[LINK] Why you may not own, or drive your vehicle in 10 years time

[Karl Auer]

On Fri, 2016-06-10 at 11:38 +1000, David Lochrin wrote:
> Way back in the 60's artificial intelligence comparable with human
> intelligence was confidently forecast to be a reality in 10 or 15
> years.  Ten years later it was another decade or so away.  We're now
> half a century further on and there are still no AI systems
> comparable with the human mind.

All true. But this discussion has little to do with AI. We are talking
driverless cars. There are many things that we once took to be the
epitome of intelligence which have turned out to be perfectly doable
using nothing but a metric shedload of data and blind processing speed.
I don't expect driverless cars to be "intelligent" except in the
marketing blurbs. Any more than a factory robot or Big Blue is
"intelligent" (though it has to be said that we do keep moving the
goalposts on our hapless candidates for machine "intelligence").

>   Eventually rule-based systems were produced, but they're a
> different thing altogether.  Games such as GO and chess are also in a
> different class because the rules are highly defined and finite, even
> though the combinatorial complexity is very high.

I used these games only as examples of things we said, with great
confidence, would never be done by computers, yet here we are. And IMHO
there is a very great deal of overlap between how those challenges were
met and the requirements for driverless cars.

> And with any road system like the one we have now, no amount of on
> -board AI is going to get around the necessity for real-world
> information outside the immediate road environment and the ability to
> understand its significance.

No, of course not (leaving aside the gratuitous use of "AI"). So that
information will be acquired, in real time, by the vehicle. Unless you
are postulating types of information that "will never be able to be
processed by computers" in which case I refer you to my first paragraph
above.

Again, look at the rain of bodies caused by human drivers on the
world's roads. I don't think humans do that good a job of dealing with
unpredictable events while driving. Or even with predictable events,
for that matter, for some value of "predictable".

>   However driverless technology might be useful in certain highly
> controlled environments.

It's been there for years - in warehouses, hospitals, factories.
Warehouses, hospitals and factories full of unpredictable humans, even.
They don't transport humans, but there's no reason we couldn't have a
driverless hospital gurney, and I'm sure the odd warehouse employee
hitches a ride every now and then down to the far end of the shed.

Heck, a tape silo is pretty much "driverless technology" for tapes. But
I don't think "tightly controlled" is the point here.

> And finally, any such technology has to work in the messy legal,
> political, economic, and social environment of the real world.

Absolutely. Like all the other things we now use computers for and that
have transformed and are transforming our legal, political, economic
and social environments.

Regards, K.

-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Karl Auer (kauer at biplane.com.au)
http://www.biplane.com.au/kauer
http://twitter.com/kauer389

GPG fingerprint: E00D 64ED 9C6A 8605 21E0 0ED0 EE64 2BEE CBCB C38B
Old fingerprint: 3C41 82BE A9E7 99A1 B931 5AE7 7638 0147 2C3C 2AC4