Fully automated vehicles - 'not in our lifetime'?

antrobus

Malthusian wrote: »

Ah, the notoriously infallible Three Laws of Robotics that failed in every single Asimov story about robots. .

That's because they were simply a device to drive Mr Asimov's stories. And there in fact four laws, now I come to think of it, He added another one later one.

Malthusian wrote: »

,,,
Saying that the Three Laws of Robotics is a great guide for AI research is like saying that Frankenstein is a great guide for constructing humans out of dead body parts. Or that we should follow Metropolis' example of how to build a modern city. It suggests you haven't read the book or seen the film. .

You obviously missed the significance of that symbol. Perhaps in future we all need to add the phrase; 'P.S. this is a joke' in order to make things doubly clear.

boliston

Boler1985 wrote: »

A single seater car would take up a 3rd of the space of a 5-seater and therefore ease congestion considerably. Probably about a 3rd of the weight and way way more energy efficient.

When the weather is good you could even have an open top single seater car that you can pedal - even more fuel efficient and say goodbye to gym fees!

Herzlos

Enterprise_1701C wrote: »

Yet.

Things change at a greater rate than anyone can ever truly predict.

I'll be surprised if we can go down the subatomic transistor route, because they'll be a nightmare to deal with, if at all. We'll hit some floor on how small a viable processor can be, at which point we'll need to see them get bigger to get more performance, somehow.



I'm referring largely to Moores law, which hasn't really been true for a while now. We're not doubling transistor counts anymore on a single chip, instead moving more to networks of processors.



They'll continue to get better, sure, but I'm not convinced this exponential growth will continue.

westernpromise

GreatApe wrote: »

it is much more reasonable to assume humans brains are not the peak of all possible intelligences that we will be surpassed.

You're assuming that humans are smart enough to design a computer that's smart enough to design other computers that are smarter than itself and humans.

Why would that be so?

The opposite is likelier. Moore's law always fizzles out. In the first 60 years of flight we went from flying 200 feet at 30mph to flying 3,000 miles at Mach 2. In the subsequent 60 years we made almost no further progress in performance. What improvement was cost, uptake, that sort of thing.

The limit isn't what we can think up, it's what we can actually build that's worth building. Just because we can think up a time traveling police box or a flying carpet or genetically recreated dinosaurs, that doesn't mean any of it will happen.

Montgolfier flew in a balloon in the 1700s. In the 1800s Edgar Allan Poe was writing about flying to the moon in a really big, really good balloon. Didn't happen.

kabayiri

Herzlos wrote: »

I'll be surprised if we can go down the subatomic transistor route, because they'll be a nightmare to deal with, if at all. We'll hit some floor on how small a viable processor can be, at which point we'll need to see them get bigger to get more performance, somehow.



I'm referring largely to Moores law, which hasn't really been true for a while now. We're not doubling transistor counts anymore on a single chip, instead moving more to networks of processors.



They'll continue to get better, sure, but I'm not convinced this exponential growth will continue.

And all of this has precisely nothing to do with automated vehicles.

I think we are well off point here.

westernpromise

The supposition was that automated vehicles that don't work today will work in the future because technology. If technology does not after all improve exponentially forever, this won't happen.

If technology did improve exponentially forever, we'd be in touch with time travelling aliens by now. Given the age and scale of the universe, perpetually exponential technology improvement anywhere would have meant someone from somewhere would have rocked up here by now.

GreatApe

Herzlos wrote: »

I'll be surprised if we can go down the subatomic transistor route, because they'll be a nightmare to deal with, if at all. We'll hit some floor on how small a viable processor can be, at which point we'll need to see them get bigger to get more performance, somehow.

I'm referring largely to Moores law, which hasn't really been true for a while now. We're not doubling transistor counts anymore on a single chip, instead moving more to networks of processors.

They'll continue to get better, sure, but I'm not convinced this exponential growth will continue.

Transistors don't necessarily have to get smaller. They could be built closer together.
Intel seems to manage to get more transistors onto the same die size at the same process node than the other fabs. There could be scope to do that for some years.

You could then just build bigger chips and 3D layer chips.
A single meter cube has a billion cubic millimeters. Give half for cooling and you have the space for 500 million cubic millimetres. A typical chip is what 100 cubic millimeters? So you could go up another 5 million fold from today by increasing the chip size to one cubic meter.

Sure that would require massive engineering challenges but that 5 million fold increase takes us another 60+ years of a very high exponential. I don't see why it would be impossible to make chips that size in the 2080s you don't hit any physical limits just building bigger chips.

Or we might jump to another medium like light or another material

Or what I think is most likely is that we will have dedicated ASICs which are thousands times more powerful

GreatApe

westernpromise wrote: »

You're assuming that humans are smart enough to design a computer that's smart enough to design other computers that are smarter than itself and humans.

Why would that be so?

The opposite is likelier. Moore's law always fizzles out. In the first 60 years of flight we went from flying 200 feet at 30mph to flying 3,000 miles at Mach 2. In the subsequent 60 years we made almost no further progress in performance. What improvement was cost, uptake, that sort of thing.

The limit isn't what we can think up, it's what we can actually build that's worth building. Just because we can think up a time traveling police box or a flying carpet or genetically recreated dinosaurs, that doesn't mean any of it will happen.

Montgolfier flew in a balloon in the 1700s. In the 1800s Edgar Allan Poe was writing about flying to the moon in a really big, really good balloon. Didn't happen.

But we already have what appears to be the first true AIs

GreatApe

westernpromise wrote: »

The supposition was that automated vehicles that don't work today will work in the future because technology. If technology does not after all improve exponentially forever, this won't happen.

If technology did improve exponentially forever, we'd be in touch with time travelling aliens by now. Given the age and scale of the universe, perpetually exponential technology improvement anywhere would have meant someone from somewhere would have rocked up here by now.

Yes that's right so either we are the first to get there or the others have decided not to show themselves. Most likely we are the first that might sound arrogant but of the many hundreds of millions of species to have exited only one has mastered technology.

Also we don't need to exponentially improve the number of transistors on a given sized chip we could just move from that to just building more transistors on larger chips. That is what the super intelligence will likely do. It will exponentially consume almost all the matter in the milkyway Galaxy.

GreatApe

westernpromise wrote: »

The supposition was that automated vehicles that don't work today will work in the future because technology. If technology does not after all improve exponentially forever, this won't happen.

If technology did improve exponentially forever, we'd be in touch with time travelling aliens by now. Given the age and scale of the universe, perpetually exponential technology improvement anywhere would have meant someone from somewhere would have rocked up here by now.

We don't need AI to do exponential

You could imagine humans create a computer which replicates itself. A dumb non AI computer. It just knows how to mine refine build and repeat. Release it and it will exponentially multiply. Give it the ability to build rockets and it can do the same for the solar system and then the milky way etc

This is called the fermi Paradox and it suggests we humans are the first to get to this stage that there are no intelligent aliens at least not in nearby galaxies