The Car of 2050

[thspfc]

I expect that the car of tomorrow will cost thousands of dollars rather than tens of thousands. The same AI advances that will enable cars to drive themselves will also allow them to be built with almost no human labor. The next opportunity for cost reductions comes from on-board electronics. A Deloitte study from 2017 estimated that a staggering 40% of the cost of a new car is from electronics. The replacement of today's complicated photolithography semiconductor production methods with nanoimprint lithography or even inkjet printing would dramatically slash that cost. The final opportunity for cost reductions comes from the substitution of steel for new alloys and polymers that are stronger and lighter and can be molded as easily as plastic. That would mean many components could be produced as one-piece rather than multiple pieces that have to be welded or bolted together.

Companies won't lower their prices just because they can afford to do so and stay in business. When production costs decrease in a rather inelastic industry such as American cars, it's producer surplus that increases, not consumer surplus.

[kernals12]

I expect that the car of tomorrow will cost thousands of dollars rather than tens of thousands. The same AI advances that will enable cars to drive themselves will also allow them to be built with almost no human labor. The next opportunity for cost reductions comes from on-board electronics. A Deloitte study from 2017 estimated that a staggering 40% of the cost of a new car is from electronics. The replacement of today's complicated photolithography semiconductor production methods with nanoimprint lithography or even inkjet printing would dramatically slash that cost. The final opportunity for cost reductions comes from the substitution of steel for new alloys and polymers that are stronger and lighter and can be molded as easily as plastic. That would mean many components could be produced as one-piece rather than multiple pieces that have to be welded or bolted together.

Companies won't lower their prices just because they can afford to do so and stay in business. When production costs decrease in a rather inelastic industry such as American cars, it's producer surplus that increases, not consumer surplus.

At least one automaker will cut their prices to bring in more sales and the other automakers will follow suit.

[formulanone]

I expect that the car of tomorrow will cost thousands of dollars rather than tens of thousands. The same AI advances that will enable cars to drive themselves will also allow them to be built with almost no human labor. The next opportunity for cost reductions comes from on-board electronics. A Deloitte study from 2017 estimated that a staggering 40% of the cost of a new car is from electronics. The replacement of today's complicated photolithography semiconductor production methods with nanoimprint lithography or even inkjet printing would dramatically slash that cost. The final opportunity for cost reductions comes from the substitution of steel for new alloys and polymers that are stronger and lighter and can be molded as easily as plastic. That would mean many components could be produced as one-piece rather than multiple pieces that have to be welded or bolted together.

Companies won't lower their prices just because they can afford to do so and stay in business. When production costs decrease in a rather inelastic industry such as American cars, it's producer surplus that increases, not consumer surplus.

At least one automaker will cut their prices to bring in more sales and the other automakers will follow suit.

Or lobby like hell to prevent it from being sold in the US.

[kernals12]

I expect that the car of tomorrow will cost thousands of dollars rather than tens of thousands. The same AI advances that will enable cars to drive themselves will also allow them to be built with almost no human labor. The next opportunity for cost reductions comes from on-board electronics. A Deloitte study from 2017 estimated that a staggering 40% of the cost of a new car is from electronics. The replacement of today's complicated photolithography semiconductor production methods with nanoimprint lithography or even inkjet printing would dramatically slash that cost. The final opportunity for cost reductions comes from the substitution of steel for new alloys and polymers that are stronger and lighter and can be molded as easily as plastic. That would mean many components could be produced as one-piece rather than multiple pieces that have to be welded or bolted together.

Companies won't lower their prices just because they can afford to do so and stay in business. When production costs decrease in a rather inelastic industry such as American cars, it's producer surplus that increases, not consumer surplus.

At least one automaker will cut their prices to bring in more sales and the other automakers will follow suit.

Or lobby like hell to prevent it from being sold in the US.

How many different automakers are there in the US today? One of the biggest ones, Tesla, didn't even exist 20 years ago.

[SectorZ]

Kernals is the type that watches Demolition Man and thinks its an awesome future. We'll be in our self-driving cars 30 years from now, going to Taco Bell, singing along to Ozempic ads emanating from our car stereos.

[Max Rockatansky]

Kernals is the type that watches Demolition Man and thinks its an awesome future. We'll be in our self-driving cars 30 years from now, going to Taco Bell, singing along to Ozempic ads emanating from our car stereos.

I wonder what his boggle is?

[formulanone]

I expect that the car of tomorrow will cost thousands of dollars rather than tens of thousands. The same AI advances that will enable cars to drive themselves will also allow them to be built with almost no human labor. The next opportunity for cost reductions comes from on-board electronics. A Deloitte study from 2017 estimated that a staggering 40% of the cost of a new car is from electronics. The replacement of today's complicated photolithography semiconductor production methods with nanoimprint lithography or even inkjet printing would dramatically slash that cost. The final opportunity for cost reductions comes from the substitution of steel for new alloys and polymers that are stronger and lighter and can be molded as easily as plastic. That would mean many components could be produced as one-piece rather than multiple pieces that have to be welded or bolted together.

Companies won't lower their prices just because they can afford to do so and stay in business. When production costs decrease in a rather inelastic industry such as American cars, it's producer surplus that increases, not consumer surplus.

At least one automaker will cut their prices to bring in more sales and the other automakers will follow suit.

Or lobby like hell to prevent it from being sold in the US.

How many different automakers are there in the US today? One of the biggest ones, Tesla, didn't even exist 20 years ago.

That's true, they're one of the few manufacturers to actually continuously make vehicles for the last 15 years. But prices have not come down significantly despite the competition (that's also a function of inflation, unyielding consumer demand, supply costs, labor, greed, et cetera).

Automakers, suppliers, marketers, personal tastes, and dealerships are keen to upsell ever larger vehicles and they always cost more than something smaller.

[kalvado]

I expect that the car of tomorrow will cost thousands of dollars rather than tens of thousands. The same AI advances that will enable cars to drive themselves will also allow them to be built with almost no human labor. The next opportunity for cost reductions comes from on-board electronics. A Deloitte study from 2017 estimated that a staggering 40% of the cost of a new car is from electronics. The replacement of today's complicated photolithography semiconductor production methods with nanoimprint lithography or even inkjet printing would dramatically slash that cost. The final opportunity for cost reductions comes from the substitution of steel for new alloys and polymers that are stronger and lighter and can be molded as easily as plastic. That would mean many components could be produced as one-piece rather than multiple pieces that have to be welded or bolted together.

Companies won't lower their prices just because they can afford to do so and stay in business. When production costs decrease in a rather inelastic industry such as American cars, it's producer surplus that increases, not consumer surplus.

At least one automaker will cut their prices to bring in more sales and the other automakers will follow suit.

Or lobby like hell to prevent it from being sold in the US.

How many different automakers are there in the US today? One of the biggest ones, Tesla, didn't even exist 20 years ago.

That's true, they're one of the few manufacturers to actually continuously make vehicles for the last 15 years. But prices have not come down significantly despite the competition (that's also a function of inflation, unyielding consumer demand, supply costs, labor, greed, et cetera).

Last time I heard, tesla was selling cars at a loss...

[GaryV]

Kernals is the type that watches Demolition Man and thinks its an awesome future. We'll be in our self-driving cars 30 years from now, going to Taco Bell, singing along to Ozempic ads emanating from our car stereos.

AM or FM?

[SectorZ]

Kernals is the type that watches Demolition Man and thinks its an awesome future. We'll be in our self-driving cars 30 years from now, going to Taco Bell, singing along to Ozempic ads emanating from our car stereos.

AM or FM?

Streaming in Adify.

[Scott5114]

I remember futurist crap from the 1980s.  Same bat time, same bat channel...

Which more or less was just a rehash of 1950s futurism.  I did enjoy thinking 2015 was going to be way more interesting than it eventually turned out as a kid though.

It's now 2023 and we have Wegovy and ChatGPT. Is it so hard to believe in rapid scientific progress?

The neural network models used in modern AI are not a panacea. It's important to remember that what is popularly called AI is not actually intelligence operating the same way that human intelligence does. While they work in some limited circumstances, modern AI algorithms are limited to probabilistic prediction and do not have the capability to handle anything novel that isn't a synthesis of information already in the training corpus. Asking for one to handle anything that wasn't included in the training corpus generally causes them to fall down in a pretty spectacular fashion. Using one in a safety-critical application like full self-driving introduces an unacceptable level of risk in a situation when the machine encounters a blind spot in the training corpus. And such a blind spot is nigh-inevitable—there is not enough money in the world to pay for the human labor that would be needed to compile a corpus consisting of every possible situation that may arise on the road.

This is why I inquired upthread as to your programming experience. It would seem you intentionally dodged the question, since answering in the negative would underscore your unfamiliarity with what is actually going on "under the hood" with modern AI, as well as what is and isn't generally possible with software.

Driverless cars don't need to be 100% safe, they just need to be noticeably safer than human drivers.

As someone who dabbles in programming as part of my job, I feel fairly secure in stating that this is probably not going to be achievable without some sort of groundbreaking revolution in the way software is written. There are just too many problems involved in driving that are impossible to solve procedurally.

[thspfc]

I expect that the car of tomorrow will cost thousands of dollars rather than tens of thousands. The same AI advances that will enable cars to drive themselves will also allow them to be built with almost no human labor. The next opportunity for cost reductions comes from on-board electronics. A Deloitte study from 2017 estimated that a staggering 40% of the cost of a new car is from electronics. The replacement of today's complicated photolithography semiconductor production methods with nanoimprint lithography or even inkjet printing would dramatically slash that cost. The final opportunity for cost reductions comes from the substitution of steel for new alloys and polymers that are stronger and lighter and can be molded as easily as plastic. That would mean many components could be produced as one-piece rather than multiple pieces that have to be welded or bolted together.

Companies won't lower their prices just because they can afford to do so and stay in business. When production costs decrease in a rather inelastic industry such as American cars, it's producer surplus that increases, not consumer surplus.

At least one automaker will cut their prices to bring in more sales and the other automakers will follow suit.

Company A drops their price by $5000. Company A's sales spike for a bit, but companies B, C, D, E, F, G, and H soon match the price. Now company A is selling the same number of cars again, but at a lower profit margin. Less money.

[CtrlAltDel]

There are just too many problems involved in driving that are impossible to solve procedurally.

What do you have in mind here? And how are they solved with people drivers?

[CtrlAltDel]

It is certainly possible that everything there is to know about batteries is known.

No, it isn't. There are people out there working right now on things we haven't figured out yet.

[Scott5114]

There are just too many problems involved in driving that are impossible to solve procedurally.

What do you have in mind here? And how are they solved with people drivers?

It's hard, but possible, to teach a computer what a traffic cone is. Once it knows that, it's easy to tell it what to do when it sees a traffic cone. It is very difficult to tell it that it can safely ignore cones being hauled in the back of a DOT truck without getting it to also ignore cones it shouldn't.

A human doesn't even consider dodging to avoid a cone in the back of a DOT truck because human brains reflexively understand context, and procedural code does not.

[kalvado]

There are just too many problems involved in driving that are impossible to solve procedurally.

What do you have in mind here? And how are they solved with people drivers?

It's hard, but possible, to teach a computer what a traffic cone is. Once it knows that, it's easy to tell it what to do when it sees a traffic cone. It is very difficult to tell it that it can safely ignore cones being hauled in the back of a DOT truck without getting it to also ignore cones it shouldn't.

A human doesn't even consider dodging to avoid a cone in the back of a DOT truck because human brains reflexively understand context, and procedural code does not.

While I am not an expert, AI - neural networks - are not procedural. They are, from my understanding, some flavor of black box filled with black magic. Not unlike human brain, actually - except for those being filled with gray matter.

[kernals12]

I expect that the car of tomorrow will cost thousands of dollars rather than tens of thousands. The same AI advances that will enable cars to drive themselves will also allow them to be built with almost no human labor. The next opportunity for cost reductions comes from on-board electronics. A Deloitte study from 2017 estimated that a staggering 40% of the cost of a new car is from electronics. The replacement of today's complicated photolithography semiconductor production methods with nanoimprint lithography or even inkjet printing would dramatically slash that cost. The final opportunity for cost reductions comes from the substitution of steel for new alloys and polymers that are stronger and lighter and can be molded as easily as plastic. That would mean many components could be produced as one-piece rather than multiple pieces that have to be welded or bolted together.

Companies won't lower their prices just because they can afford to do so and stay in business. When production costs decrease in a rather inelastic industry such as American cars, it's producer surplus that increases, not consumer surplus.

At least one automaker will cut their prices to bring in more sales and the other automakers will follow suit.

Company A drops their price by $5000. Company A's sales spike for a bit, but companies B, C, D, E, F, G, and H soon match the price. Now company A is selling the same number of cars again, but at a lower profit margin. Less money.

Nor necessarily. If cars are cheaper, people will replace their cars more frequently

[SectorZ]

I expect that the car of tomorrow will cost thousands of dollars rather than tens of thousands. The same AI advances that will enable cars to drive themselves will also allow them to be built with almost no human labor. The next opportunity for cost reductions comes from on-board electronics. A Deloitte study from 2017 estimated that a staggering 40% of the cost of a new car is from electronics. The replacement of today's complicated photolithography semiconductor production methods with nanoimprint lithography or even inkjet printing would dramatically slash that cost. The final opportunity for cost reductions comes from the substitution of steel for new alloys and polymers that are stronger and lighter and can be molded as easily as plastic. That would mean many components could be produced as one-piece rather than multiple pieces that have to be welded or bolted together.

Companies won't lower their prices just because they can afford to do so and stay in business. When production costs decrease in a rather inelastic industry such as American cars, it's producer surplus that increases, not consumer surplus.

At least one automaker will cut their prices to bring in more sales and the other automakers will follow suit.

Company A drops their price by $5000. Company A's sales spike for a bit, but companies B, C, D, E, F, G, and H soon match the price. Now company A is selling the same number of cars again, but at a lower profit margin. Less money.

Nor necessarily. If cars are cheaper, people will replace their cars more frequently

Well that is the goal with futurists, make everything more disposable. Maybe just let people lease cars at this point. You will own nothing and like it.

[thspfc]

I expect that the car of tomorrow will cost thousands of dollars rather than tens of thousands. The same AI advances that will enable cars to drive themselves will also allow them to be built with almost no human labor. The next opportunity for cost reductions comes from on-board electronics. A Deloitte study from 2017 estimated that a staggering 40% of the cost of a new car is from electronics. The replacement of today's complicated photolithography semiconductor production methods with nanoimprint lithography or even inkjet printing would dramatically slash that cost. The final opportunity for cost reductions comes from the substitution of steel for new alloys and polymers that are stronger and lighter and can be molded as easily as plastic. That would mean many components could be produced as one-piece rather than multiple pieces that have to be welded or bolted together.

Companies won't lower their prices just because they can afford to do so and stay in business. When production costs decrease in a rather inelastic industry such as American cars, it's producer surplus that increases, not consumer surplus.

At least one automaker will cut their prices to bring in more sales and the other automakers will follow suit.

Company A drops their price by $5000. Company A's sales spike for a bit, but companies B, C, D, E, F, G, and H soon match the price. Now company A is selling the same number of cars again, but at a lower profit margin. Less money.

Nor necessarily. If cars are cheaper, people will replace their cars more frequently

But they won't be cheaper.

[hotdogPi]

Computer prices went way down from 1980 to 2000. Why not cars?

[Max Rockatansky]

Computers were largely a new consumer item in the 1980s.  I would expect EV prices to fall over time as the demand increases and technology improves, probably not so much for the average car though.

[thspfc]

Computer prices went way down from 1980 to 2000. Why not cars?

Because demand for computers at that time was far more elastic than demand for cars is now. (Elastic meaning subject to change based on price.)

Computers at that time were a want, not a need, for most people. Cars, in the US, are seen as a need for almost everyone except those in a handful of urban areas such as New York.

The reason why this matters is because when demand is more elastic, companies can increase profits by lowering the price, as the increased number of units sold makes up for a smaller margin. Therefore they lower the price. In the case of American cars, demand is less elastic, and the increased cars sold at a lower price likely would not make up for the smaller margin (obviously I'm not running the numbers at Ford or GM but it's pretty obvious that cars are seen as a need, and almost all people who need one already have one.)

Another example along the lines of early computers is flying. When widespread flying became possible, it was in airlines' best interest to decrease prices because at the previous price point it was only available to the richest people. By making it financially available to the middle class, they made more money in total.

[kalvado]

Computer prices went way down from 1980 to 2000. Why not cars?

Because demand for computers at that time was far more elastic than demand for cars is now. (Elastic meaning subject to change based on price.)

Computers at that time were a want, not a need, for most people. Cars, in the US, are seen as a need for almost everyone except those in a handful of urban areas such as New York.

The reason why this matters is because when demand is more elastic, companies can increase profits by lowering the price, as the increased number of units sold makes up for a smaller margin. Therefore they lower the price. In the case of American cars, demand is less elastic, and the increased cars sold at a lower price likely would not make up for the smaller margin (obviously I'm not running the numbers at Ford or GM but it's pretty obvious that cars are seen as a need, and almost all people who need one already have one.)

Another example along the lines of early computers is flying. When widespread flying became possible, it was in airlines' best interest to decrease prices because at the previous price point it was only available to the richest people. By making it financially available to the middle class, they made more money in total.

oh, and you never heard about airline deregulation?

As for demand elasticity for cars, there are options for pushing for longer service time. That provides enough short-term elasticity.
We certainly bought both our current  cars while previous ones had fixable problems - but were coming to the point of "is it cost-efficient to fix?". 

[Scott5114]

There are just too many problems involved in driving that are impossible to solve procedurally.

What do you have in mind here? And how are they solved with people drivers?

It's hard, but possible, to teach a computer what a traffic cone is. Once it knows that, it's easy to tell it what to do when it sees a traffic cone. It is very difficult to tell it that it can safely ignore cones being hauled in the back of a DOT truck without getting it to also ignore cones it shouldn't.

A human doesn't even consider dodging to avoid a cone in the back of a DOT truck because human brains reflexively understand context, and procedural code does not.

While I am not an expert, AI - neural networks - are not procedural. They are, from my understanding, some flavor of black box filled with black magic. Not unlike human brain, actually - except for those being filled with gray matter.

At a low enough level, all code is procedural, even black magic, because processors are procedural.

This is how an AI neural network works: The computer uses a random number generator to produce a few random procedures, then runs them and produces the results of each for a human to inspect. The human selects the one that produced the most correct result. The computer then uses the RNG to produce a few random variations on that procedure, and presents the human with the results of each. The human picks the most correct one, and so on and so forth. Eventually you end up with a super complicated procedure that looks like magic, and that's what ships as "the AI" (or more formally as a checkpoint). But if you break it open it's a bunch of if/then do/while like anything else, it's just way more complex than a human could have written unassisted.

If you ever get a chance to fool around with the Stable Diffusion AI image generator, you can see this play out in real time, as it uses a similar algorithm to generate its images (it starts with a picture of white noise, and then iteratively selects the variation that best matches the prompt, so you can see the image take shape out of the static).

[Road Hog]

The car of 1950 isn't much different than the car of 2000. They both have four wheels, a steering wheel, a transmission, a driveshaft, a windshield, a speedometer, and an engine.

The exterior stylings are clearly different as well as the engineering improvements, but that is all largely background. My uncle, a mechanic's mechanic, tried and failed to get a 1942 Ford pickup running but the interior wasn't much different from a modern car.