at which they fly East or West, North or South.  There are many fewer of them in the air at any given time than there are cars on the road in your average city.  Planes do not have bad road conditions to deal with or for the most part drunk pilots in the air.

The roads are a very complex place with a large number of vehicles, many different parameters and constantly changing environment.  You can go from dry to a t-storm in seconds, you can go from decent traction to black ice in nano-seconds.  Programming all those possibilities in advance is about impossible, and the AI is not there to make split second decisions that can save your life.  I know, because I made those decisions, and when I went AIRBORNE off the Glenn Highway, no AI could have done what I did.  Only a well trained Homo Sap driver could have done that.

While you are quite correct about the complexity of driving a car, RE, I think you underestimate the complexity of flying a plane.  I think Eddie is correct that flying is more complex than driving, at the same traffic density.  However, the fact that there are many fewer planes in the air makes it much simpler.  But, the bigger point that Eddie misses is NOT ONE DAMN PLANE FLIES ITSELF, at least in the US.  Maybe over the open ocean.  One that did is likely to get itself shot down.  What really keeps planes safe in the air are the AIR TRAFFIC CONTROLLERS.  They assign the corridors for the planes to follow.  They're the ones making sure the planes don't come close enough to each other to risk collisions.

• Tweet