I was wrong about robocars

[Ken from Chicago]

I thought robocars wouldn't been on the streets until 2025.

https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/ijuvQ9oJpuUJ

Maybe some prototypes by 2020.

Maybe even some on test tracks as early as 2015.

I was wrong:

http://news.cnet.com/8301-11386_3-57406639-76/google-self-driving-car-chauffeurs-legally-blind-man/

(Here's a shorter link:)

http://www.pcmag.com/article2/0,2817,2402340,00.asp

-- Ken from Chicago

[Duggy]

On Mar 30, 7:59 am, "Ken from Chicago" <kwicker1b_nos...@comcast.net>
wrote:

> I thought robocars wouldn't been on the streets until 2025.
>

> https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/ijuvQ...

>
> Maybe some prototypes by 2020.
>
> Maybe even some on test tracks as early as 2015.
>
> I was wrong:
>

> http://news.cnet.com/8301-11386_3-57406639-76/google-self-driving-car...

>
> (Here's a shorter link:)
>
> http://www.pcmag.com/article2/0,2817,2402340,00.asp

Weren't they around about 2 or 3 years ago?

===
= DUG.
===

[Lynn McGuire]

Maybe they will be better drivers than most of the
drivers around Houston ...

Lynn

[Ken from Chicago]

"Lynn McGuire" <l...@winsim.com> wrote in message
news:jl2ps3$nuv$1...@dont-email.me...

Not everyone in Houston has driven on snow or ice before.

-- Ken from Chicago (who still chortles when the state pert near shuts down
when they get a quarter inch of snow)

[Gutless Umbrella Carrying Sissy]

"Ken from Chicago" <kwicker1...@comcast.net> wrote in
news:jl2qk0$2dk$1...@dont-email.me:

That global warming's gonna kill us all.

--
Terry Austin

"Terry Austin: like the polio vaccine, only with more asshole."
-- David Bilek

Jesus forgives sinners, not criminals.

[Gutless Umbrella Carrying Sissy]

"Ken from Chicago" <kwicker1...@comcast.net> wrote in

news:jl2lvp$4mk$1...@dont-email.me:

Unless, of course, something happens that the computer doesn't
expect, and can't cope with. What are the odds of that, on a random
public street?

Plus, they're still in the alpha stage anyway.

[Joseph Nebus]

In <XnsA025AC15865...@69.16.186.7> Gutless Umbrella Carrying Sissy <taus...@gmail.com> writes:

>"Ken from Chicago" <kwicker1...@comcast.net> wrote in
>news:jl2lvp$4mk$1...@dont-email.me:

>> (Here's a shorter link:)
>>
>> http://www.pcmag.com/article2/0,2817,2402340,00.asp
>>
>Unless, of course, something happens that the computer doesn't
>expect, and can't cope with. What are the odds of that, on a random
>public street?

>Plus, they're still in the alpha stage anyway.

Surely, the existence of self-driving cars in an acceptable
alpha stage demonstrates that they will never be able to function as
a generally distributed technology.

--
http://nebusresearch.wordpress.com/ Joseph Nebus
Current Entry: How To Forget The Area Of A Trapezoid http://wp.me/p1RYhY-9d
------------------------------------------------------------------------------

[Sea Wasp (Ryk E. Spoor)]

On 3/29/12 5:59 PM, Ken from Chicago wrote:
> I thought robocars wouldn't been on the streets until 2025.
>
> https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/ijuvQ9oJpuUJ
>
>
> Maybe some prototypes by 2020.
>
> Maybe even some on test tracks as early as 2015.
>
> I was wrong:

That's a severely constrained test under very controlled conditions --
and about a thousand things could go wrong with it even then. *I*
wouldn't have taken that risk, unless they had a police guard clearing
the road ahead.

It's a nice stunt, but automated cars have a *LONG* way to go.


--
Sea Wasp
/^\
;;;
Website: http://www.grandcentralarena.com Blog:
http://seawasp.livejournal.com

[Your Name]

In article <XnsA025AC15865...@69.16.186.7>, Gutless Umbrella

Carrying Sissy <taus...@gmail.com> wrote:
> "Ken from Chicago" <kwicker1...@comcast.net> wrote in
> news:jl2lvp$4mk$1...@dont-email.me:
> >
> > I thought robocars wouldn't been on the streets until 2025.
> >
> > https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/i
> > juvQ9oJpuUJ
> >
> > Maybe some prototypes by 2020.
> >
> > Maybe even some on test tracks as early as 2015.
> >
> > I was wrong:
> >
> > http://news.cnet.com/8301-11386_3-57406639-76/google-self-driving
> > -car-chauffeurs-legally-blind-man/
> >
> > (Here's a shorter link:)
> >
> > http://www.pcmag.com/article2/0,2817,2402340,00.asp
>
> Unless, of course, something happens that the computer doesn't
> expect, and can't cope with. What are the odds of that, on a random
> public street?
>
> Plus, they're still in the alpha stage anyway.

If Microsoft have ANY involvement, the car simply won't start somedays,
won't stop others, freeze in the middle of a journey, the tyres will
auto-deflate every time you put refuel / recharge it, ...


One of the biggest problems for auto-drive cars the useless GPS systems
that send you the wrong way up one way streets, through fields or lakes,
off cliffs, on huge detours (including going through different countries
just to get to the other side of the same city you're already in).

[Greg Goss]

"Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote:

>On 3/29/12 5:59 PM, Ken from Chicago wrote:
>> I thought robocars wouldn't been on the streets until 2025.
>>
>> https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/ijuvQ9oJpuUJ
>>
>>
>> Maybe some prototypes by 2020.
>>
>> Maybe even some on test tracks as early as 2015.
>>
>> I was wrong:
>
> That's a severely constrained test under very controlled conditions --
>and about a thousand things could go wrong with it even then. *I*
>wouldn't have taken that risk, unless they had a police guard clearing
>the road ahead.

Notice that the local police department gets a co-credit at the end.
--
I used to own a mind like a steel trap.
Perhaps if I'd specified a brass one, it
wouldn't have rusted like this.

[Ken from Chicago]

"Gutless Umbrella Carrying Sissy" <taus...@gmail.com> wrote in message
news:XnsA025AC15865...@69.16.186.7...

> "Ken from Chicago" <kwicker1...@comcast.net> wrote in
> news:jl2lvp$4mk$1...@dont-email.me:
>
>> I thought robocars wouldn't been on the streets until 2025.
>>
>> https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/i
>> juvQ9oJpuUJ
>>
>> Maybe some prototypes by 2020.
>>
>> Maybe even some on test tracks as early as 2015.
>>
>> I was wrong:
>>
>> http://news.cnet.com/8301-11386_3-57406639-76/google-self-driving
>> -car-chauffeurs-legally-blind-man/
>>
>> (Here's a shorter link:)
>>
>> http://www.pcmag.com/article2/0,2817,2402340,00.asp
>>
> Unless, of course, something happens that the computer doesn't
> expect, and can't cope with. What are the odds of that, on a random
> public street?

So it speeds up, swerves, stops or reverses direction.

> Plus, they're still in the alpha stage anyway.

Yes, but I didn't expect it to be on public streets this soon.

> --
> Terry Austin

-- Ken from Chicago

[Ken from Chicago]

"Your Name" <Your...@YourISP.com> wrote in message
news:YourName-300...@203-118-187-140.dsl.dyn.ihug.co.nz...

> In article <XnsA025AC15865...@69.16.186.7>, Gutless Umbrella
> Carrying Sissy <taus...@gmail.com> wrote:
>> "Ken from Chicago" <kwicker1...@comcast.net> wrote in
>> news:jl2lvp$4mk$1...@dont-email.me:
>> >
>> > I thought robocars wouldn't been on the streets until 2025.
>> >
>> > https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/i
>> > juvQ9oJpuUJ
>> >
>> > Maybe some prototypes by 2020.
>> >
>> > Maybe even some on test tracks as early as 2015.
>> >
>> > I was wrong:
>> >
>> > http://news.cnet.com/8301-11386_3-57406639-76/google-self-driving
>> > -car-chauffeurs-legally-blind-man/
>> >
>> > (Here's a shorter link:)
>> >
>> > http://www.pcmag.com/article2/0,2817,2402340,00.asp
>>
>> Unless, of course, something happens that the computer doesn't
>> expect, and can't cope with. What are the odds of that, on a random
>> public street?
>>
>> Plus, they're still in the alpha stage anyway.
>
> If Microsoft have ANY involvement, the car simply won't start somedays,
> won't stop others, freeze in the middle of a journey, the tyres will
> auto-deflate every time you put refuel / recharge it, ...

Google and Microsoft aren't exactly the best of friends. Microsoft lawsuit
against Android phone makers doesn't help matters.

> One of the biggest problems for auto-drive cars the useless GPS systems
> that send you the wrong way up one way streets, through fields or lakes,
> off cliffs, on huge detours (including going through different countries
> just to get to the other side of the same city you're already in).

What percentage of trips do they do that? Also, aren't those being updated?

-- Ken from Chicago

[Ken from Chicago]

"Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote in message
news:jl3fbk$lsb$1...@dont-email.me...

> On 3/29/12 5:59 PM, Ken from Chicago wrote:
>> I thought robocars wouldn't been on the streets until 2025.
>>
>> https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/ijuvQ9oJpuUJ
>>
>>
>> Maybe some prototypes by 2020.
>>
>> Maybe even some on test tracks as early as 2015.
>>
>> I was wrong:
>
> That's a severely constrained test under very controlled conditions --

It's still on a public street, which further along than I suspected.

> and about a thousand things could go wrong with it even then. *I* wouldn't
> have taken that risk, unless they had a police guard clearing the road
> ahead.

The end credit included local police.

> It's a nice stunt, but automated cars have a *LONG* way to go.
>
>
> --
> Sea Wasp
> /^\
> ;;; Website: http://www.grandcentralarena.com
> Blog: http://seawasp.livejournal.com

-- Ken from Chicago

[Sea Wasp (Ryk E. Spoor)]

On 3/30/12 1:53 AM, Ken from Chicago wrote:
> "Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote in message
> news:jl3fbk$lsb$1...@dont-email.me...
>> On 3/29/12 5:59 PM, Ken from Chicago wrote:
>>> I thought robocars wouldn't been on the streets until 2025.
>>>
>>> https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/ijuvQ9oJpuUJ
>>>
>>>
>>>
>>> Maybe some prototypes by 2020.
>>>
>>> Maybe even some on test tracks as early as 2015.
>>>
>>> I was wrong:
>>
>> That's a severely constrained test under very controlled conditions --
>
> It's still on a public street, which further along than I suspected.
>
>> and about a thousand things could go wrong with it even then. *I*
>> wouldn't have taken that risk, unless they had a police guard clearing
>> the road ahead.
>
> The end credit included local police.

Which would indicate it's not actually close to being a real robocar.

I will be impressed when your robocar can deal with situations *I* have
dealt with in traffic. Then I *might* consider it getting close to
deployable (might, because that would indicate that it was about as good
as me, but I'm not perfect, and if you want to replace humans with
robots, the robots have to be BETTER (because for damn sure they'll be
expensive).

[David Dyer-Bennet]

Have you actually read the articles on this project? They've been
driving around on public streets for years -- with human co-pilots.

The police were necessary this time, I believe, for *legal* reasons;
there was no licensed driver in the car, and no legal provision for an
automated car.

But it's been driving on real streets with other cars, without the human
overriding, for quite a while now.
--
David Dyer-Bennet, dd...@dd-b.net; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info

[Cryptoengineer]

On Mar 30, 7:53 am, "Sea Wasp (Ryk E. Spoor)"

<seaw...@sgeinc.invalid.com> wrote:
> On 3/30/12 1:53 AM, Ken from Chicago wrote:
>
>
>
>
>
>
>
>
>

> > "Sea Wasp (Ryk E. Spoor)" <seaw...@sgeinc.invalid.com> wrote in message

> >news:jl3fbk$lsb$1...@dont-email.me...
> >> On 3/29/12 5:59 PM, Ken from Chicago wrote:
> >>> I thought robocars wouldn't been on the streets until 2025.
>

> >>>https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/ijuvQ...

>
> >>> Maybe some prototypes by 2020.
>
> >>> Maybe even some on test tracks as early as 2015.
>
> >>> I was wrong:
>
> >> That's a severely constrained test under very controlled conditions --
>
> > It's still on a public street, which further along than I suspected.
>
> >> and about a thousand things could go wrong with it even then. *I*
> >> wouldn't have taken that risk, unless they had a police guard clearing
> >> the road ahead.
>
> > The end credit included local police.
>
>         Which would indicate it's not actually close to being a real robocar.
>
>         I will be impressed when your robocar can deal with situations *I* have
> dealt with in traffic. Then I *might* consider it getting close to
> deployable (might, because that would indicate that it was about as good
> as me, but I'm not perfect, and if you want to replace humans with
> robots, the robots have to be BETTER (because for damn sure they'll be
> expensive).

I strongly suspect we'll see them on freeways long before they are
routinely used on city streets, and then in special robo-only lanes.

There are already options on some cars cars for cruise control slaved
to laser or radar range finders, to maintain a constant separation,
and cars that watch lane markers to warn drivers when they wander.

Combine the two, and you're very close to being able to put a car on
'follow' for long haul driving.

It will take some more technology to mature the system; I expect
including GPS and inter-car short range mesh networking so cars can
cooperate, and react to issues that are out of the line of sight. All
this is also well into development.

Dealing with unexpected non-car issues are a major challenge; the
classic kid-runs-out-between-parked-cars situation for one. What is
the acceptable level of failure? This is why I expect we'll see smart
cars (and trucks!) on limited access roads first.

Liability issues will be huge as well; if a smart car is involved in
an accident, who is liable? At the start, I have no doubt that we'll
require a licensed driver at the wheel, ready to override the system
instantly; and who is supposed to have his/her attention on the road.

But later, if the systems seem at least as reliable as human drivers,
there will be huge pressure to allow them to be used without a driver
at the wheel.

pt

[Joseph Nebus]

Look, all this evidence of cars driving themselves is no
evidence that cars will ever be able to drive themselves, and saying
that it's evidence that cars may be able to drive themselves is just
distracting the public from the need to sponsor the manned space
programs and the helium-3 technology development essential to making
cars that can drive themselves.

[Sea Wasp (Ryk E. Spoor)]

Without them overriding EVER? Or just without them overriding when
things are going well?

If they really are that good, why does the DARPA challenge exist at
all? It would seem that Google should have long since finished that
sucker off.

[Sea Wasp (Ryk E. Spoor)]

All these STATEMENTS that the cars have driven themselves without being
overridden would appear to be at odds with the state of autonomous
vehicle research as I have come to know it -- fairly well -- since
having to do work IN that field.

If Google really has cars that can be let loose to drive themselves in
real traffic, and don't need to be overridden by human beings, why is it
that the Army is still having significant research being done just to
get a damn convoy to follow the same path automatically?

[James Silverton]

Somehow, after being scared by human drivers in two dimensions in Paris,
London and New York, I would am completely unnerved by suggestions of
three-dimensional flying cars. Perhaps, totally reliable robocars might
solve the problems.

--
Jim Silverton (Potomac, MD)

Extraneous "not" in Reply To.

[Suzanne Blom]

Off-road vs. on road. Roads in most parts have all kinds of markers
that let one know what is going on. Further, when there's a hole in the
road ahead, for instance, stopping and waiting until the cops signal how
to go is usually perfectly acceptable. DARPA doubtless does not want
their things to stop and wait for further instructions.

[Bill Snyder]

And guys trying to flag them down might not be cops.

--
Bill Snyder [This space unintentionally left blank]

[Derek Lyons]

"Ken from Chicago" <kwicker1...@comcast.net> wrote:

>"Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote in message
>

>> That's a severely constrained test under very controlled conditions --
>
>It's still on a public street, which further along than I suspected.

It's on a public street - but it's not out in public. The streets
appear to have been closed and cleared in advance of the "test".

In other words, no different than those car commercials showing a car
flooring it through an urban landscape - "professional driver on a
closed course, do not try this at home". I.E. not representative of
real conditions, either of the road or of what the car would legally
be allowed to do. Complete fiction meant to impress the
impressionable.

>> and about a thousand things could go wrong with it even then. *I* wouldn't
>> have taken that risk, unless they had a police guard clearing the road
>> ahead.
>
>The end credit included local police.

I should think so. In most municipalities, they're the ones
responsible for closing and clearing the streets when someone applies
for a permit to do so. Equally, they're responsible when special
cargo, vehicles, or personages are transiting the local roadways.
Their presence in the credits is the same as their presence in movie
credits, it means they were there and provided cooperation and
operational support. It does not convey approval of or support for
robo cars.

D.
--
Touch-twice life. Eat. Drink. Laugh.

http://derekl1963.livejournal.com/

-Resolved: To be more temperate in my postings.
Oct 5th, 2004 JDL

[David Dyer-Bennet]

Sorry, you're quite right, and I shouldn't let myself get distracted
from the real issue that way!

[David Dyer-Bennet]

I wouldn't be shocked if the press releases emphasized the positive and
played down the negative. But I don't think there's any question that
they've been running their cars, with human co-pilots, on the street
quite a lot for quite a while. Just exactly how often the co-pilots
overrode is a matter that could be shaded to emphasize the positive,
certainly.

> If Google really has cars that can be let loose to drive
> themselves in real traffic, and don't need to be overridden by human
> beings, why is it that the Army is still having significant research
> being done just to get a damn convoy to follow the same path
> automatically?

Who knows? I'd *expect* Google researchers to do hugely better than the
Army on software issues, though.

[David Dyer-Bennet]

Google isn't claiming they're ready to productize and sell. It's just
that this theoretical thing, driving on a real street with real traffic,
that people say we won't know how to do for decades, they've already
done.

I can only report my memories of some articles I've read; you might as
well go read them first-hand, I'm sure they're easy to find, and get
something better than my second-hand retelling with errors. But I
believe there have been trips without override. Not *all* trips :-) .

> If they really are that good, why does the DARPA challenge
> exist at all? It would seem that Google should have long since
> finished that sucker off.

No idea; they probably don't fully meet the requirements of the
challenge yet .

[David Dyer-Bennet]

fair...@gmail.com (Derek Lyons) writes:

> "Ken from Chicago" <kwicker1...@comcast.net> wrote:
>
>>"Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote in message
>>
>>> That's a severely constrained test under very controlled conditions --
>>
>>It's still on a public street, which further along than I suspected.
>
> It's on a public street - but it's not out in public. The streets
> appear to have been closed and cleared in advance of the "test".

*This* test, perhaps; but they've been on public streets with traffic in
earlier work (with a co-pilot). This test was special in having no
licensed driver on board. Regardless of the state of the car, they
*HAD TO* close the streets and get police cooperation to be legal.

[Walter Bushell]

In article <jl2ps3$nuv$1...@dont-email.me>,

Considering the average driver anywhere, shouldn't be too long, if
it's not already here. A couple of months ago, I was almost run down
crossing the street with the light, apparently the driver hadn't
noticed the light had changed. I was about half way across the street
and I had speeded up to cross when the light changed and I was a few
feet from the curb.

--
This space unintentionally left blank.

[Ken from Chicago]

"Joseph Nebus" <nebusj-@-rpi-.edu> wrote in message
news:jl4en2$2dk$1...@reader1.panix.com...

<snip>

> Look, all this evidence of cars driving themselves is no
> evidence that cars will ever be able to drive themselves, and saying
> that it's evidence that cars may be able to drive themselves is just
> distracting the public from the need to sponsor the manned space
> programs and the helium-3 technology development essential to making
> cars that can drive themselves.
>
> --
> http://nebusresearch.wordpress.com/ Joseph
> Nebus
> Current Entry: How To Forget The Area Of A Trapezoid
> http://wp.me/p1RYhY-9d
> ------------------------------------------------------------------------------

This project doesn't prevent space exploration projects. There are several
countries working on that, not to mention several companies, such as Virgin
Galactic and SpaceX (which is prepping a launch to the ISS for the end of
April).

-- Ken from Chicago

[Ken from Chicago]

"Derek Lyons" <fair...@gmail.com> wrote in message
news:4f78e67d....@news.supernews.com...

Reportedly a police office monitored the drive but did not stop traffic or
interfere:

http://www.pcmag.com/article2/0,2817,2402380,00.asp

And there were triple-redundancies, a Google rep who's a licensed driver in
the front passenger with a passenger-side brake pedal, an emergency brake
button in the center column that anyone could press and the rep had a laptop
that could stop the car--altho there was were no remote control.

And Google's robocar(s) had driven 200,000 miles.

I agree, it's far from being sold to the public, but it's much further along
than I would have ever guessed.

-- Ken from Chicago

[Sea Wasp (Ryk E. Spoor)]

Driving as in "following along the street" and obeying reasonable local
laws is a fine achievement, but it's not tackling the hard problems.

If they've driven for, say, hundreds of hours in real city traffic with
no intervention? Okay, I'm impressed then. But from what I've seen, my
impression is that what they have is a "smart cruise control", so to
speak -- it'll drive itself, mostly, based on reasonable assumptions,
and the human is there to perform the harder tasks and intervene in
situations the normal cruise doesn't address.

Wonder if they've published on it... don't immediately see any journal
articles, but Google is a pain in the ass word to search on for obvious
reasons.

I did see a more casual IEEE article, with several interesting points:

"Urmson, who is the tech lead for the project, said that the "heart of
our system" is a laser range finder mounted on the roof of the car. The
device, a Velodyne 64-beam laser, generates a detailed 3D map of the
environment. "

Whoa. Yeah, that's gonna help. And HURT. Velodyne's system costs about
$75k. We use laser scanners in our unmanned work, and they have some...
very interesting limitations.They are very good at getting rangefinding
scans of the environment around them -- under good conditions. Also,
mechanically there's a whole bunch of questions in terms of long term
operation, but that's basically engineering (though not easy engineering).


But there are some conditions that kill them pretty dead -- in terms of
scanning, that is, regardless of how well it works; I'd want them backed
up with other sensors, heavily.

"The second thing is that, before sending the self-driving car on a road
test, Google engineers drive along the route one or more times to gather
data about the environment."

Yes, having a human scout go ahead and survey your territory, sometimes
multiple times, prior to sending the robot down, cuts way down on the
challenge.


They don't give a clear idea of how much human intervention is
required, unfortunately, which makes it impossible to judge directly.

>
>> If Google really has cars that can be let loose to drive
>> themselves in real traffic, and don't need to be overridden by human
>> beings, why is it that the Army is still having significant research
>> being done just to get a damn convoy to follow the same path
>> automatically?
>
> Who knows? I'd *expect* Google researchers to do hugely better than the
> Army on software issues, though.

Not on targeted software for specific solutions. The Army doesn't do
this stuff by itself; it hires people. Say, like Google... to do the
work for them. They'll PAY for the research, and it's done under their
banner, but the work's generally done by civilian contractors. Like the
one I work for. ;)

[Leif Roar Moldskred]

In rec.arts.sf.written "Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote:
>
> Driving as in "following along the street" and obeying reasonable local
> laws is a fine achievement, but it's not tackling the hard problems.

I'd expect that some of the hard problems with AI control of machinery
is to ensure they behave well when things go pear-shaped, right?
Driving along with the normal flow of traffic is one thing, but how
does it react when the car in front blows a tire or the engine
suddenly stalls? Will it recognize grazing deer by the wayside as a
reason to slow down? (To be fair, it only have to be substantially
better than human drivers at this, which makes it easier.) Then there
is the issue of _weather_.

--
Leif Roar Moldskred

[Sea Wasp (Ryk E. Spoor)]

On 3/30/12 4:18 PM, Leif Roar Moldskred wrote:
> In rec.arts.sf.written "Sea Wasp (Ryk E. Spoor)"<sea...@sgeinc.invalid.com> wrote:
>>
>> Driving as in "following along the street" and obeying reasonable local
>> laws is a fine achievement, but it's not tackling the hard problems.
>
> I'd expect that some of the hard problems with AI control of machinery
> is to ensure they behave well when things go pear-shaped, right?

Yes. Following a known route and following rules of known conduct are
not trivial, but are well within the areas that computers excel.

Decisionmaking in unbounded conditions they are *terrible* at. Some
humans are very good at it, most of us are okay, a few are terrible.

Unfortunately, those are exactly the conditions your robocar needs to
perform BEST in. As in, they'd better be AT LEAST as good as the best
human driver, and probably better, or they will be blamed for any
accident they get involved in, which will set back the deployment of
such vehicles significantly.

Weather is a different challenge in many ways. You could drastically
improve human performance in bad weather with the addition of better
sensors and displays. The computer has a major advantage over human
there in that a computer doesn't care about the particular perception
mode used; on the other hand, the REASON for that is one of the
computer's primary weaknesses; it has no understanding of what it sees.
That is, it may have the data that shows the existence of a curb, and of
cars ahead of it, and have rules on what to do around those things, but
it actually doesn't know WHAT those things are, which means it cannot
make judgments based on what the things are.

[Your Name]

In article <jl3fbk$lsb$1...@dont-email.me>, "Sea Wasp (Ryk E. Spoor)"
<sea...@sgeinc.invalid.com> wrote:
>
> ... unless they had a police guard clearing the road ahead.

Sounds like a "good" Government job creation scheme to reduce unemployment
- every auto-drive car comes with a "Road Assistance Manager" who walks 10
feet in front removing obstacles, carrying a warning flag / light, etc.
;-)

[Your Name]

In article <jl3hji$u6u$1...@dont-email.me>, "Ken from Chicago"

They don't have to be working together on the same car / system. Ford
could use Microsoft and Honda could use Google.

> > One of the biggest problems for auto-drive cars the useless GPS systems
> > that send you the wrong way up one way streets, through fields or lakes,
> > off cliffs, on huge detours (including going through different countries
> > just to get to the other side of the same city you're already in).
>
> What percentage of trips do they do that? Also, aren't those being updated?

And they'll still be constantly out-of-date, as well as having competitors
with different versions, etc.

Auto-drive cars in any real sense is just as stupid as opening up the
airways to flying cars for the average public driver - there's enough
mess using a bascially 2D roading system, let alone giving the morons
access to a 3D environment.

Auto-drive cars MIGHT work in a limited scope of a inner-city-based taxi
service, in a similar sense to the long-dreamt personal Pod systems, but
even then you'd have to keep them separated, which is why many previous
ideas for such systems have basically been monorails.

[Bill Snyder]

"EmergencyBrakingAlgorithm.exe has encountered a problem and needs
to close."

[Gutless Umbrella Carrying Sissy]

nebusj-@-rpi-.edu (Joseph Nebus) wrote in
news:jl33fm$f57$1...@reader1.panix.com:

> In <XnsA025AC15865...@69.16.186.7> Gutless Umbrella

> Carrying Sissy <taus...@gmail.com> writes:
>
>>"Ken from Chicago" <kwicker1...@comcast.net> wrote in
>>news:jl2lvp$4mk$1...@dont-email.me:
>

>>> (Here's a shorter link:)
>>>
>>> http://www.pcmag.com/article2/0,2817,2402340,00.asp
>>>
>>Unless, of course, something happens that the computer doesn't
>>expect, and can't cope with. What are the odds of that, on a
>>random public street?
>
>>Plus, they're still in the alpha stage anyway.
>

> Surely, the existence of self-driving cars in an
> acceptable
> alpha stage demonstrates that they will never be able to
> function as a generally distributed technology.
>
If you say so. I was commenting on Ken's comment that things were
happening a lot faster than he expected. My comment is that we can't
really tell yet, because they're not really at the protoype stage
yet.

--
Terry Austin

"Terry Austin: like the polio vaccine, only with more asshole."
-- David Bilek

Jesus forgives sinners, not criminals.

[Gutless Umbrella Carrying Sissy]

"Ken from Chicago" <kwicker1...@comcast.net> wrote in

news:jl3hji$u6u$1...@dont-email.me:

> "Your Name" <Your...@YourISP.com> wrote in message
> news:YourName-300...@203-118-187-140.dsl.dyn.ihug.co.nz
> ...

>> In article <XnsA025AC15865...@69.16.186.7>, Gutless

>> Umbrella Carrying Sissy <taus...@gmail.com> wrote:
>>> "Ken from Chicago" <kwicker1...@comcast.net> wrote in
>>> news:jl2lvp$4mk$1...@dont-email.me:
>>> >

>>> > I thought robocars wouldn't been on the streets until 2025.
>>> >
>>> > https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIP
>>> > P0/i juvQ9oJpuUJ
>>> >
>>> > Maybe some prototypes by 2020.
>>> >
>>> > Maybe even some on test tracks as early as 2015.
>>> >
>>> > I was wrong:
>>> >
>>> > http://news.cnet.com/8301-11386_3-57406639-76/google-self-dri
>>> > ving -car-chauffeurs-legally-blind-man/
>>> >

>>> > (Here's a shorter link:)
>>> >
>>> > http://www.pcmag.com/article2/0,2817,2402340,00.asp
>>>
>>> Unless, of course, something happens that the computer doesn't
>>> expect, and can't cope with. What are the odds of that, on a
>>> random public street?
>>>
>>> Plus, they're still in the alpha stage anyway.
>>

>> If Microsoft have ANY involvement, the car simply won't start
>> somedays, won't stop others, freeze in the middle of a journey,
>> the tyres will auto-deflate every time you put refuel /
>> recharge it, ...
>
> Google and Microsoft aren't exactly the best of friends.
> Microsoft lawsuit against Android phone makers doesn't help
> matters.
>

>> One of the biggest problems for auto-drive cars the useless GPS
>> systems that send you the wrong way up one way streets, through
>> fields or lakes, off cliffs, on huge detours (including going
>> through different countries just to get to the other side of
>> the same city you're already in).
>
> What percentage of trips do they do that? Also, aren't those
> being updated?
>

What percentage is acceptable when a failure means a very serious
risk of death?

[Gutless Umbrella Carrying Sissy]

"Ken from Chicago" <kwicker1...@comcast.net> wrote in

news:jl3hd2$tg5$1...@dont-email.me:

> "Gutless Umbrella Carrying Sissy" <taus...@gmail.com> wrote in
> message news:XnsA025AC15865...@69.16.186.7...

>> "Ken from Chicago" <kwicker1...@comcast.net> wrote in
>> news:jl2lvp$4mk$1...@dont-email.me:
>>
>>> I thought robocars wouldn't been on the streets until 2025.
>>>
>>> https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0
>>> /i juvQ9oJpuUJ
>>>
>>> Maybe some prototypes by 2020.
>>>
>>> Maybe even some on test tracks as early as 2015.
>>>
>>> I was wrong:
>>>
>>> http://news.cnet.com/8301-11386_3-57406639-76/google-self-drivi
>>> ng -car-chauffeurs-legally-blind-man/
>>>
>>> (Here's a shorter link:)
>>>
>>> http://www.pcmag.com/article2/0,2817,2402340,00.asp
>>>
>> Unless, of course, something happens that the computer doesn't
>> expect, and can't cope with. What are the odds of that, on a
>> random public street?
>

> So it speeds up, swerves, stops or reverses direction.

As it is programmed to do. Are you prepared to bet your life -
literally - that it will _always_ have the proper response?

>
>> Plus, they're still in the alpha stage anyway.
>

> Yes, but I didn't expect it to be on public streets this soon.
>
I'm a little surprised. It's taken law changes to make even this
level of testing acceptable. Until there's a fatal accident that is
traced back to the computer doing something that a live driver
probably wouldn't have. Then, we see the real obstacle to self-
driving cars:

Who do you sue? The driver? There isn't one. The passenger? He's
not in control, so he can't be (legally) resonsible. No, the ones
to sue will be the manufacturer of the car (and the driving system,
if that's a seperate company), because car makers have deep, deep
pockets. And they know it.

We may see some very clever prototypes within the next few years,
but it will take a long, long time before companies with multi-
billions in revenues open that revenue up to a lawsuit over _ever_
accident that happens.

[Your Name]

In article <ns7cn7lghh2o9sshu...@4ax.com>, Bill Snyder

Yep, bring a whole new meaning to the term "Blue screen of death". :-(

To be fair, although Microsoft is atroociously bad at programming, almost
half of Windows' issues are thanks to equally abysmal programming in
things like Adobe's Flash ... of course auto-drive cars will probably use
that to control the indicators. ;-)

[Bill Snyder]

On Fri, 30 Mar 2012 15:11:32 -0700, Gutless Umbrella Carrying

Your death, or my death?

[tphile2]

So what was the first robot car or super car in fiction? The
Batmobile? Something Tom Swift invented?

[Juho Julkunen]

In article <jl4g2q$d0n$2...@dont-email.me>, sea...@sgeinc.invalid.com
says...

> If they really are that good, why does the DARPA challenge exist at
> all? It would seem that Google should have long since finished that
> sucker off.

Is there still DARPA cahllenge? I haven't heard anything since 2007
Urban challenge (which six teams finished successfully) and assumed
they figured they were ready to start actually developing the suckers.

--
Juho Julkunen

[Robert Bannister]

On 30/03/12 2:13 PM, Your Name wrote:
> In article<XnsA025AC15865...@69.16.186.7>, Gutless Umbrella
> Carrying Sissy<taus...@gmail.com> wrote:
>> "Ken from Chicago"<kwicker1...@comcast.net> wrote in
>> news:jl2lvp$4mk$1...@dont-email.me:
>>>
>>> I thought robocars wouldn't been on the streets until 2025.
>>>
>>> https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/i
>>> juvQ9oJpuUJ
>>>
>>> Maybe some prototypes by 2020.
>>>
>>> Maybe even some on test tracks as early as 2015.
>>>
>>> I was wrong:
>>>
>>> http://news.cnet.com/8301-11386_3-57406639-76/google-self-driving
>>> -car-chauffeurs-legally-blind-man/
>>>
>>> (Here's a shorter link:)
>>>
>>> http://www.pcmag.com/article2/0,2817,2402340,00.asp
>>
>> Unless, of course, something happens that the computer doesn't
>> expect, and can't cope with. What are the odds of that, on a random
>> public street?
>>
>> Plus, they're still in the alpha stage anyway.
>
> If Microsoft have ANY involvement, the car simply won't start somedays,
> won't stop others, freeze in the middle of a journey, the tyres will
> auto-deflate every time you put refuel / recharge it, ...
>
>

> One of the biggest problems for auto-drive cars the useless GPS systems
> that send you the wrong way up one way streets, through fields or lakes,
> off cliffs, on huge detours (including going through different countries
> just to get to the other side of the same city you're already in).

GPS is a real problem in my town with taxis - you phone for one, but it
doesn't come, not because the driver couldn't be bothered, but because
his GPS told him to turn one street earlier or later than he should have
and so he doesn't find you.

--
Robert Bannister

[Robert Bannister]

Mind you, would they be worse than the many human drivers who obviously
need a reboot?

--
Robert Bannister

[Sea Wasp (Ryk E. Spoor)]

Hmm. I don't see a new one at the moment. Perhaps not.

[David DeLaney]

Ken from Chicago <kwicker1...@comcast.net> wrote:
>"Joseph Nebus" <nebusj-@-rpi-.edu> wrote in message

>> Look, all this evidence of cars driving themselves is no
>> evidence that cars will ever be able to drive themselves, and saying
>> that it's evidence that cars may be able to drive themselves is just
>> distracting the public from the need to sponsor the manned space
>> programs and the helium-3 technology development essential to making
>> cars that can drive themselves.
>

>This project doesn't prevent space exploration projects. There are several
>countries working on that, not to mention several companies, such as Virgin
>Galactic and SpaceX (which is prepping a launch to the ISS for the end of
>April).

I am glad to hear that we have multiple countries working on preventing
space exploration projects! This will lower the probability of one set
of Grim Meathook Futures where endless wealth actually rains from the sky
and space stations can be set up to shuffle dissidents and terrorists off
to so that Earth can settle into its middle-age comfortably!

Dave, now we just have to make sure the FTL projects don't come up with
anything useful either
--
\/David DeLaney posting from d...@vic.com "It's not the pot that grows the flower
It's not the clock that slows the hour The definition's plain for anyone to see
Love is all it takes to make a family" - R&P. VISUALIZE HAPPYNET VRbeable<BLINK>
http://www.vic.com/~dbd/ - net.legends FAQ & Magic / I WUV you in all CAPS! --K.

[Derek Lyons]

For values of "easier" that actually equate to "incredibly hard".
Despite all the anecdotal evidence tossed around, humans are on
average actually pretty dang good at this driving thing.

[Derek Lyons]

David Dyer-Bennet <dd...@dd-b.net> wrote:

>fair...@gmail.com (Derek Lyons) writes:
>
>> "Ken from Chicago" <kwicker1...@comcast.net> wrote:
>>
>>>"Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote in message
>>>
>>>> That's a severely constrained test under very controlled conditions --
>>>
>>>It's still on a public street, which further along than I suspected.
>>
>> It's on a public street - but it's not out in public. The streets
>> appear to have been closed and cleared in advance of the "test".
>
>*This* test, perhaps; but they've been on public streets with traffic in
>earlier work (with a co-pilot). This test was special in having no
>licensed driver on board. Regardless of the state of the car, they
>*HAD TO* close the streets and get police cooperation to be legal.

Right - they *had to*. There was JUST NO WAY a four passenger car
could have a co-pilot onboard. (tl;dr version: Bullshit.)

[Derek Lyons]

>Reportedly a police office monitored the drive but did not stop traffic or
>interfere:
>
>http://www.pcmag.com/article2/0,2817,2402380,00.asp

Then that's some of the emptiest streets I've ever seen... I.E.
something mose of us don't.

>And there were triple-redundancies, a Google rep who's a licensed driver in
>the front passenger with a passenger-side brake pedal, an emergency brake
>button in the center column that anyone could press and the rep had a laptop
>that could stop the car--altho there was were no remote control.

So now, the story changes - earlier it was "the police had to clear
the streets as there was no licensed drivers onboard". My suspicion
level increases.

[Leif Roar Moldskred]

In rec.arts.sf.written Derek Lyons <fair...@gmail.com> wrote:
> Leif Roar Moldskred <le...@dimnakorr.com> wrote:
>>
>>I'd expect that some of the hard problems with AI control of machinery
>>is to ensure they behave well when things go pear-shaped, right?
>>Driving along with the normal flow of traffic is one thing, but how
>>does it react when the car in front blows a tire or the engine
>>suddenly stalls? Will it recognize grazing deer by the wayside as a
>>reason to slow down? (To be fair, it only have to be substantially
>>better than human drivers at this, which makes it easier.)
>
> For values of "easier" that actually equate to "incredibly hard".

Oh, absolutely.

> Despite all the anecdotal evidence tossed around, humans are on
> average actually pretty dang good at this driving thing.

Yes, but frightfully inconsistent. One day they do emergency braking
on a split-second below-conscious decision without killing _anyone_,
and the next they back over their own picket fence. An AI system that
could drive _consistently_ at the average skill of even a poor driver
would be quite an accomplishment, and very useful.

--
Leif Roar Moldskred

[Your Name]

In article <MPG.29e07e9a5...@news.kolumbus.fi>, Juho Julkunen

You haven't heard anything because the cars:
- ran out of fuel driving around in circles,
- hit brick walls,
- ran over all the reporters,
- drove into the middle of a lake,
- ran out of fuel in the middle of Russia after following a GPS directions
- ...

None of them have actually made it to the finish line yet. ;-)

[Dimensional Traveler]

On 3/29/2012 10:47 PM, Ken from Chicago wrote:
> "Gutless Umbrella Carrying Sissy" <taus...@gmail.com> wrote in message
> news:XnsA025AC15865...@69.16.186.7...

>> "Ken from Chicago" <kwicker1...@comcast.net> wrote in
>> news:jl2lvp$4mk$1...@dont-email.me:
>>
>>> I thought robocars wouldn't been on the streets until 2025.
>>>
>>> https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/i
>>> juvQ9oJpuUJ
>>>
>>> Maybe some prototypes by 2020.
>>>
>>> Maybe even some on test tracks as early as 2015.
>>>
>>> I was wrong:
>>>
>>> http://news.cnet.com/8301-11386_3-57406639-76/google-self-driving
>>> -car-chauffeurs-legally-blind-man/
>>>
>>> (Here's a shorter link:)
>>>
>>> http://www.pcmag.com/article2/0,2817,2402340,00.asp
>>>
>> Unless, of course, something happens that the computer doesn't
>> expect, and can't cope with. What are the odds of that, on a random
>> public street?
>

> So it speeds up, swerves, stops or reverses direction.
>

At least three out of four of those will be the wrong response. What's
worse is that the one right answer won't be the same all the time.

[Dimensional Traveler]

On 3/29/2012 10:50 PM, Ken from Chicago wrote:
> "Your Name" <Your...@YourISP.com> wrote in message

> news:YourName-300...@203-118-187-140.dsl.dyn.ihug.co.nz...
>> In article <XnsA025AC15865...@69.16.186.7>, Gutless Umbrella

>> Carrying Sissy <taus...@gmail.com> wrote:
>>> "Ken from Chicago" <kwicker1...@comcast.net> wrote in
>>> news:jl2lvp$4mk$1...@dont-email.me:
>>> >
>>> > I thought robocars wouldn't been on the streets until 2025.
>>> >
>>> > https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/i
>>> > juvQ9oJpuUJ
>>> >
>>> > Maybe some prototypes by 2020.
>>> >
>>> > Maybe even some on test tracks as early as 2015.
>>> >
>>> > I was wrong:
>>> >
>>> > http://news.cnet.com/8301-11386_3-57406639-76/google-self-driving
>>> > -car-chauffeurs-legally-blind-man/
>>> >
>>> > (Here's a shorter link:)
>>> >
>>> > http://www.pcmag.com/article2/0,2817,2402340,00.asp
>>>
>>> Unless, of course, something happens that the computer doesn't
>>> expect, and can't cope with. What are the odds of that, on a random
>>> public street?
>>>

>>> Plus, they're still in the alpha stage anyway.
>>
>> If Microsoft have ANY involvement, the car simply won't start somedays,
>> won't stop others, freeze in the middle of a journey, the tyres will
>> auto-deflate every time you put refuel / recharge it, ...
>
> Google and Microsoft aren't exactly the best of friends. Microsoft
> lawsuit against Android phone makers doesn't help matters.
>
>> One of the biggest problems for auto-drive cars the useless GPS systems
>> that send you the wrong way up one way streets, through fields or lakes,
>> off cliffs, on huge detours (including going through different countries
>> just to get to the other side of the same city you're already in).
>
> What percentage of trips do they do that? Also, aren't those being updated?
>

I surprisingly high percentage. I once had a map/route calculation
software send me on a 100 mile roundabout because the map file had a
five foot gap in a road where it crossed a railroad track. I spent an
hour trying to figure out what the bleep the problem was and didn't
until I zoomed down so far that the map only showed a fifty foot section
of road.

[Dimensional Traveler]

It would give a new meaning to the phrase "your system has crashed". :/

[Ken from Chicago]

"Gutless Umbrella Carrying Sissy" <taus...@gmail.com> wrote in message

news:XnsA0269B2A840...@69.16.186.7...

> "Ken from Chicago" <kwicker1...@comcast.net> wrote in
> news:jl3hd2$tg5$1...@dont-email.me:
>
>> "Gutless Umbrella Carrying Sissy" <taus...@gmail.com> wrote in
>> message news:XnsA025AC15865...@69.16.186.7...

<snip>

>>> Unless, of course, something happens that the computer doesn't
>>> expect, and can't cope with. What are the odds of that, on a
>>> random public street?
>>
>> So it speeds up, swerves, stops or reverses direction.
>
> As it is programmed to do. Are you prepared to bet your life -
> literally - that it will _always_ have the proper response?

How does that differ from now trusting that brakes, steering columns,
accelerators will always have the proper response? Or that of EVERY other
vehicle around us? Trusting the machines genie is already out of the bottle.

>>> Plus, they're still in the alpha stage anyway.
>>
>> Yes, but I didn't expect it to be on public streets this soon.
>>
> I'm a little surprised. It's taken law changes to make even this
> level of testing acceptable. Until there's a fatal accident that is
> traced back to the computer doing something that a live driver
> probably wouldn't have. Then, we see the real obstacle to self-
> driving cars:
>
> Who do you sue? The driver? There isn't one. The passenger? He's
> not in control, so he can't be (legally) resonsible. No, the ones
> to sue will be the manufacturer of the car (and the driving system,
> if that's a seperate company), because car makers have deep, deep
> pockets. And they know it.

It's the same thing that happens already. If the owners didn't maintain the
car well then they are at fault. If the mechanics didn't service the car
well then they are at fault. If it's something common to a robocars in
general then the manufacturer's at fault.

This isn't the first time a lawsuit has arising over mechanical objects
causing loss of life or harm to someone. This stuff already happens to the
car industry.

> We may see some very clever prototypes within the next few years,
> but it will take a long, long time before companies with multi-
> billions in revenues open that revenue up to a lawsuit over _ever_
> accident that happens.
>
> --
> Terry Austin

Every year even regular cars are increasingly automated. Sure, some are
retro where the driver has more direct control and responsibility, but the
vast majority of cars have more automated systems to assist or even take
over aspects of driving.

I've long said we could have robocars on the street--if we were willing to
accept a top speed of 10-20 mph. That's plenty of time for then-current
computers to react--plus even if they failed, a crash at that speed is still
easily survivable by both passenger and pedestrian.

-- Ken from Chicago (who wouldn't mind a 10 mph robocar for local travel and
switches to manual mode for longer distances)

[tphile2]

On Mar 31, 1:28 am, "Ken from Chicago" <kwicker1b_nos...@comcast.net>
wrote:
> "Gutless Umbrella Carrying Sissy" <tausti...@gmail.com> wrote in messagenews:XnsA0269B2A840...@69.16.186.7...
>
> > "Ken from Chicago" <kwicker1b_nos...@comcast.net> wrote in
> >news:jl3hd2$tg5$1...@dont-email.me:
>
> >> "Gutless Umbrella Carrying Sissy" <tausti...@gmail.com> wrote in
> >> messagenews:XnsA025AC15865...@69.16.186.7...

I wouldn't care for a robocar even if it could transform into a giant
robot.
I want the robot driver. Now Robby The Robot was cool but I want
mine to look like Kato(hornet not panther version). or best of all
The Angelina Jolie wearing a Kato uniform model. the leg showing
skirt version of course.

[Kip Williams]

Turns out most Twits are actually shot by other Twits before they reach
the last stage.


Kip W
rasfw

[Steven L.]

"Ken from Chicago" <ken...@ymail.com> wrote in message
news:jl3hji$u6u$1...@dont-email.me:

> "Your Name" <Your...@YourISP.com> wrote in message
> news:YourName-300...@203-118-187-140.dsl.dyn.ihug.co.nz...

> > In article <XnsA025AC15865...@69.16.186.7>, Gutless Umbrella

> > Carrying Sissy <taus...@gmail.com> wrote:
> >> "Ken from Chicago" <kwicker1...@comcast.net> wrote in

> >> news:jl2lvp$4mk$1...@dont-email.me:
> >> >
> >> > I thought robocars wouldn't been on the streets until 2025.
> >> >
> >> > https://groups.google.com/d/msg/rec.arts.sf.written/gMOHEIaIPP0/i
> >> > juvQ9oJpuUJ
> >> >
> >> > Maybe some prototypes by 2020.
> >> >
> >> > Maybe even some on test tracks as early as 2015.
> >> >
> >> > I was wrong:
> >> >
> >> > http://news.cnet.com/8301-11386_3-57406639-76/google-self-driving
> >> > -car-chauffeurs-legally-blind-man/
> >> >
> >> > (Here's a shorter link:)
> >> >
> >> > http://www.pcmag.com/article2/0,2817,2402340,00.asp
> >>

> >> Unless, of course, something happens that the computer doesn't
> >> expect, and can't cope with. What are the odds of that, on a random
> >> public street?
> >>

> >> Plus, they're still in the alpha stage anyway.
> >

> > If Microsoft have ANY involvement, the car simply won't start somedays,
> > won't stop others, freeze in the middle of a journey, the tyres will
> > auto-deflate every time you put refuel / recharge it, ...
>
> Google and Microsoft aren't exactly the best of friends. Microsoft lawsuit
> against Android phone makers doesn't help matters.
>

> > One of the biggest problems for auto-drive cars the useless GPS systems
> > that send you the wrong way up one way streets, through fields or lakes,
> > off cliffs, on huge detours (including going through different countries
> > just to get to the other side of the same city you're already in).
>
> What percentage of trips do they do that? Also, aren't those being updated?

One major problem that GPS has is the inability to adjust to route
changes in real time.

I had an important appointment on a day when there was a snow emergency.
Different streets and roads were being opened and closed at different
times to allow snow plows to plow the snow. My GPS navigator didn't
know about any of that, of course, and kept directing me onto streets
that were temporarily closed to allow the snow plows to do their job.

In a more recent windstorm, a tree had fallen across the road that my
GPS navigator kept directing me onto.



-- Steven L.

[Steven L.]

"Gutless Umbrella Carrying Sissy" <taus...@gmail.com> wrote in message
news:XnsA0269A8BA54...@69.16.186.7:

What would a self-driving car do when a road is temporarily obstructed
due to construction work, and a police officer is detouring traffic
manually? Just plow right into him?




-- Steven L.

[Steven L.]

"Gutless Umbrella Carrying Sissy" <taus...@gmail.com> wrote in message

news:XnsA0269B2A840...@69.16.186.7:

The cars are equipped with manual override.

If the driver didn't override the computer, then he's responsible.

It's no different than cruise control now. If you get a speeding ticket
while you've set your cruise control to go 75 mph, then YOU are
responsible, not the manufacturer of the cruise control for "allowing"
speeding.

Note also that Category III autopilots can automatically land the
airliner at properly equipped airports. Most of us have already flown
on "self-landing" airliners.




-- Steven L.

[Leif Roar Moldskred]

In rec.arts.sf.written Steven L. <sdli...@earthlink.net> wrote:

> Note also that Category III autopilots can automatically land the
> airliner at properly equipped airports. Most of us have already flown
> on "self-landing" airliners.

All planes are self-landing. Once.

--
Leif Roar Moldskred

[Steven L.]

"Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote in message

news:jl4g6o$d0n$3...@dont-email.me:

> On 3/30/12 10:07 AM, Joseph Nebus wrote:
> > In<ylfk1uoa...@dd-b.net> David Dyer-Bennet<dd...@dd-b.net> writes:
> >
> >> "Sea Wasp (Ryk E. Spoor)"<sea...@sgeinc.invalid.com> writes:

> >
> >>> On 3/30/12 1:53 AM, Ken from Chicago wrote:
> >>>> It's still on a public street, which further along than I suspected.
> >>>>

> >>>>> and about a thousand things could go wrong with it even then. *I*
> >>>>> wouldn't have taken that risk, unless they had a police guard clearing
> >>>>> the road ahead.
> >>>>
> >>>> The end credit included local police.
> >>>

> >>> Which would indicate it's not actually close to being a real robocar.
> >>>
> >>> I will be impressed when your robocar can deal with situations
> >>> *I* have dealt with in traffic. Then I *might* consider it getting
> >>> close to deployable (might, because that would indicate that it was
> >>> about as good as me, but I'm not perfect, and if you want to replace
> >>> humans with robots, the robots have to be BETTER (because for damn
> >>> sure they'll be expensive).
> >
> >> Have you actually read the articles on this project? They've been
> >> driving around on public streets for years -- with human co-pilots.
> >
> >> The police were necessary this time, I believe, for *legal* reasons;
> >> there was no licensed driver in the car, and no legal provision for an
> >> automated car.
> >
> >> But it's been driving on real streets with other cars, without the human
> >> overriding, for quite a while now.

> >
> > Look, all this evidence of cars driving themselves is no
> > evidence that cars will ever be able to drive themselves,
>
>

> All these STATEMENTS that the cars have driven themselves without being
> overridden would appear to be at odds with the state of autonomous
> vehicle research as I have come to know it -- fairly well -- since
> having to do work IN that field.
>
> If Google really has cars that can be let loose to drive themselves in
> real traffic, and don't need to be overridden by human beings, why is it
> that the Army is still having significant research being done just to
> get a damn convoy to follow the same path automatically?

The Army's convoys probably need to go off-road (remember the Gulf War
battles in the desert and the battles in rough Afghanistan terrain?),
rather than stay on smooth roads with markers, road lines, and traffic
lights.

A self-driving all-terrain vehicle is harder than a self-driving
automobile for ordinary civilian use.

The thing I want to see the self-driving car do is choose a suitable
parking space, slow down, stop, and parallel park.




-- Steven L.

[Sea Wasp (Ryk E. Spoor)]

Actually, the described application involved roads.

> A self-driving all-terrain vehicle is harder than a self-driving
> automobile for ordinary civilian use.
>
> The thing I want to see the self-driving car do is choose a suitable
> parking space, slow down, stop, and parallel park.

That's not much of a challenge in terms of what's needed for an
automated car.

What I want to see one do is what I have done multiple times: note a
potential problem and make allowance for it.

For example:

I was once driving down Route 7 enroute to Schenectady. Ahead of me was
a truck for a gardening company. On the back of the pickup truck the
owner had customized a little shed, so his truck looked like a garden
shed -- good advertising.

I noticed that the shed seemed to be vibrating oddly, so I decided to
slow down, give him room.

A few seconds later, the entire shed LEAPED off the back of the truck
and smashed down -- exactly where I would have been had I been following
him at the normal distance.

I have been in similar circumstances, though not quite as amusing to
tell, several other times.

THAT is what a robocar needs to be able to do to convince me it's ready
for prime time.

[James Silverton]

"A landing you can walk away from is a *good landing*."

--
Jim Silverton (Potomac, MD)

Extraneous "not" in Reply To.

[Matthias Warkus]

Am 31.03.12 17:10, schrieb Steven L.:

> A self-driving all-terrain vehicle is harder than a self-driving
> automobile for ordinary civilian use.
>
> The thing I want to see the self-driving car do is choose a suitable
> parking space, slow down, stop, and parallel park.

The last part is already available in production cars, My parents
decided against the (expensive) autoparking option when buying a new VW
Golf last year. They test-drove it, though.

When driving past the parking space, the autoparking system gives you a
hint that it's big enough for your car. You then stop and reverse
slowly, then go forward a little again if necessary. All steering
throughout the process is automatic. The thing has been available since
at least 2008.

mawa
--
http://www.prellblog.de

[Greg Goss]

"Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote:
>On 3/31/12 11:10 AM, Steven L. wrote:

>> The thing I want to see the self-driving car do is choose a suitable
>> parking space, slow down, stop, and parallel park.
>
> That's not much of a challenge in terms of what's needed for an
>automated car.

The Ford Focus is not a high-luxury car.

"Parking space detected. Please release the wheel." (or something
similar.)

They demo-ed it on a recent instruction on "The Amazing Race".

"go to such and such location, and get your car to parallel park
itself."
"When we get home, I'm definitely getting a Ford."
--
I used to own a mind like a steel trap.
Perhaps if I'd specified a brass one, it
wouldn't have rusted like this.

[Sea Wasp (Ryk E. Spoor)]

On 3/31/12 12:17 PM, Greg Goss wrote:
> "Sea Wasp (Ryk E. Spoor)"<sea...@sgeinc.invalid.com> wrote:
>> On 3/31/12 11:10 AM, Steven L. wrote:
>
>>> The thing I want to see the self-driving car do is choose a suitable
>>> parking space, slow down, stop, and parallel park.
>>
>> That's not much of a challenge in terms of what's needed for an
>> automated car.
>
> The Ford Focus is not a high-luxury car.
>
> "Parking space detected. Please release the wheel." (or something
> similar.)

Although this doesn't quite fit the specifics, in which the car itself
makes all the decisions.

It is, however, a VERY nicely bounded problem which can be solved by
direct application of physics and calculations based on the known
dimensions of the car. That's one of the EASY ones for cars.

It's live interaction with other, unpredictable moving vehicles in a
non-controlled world that's the problem.

[Suzanne Blom]

I believe somebody, probably engineers who believe in their system, are
going to be paid a fair amount to stand in the road to make sure that
doesn't happen before the systems are certified.

[Your Name]

In article <jl7805$7bg$1...@dont-email.me>, not.jim....@verizon.net wrote:
> On 3/31/2012 11:09 AM, Leif Roar Moldskred wrote:
> > In rec.arts.sf.written Steven L.<sdli...@earthlink.net> wrote:
> >
> >> Note also that Category III autopilots can automatically land the
> >> airliner at properly equipped airports. Most of us have already flown
> >> on "self-landing" airliners.
> >
> > All planes are self-landing. Once.
>
> "A landing you can walk away from is a *good landing*."

A landing the plane can fly away from is a better one. ;-)

[Your Name]

In article <jl7ikh$a39$1...@dont-email.me>, "Sea Wasp (Ryk E. Spoor)"

<sea...@sgeinc.invalid.com> wrote:
> On 3/31/12 12:17 PM, Greg Goss wrote:
> > "Sea Wasp (Ryk E. Spoor)"<sea...@sgeinc.invalid.com> wrote:
> >> On 3/31/12 11:10 AM, Steven L. wrote:
> >
> >>> The thing I want to see the self-driving car do is choose a suitable
> >>> parking space, slow down, stop, and parallel park.
> >>
> >> That's not much of a challenge in terms of what's needed for an
> >> automated car.
> >
> > The Ford Focus is not a high-luxury car.
> >
> > "Parking space detected. Please release the wheel." (or something
> > similar.)
>
> Although this doesn't quite fit the specifics, in which the car
itself
> makes all the decisions.
>
> It is, however, a VERY nicely bounded problem which can be solved by
> direct application of physics and calculations based on the known
> dimensions of the car. That's one of the EASY ones for cars.

Even "simple" parking isn't that easy. For example, a small dog could run
out through the space being backed into - although a driver might not see
it either, they would hear the dog's owner shouting furiously for the dog
and / or the car to stop.

> It's live interaction with other, unpredictable moving vehicles in a
> non-controlled world that's the problem.

Unfortunately there are already auto-drive cars on the roads being used by
the general public. Just like the auto-park systems, there are some cars
that have auto-drive fitted which are meant to be used on highways /
motorways and slow moving traffic jams (it's anyone's guess when the
owners actually use them though!). :-(

[Duggy]

On Apr 1, 12:48 am, "Steven L." <sdlit...@earthlink.net> wrote:
> What would a self-driving car do when a road is temporarily obstructed
> due to construction work, and a police officer is detouring traffic
> manually?  Just plow right into him?

Why would self driving cars be different to human drivers?

===
= DUG.
===

[Duggy]

On Apr 1, 12:46 am, "Steven L." <sdlit...@earthlink.net> wrote:
> One major problem that GPS has is the inability to adjust to route
> changes in real time.

That's not a GPS problem, it's a reporting problem.

===
= DUG.
===

[Duggy]

On Apr 1, 12:56 am, "Steven L." <sdlit...@earthlink.net> wrote:
> > >> "Ken from Chicago" <kwicker1b_nos...@comcast.net> wrote:

> > >>>http://news.cnet.com/8301-11386_3-57406639-76/google-self-drivi
> > >>> ng -car-chauffeurs-legally-blind-man/

> > >>> (Here's a shorter link:)

> > >>>http://www.pcmag.com/article2/0,2817,2402340,00.asp

> The cars are equipped with manual override.
>
> If the driver didn't override the computer, then he's responsible.

Checking URL:

"/google-self-driving -car-chauffeurs-legally-blind-man/"

===
= DUG.
===

[Duggy]

On Apr 1, 1:10 am, "Steven L." <sdlit...@earthlink.net> wrote:
> The thing I want to see the self-driving car do is choose a suitable
> parking space,

Don't know.

> slow down, stop, and parallel park.

http://www.youtube.com/watch?v=dTa3d0zKqAY

===
= DUG.
===

[Duggy]

On Apr 1, 6:49 am, Suzanne Blom <bo...@sueblom.net> wrote:
> I believe somebody, probably engineers who believe in their system, are
> going to be paid a fair amount to stand in the road to make sure that
> doesn't happen before the systems are certified.

It would be an easy matter to add a wi-fi signal or such to hand signs
for roadworks and lollypop ladies, and something similar for police.

With very few cars available to roll-out should be simple.

===
= DUG.
===

[David Dyer-Bennet]

fair...@gmail.com (Derek Lyons) writes:

> David Dyer-Bennet <dd...@dd-b.net> wrote:
>
>>fair...@gmail.com (Derek Lyons) writes:
>>
>>> "Ken from Chicago" <kwicker1...@comcast.net> wrote:
>>>
>>>>"Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote in message
>>>>
>>>>> That's a severely constrained test under very controlled conditions --

>>>>
>>>>It's still on a public street, which further along than I suspected.
>>>

>>> It's on a public street - but it's not out in public. The streets
>>> appear to have been closed and cleared in advance of the "test".
>>
>>*This* test, perhaps; but they've been on public streets with traffic in
>>earlier work (with a co-pilot). This test was special in having no
>>licensed driver on board. Regardless of the state of the car, they
>>*HAD TO* close the streets and get police cooperation to be legal.
>
> Right - they *had to*. There was JUST NO WAY a four passenger car
> could have a co-pilot onboard. (tl;dr version: Bullshit.)

Think it through. Of course there *could have* been a co-polit on
board. But then it wouldn't have been newsworthy.

This was a publicity stunt -- a blind man in a self-driving car. If you
have a co-pilot there, it becomes dull -- just the same thing they've
been doing for a couple of years now.
--
David Dyer-Bennet, dd...@dd-b.net; http://dd-b.net/
Snapshots: http://dd-b.net/dd-b/SnapshotAlbum/data/
Photos: http://dd-b.net/photography/gallery/
Dragaera: http://dragaera.info

[David Dyer-Bennet]

James Silverton <jim.si...@verizon.net> writes:

> On 3/31/2012 11:09 AM, Leif Roar Moldskred wrote:
>> In rec.arts.sf.written Steven L.<sdli...@earthlink.net> wrote:
>>
>>> Note also that Category III autopilots can automatically land the
>>> airliner at properly equipped airports. Most of us have already flown
>>> on "self-landing" airliners.
>>
>> All planes are self-landing. Once.
>>
>
> "A landing you can walk away from is a *good landing*."

A landing where you can use the plane again is a *great* landing.

[Dimensional Traveler]

On 3/31/2012 11:32 AM, Sea Wasp (Ryk E. Spoor) wrote:
> On 3/31/12 12:17 PM, Greg Goss wrote:
>> "Sea Wasp (Ryk E. Spoor)"<sea...@sgeinc.invalid.com> wrote:
>>> On 3/31/12 11:10 AM, Steven L. wrote:
>>
>>>> The thing I want to see the self-driving car do is choose a suitable
>>>> parking space, slow down, stop, and parallel park.
>>>
>>> That's not much of a challenge in terms of what's needed for an
>>> automated car.
>>
>> The Ford Focus is not a high-luxury car.
>>
>> "Parking space detected. Please release the wheel." (or something
>> similar.)
>
> Although this doesn't quite fit the specifics, in which the car itself
> makes all the decisions.
>
> It is, however, a VERY nicely bounded problem which can be solved by
> direct application of physics and calculations based on the known
> dimensions of the car. That's one of the EASY ones for cars.
>
> It's live interaction with other, unpredictable moving vehicles in a
> non-controlled world that's the problem.
>

So you're saying it has to be ALL robocars or none at all. ;)

('Cause if they're all robocars then you can just have them talk to each
other via blutooth so they can form a consensus about how late to work
we all are going to be.)

[David DeLaney]

James Silverton <jim.si...@verizon.net> wrote:
>On 3/31/2012 11:09 AM, Leif Roar Moldskred wrote:
>> In rec.arts.sf.written Steven L.<sdli...@earthlink.net> wrote:
>>> Note also that Category III autopilots can automatically land the
>>> airliner at properly equipped airports. Most of us have already flown
>>> on "self-landing" airliners.
>>
>> All planes are self-landing. Once.
>
>"A landing you can walk away from is a *good landing*."

So this research would just extend this capability to "can automatically
land the airliner on the nearest interstate highway", right?

Dave, will drag hooks for food
--
\/David DeLaney posting from d...@vic.com "It's not the pot that grows the flower
It's not the clock that slows the hour The definition's plain for anyone to see
Love is all it takes to make a family" - R&P. VISUALIZE HAPPYNET VRbeable<BLINK>
http://www.vic.com/~dbd/ - net.legends FAQ & Magic / I WUV you in all CAPS! --K.

[Sea Wasp (Ryk E. Spoor)]

All robocars would make it easier, but even that wouldn't deal with the
unpredictable. How does a robocar know that some kid's about to run out
in front of it? A human being can SEE the movement and understand what
it means. How does a robocar look ahead and see a tree coming down, and
know it has to brake NOW because the other cars ahead of it CAN'T stop
in time?

That's actual *understanding* of a scene. None of the robocar research
is close to cracking THAT problem.

[Joseph Nebus]

And if there's anything that a half-century's experience with
computer programming proves, it's that computers can never be programmed
to make the kinds of decisions that trained humans can.

--
http://nebusresearch.wordpress.com/ Joseph Nebus
Current Entry: How To Forget The Area Of A Trapezoid http://wp.me/p1RYhY-9d
------------------------------------------------------------------------------

[Ken from Chicago]

"Steven L." <sdli...@earthlink.net> wrote in message
news:yt6dndZQWaz5hOrS...@earthlink.com...

<snip>

> What would a self-driving car do when a road is temporarily obstructed due
> to construction work, and a police officer is detouring traffic manually?
> Just plow right into him?
>
>
>
>
> -- Steven L.

The same thing they'd do if someone runs out into the street or if a car is
simply parked in the middle of the street--detour.

It's not like this is some insurmountable or even unimaginable problem.
Basic video cameras could detect the police officer, or construction worker,
or just a regular person in the middle of the street and scan their
gesturing in a certain direction and go in the direction, recalculating the
path back to its destination.

-- Ken from Chicago

[Gene Wirchenko]

On Sat, 31 Mar 2012 14:46:32 +0000, "Steven L."
<sdli...@earthlink.net> wrote:

[snip]

>One major problem that GPS has is the inability to adjust to route
>changes in real time.

TheGPS that I used in the trip I wrote up -- see below -- did
adjust. There was no way to say that a street was not open though.

>I had an important appointment on a day when there was a snow emergency.
> Different streets and roads were being opened and closed at different
>times to allow snow plows to plow the snow. My GPS navigator didn't
>know about any of that, of course, and kept directing me onto streets
>that were temporarily closed to allow the snow plows to do their job.
>
>In a more recent windstorm, a tree had fallen across the road that my
>GPS navigator kept directing me onto.

I had fun with a GPS unit on one trip. I wrote it up for RISKS:
http://catless.ncl.ac.uk/Risks/25.88.html#subj11.1

Sincerely,

Gene Wirchenko

[Ken from Chicago]

"Dimensional Traveler" <dtr...@sonic.net> wrote in message
news:4f7697ec$0$49348$742e...@news.sonic.net...
> On 3/30/2012 2:04 PM, Bill Snyder wrote:
>> On Sat, 31 Mar 2012 09:01:26 +1200, Your...@YourISP.com (Your
>> Name) wrote:
>>
>>> In article<jl3hji$u6u$1...@dont-email.me>, "Ken from Chicago"
>>> <ken...@ymail.com> wrote:

<snip>

>>>> Google and Microsoft aren't exactly the best of friends. Microsoft
>>>> lawsuit
>>>> against Android phone makers doesn't help matters.
>>>

>>> They don't have to be working together on the same car / system. Ford
>>> could use Microsoft and Honda could use Google.
>>
>> "EmergencyBrakingAlgorithm.exe has encountered a problem and needs
>> to close."
>>
> It would give a new meaning to the phrase "your system has crashed". :/

"Start me up."

http://www.youtube.com/watch?v=5VPFKnBYOSI

"Where do you want to go today?"

http://www.youtube.com/watch?v=qwXYJIZOyyk

-- Ken from Chicago

[Ken from Chicago]

"Steven L." <sdli...@earthlink.net> wrote in message

news:bP2dnXUp4LrKg-rS...@earthlink.com...
>

<snip>

> The Army's convoys probably need to go off-road (remember the Gulf War
> battles in the desert and the battles in rough Afghanistan terrain?),
> rather than stay on smooth roads with markers, road lines, and traffic
> lights.
>
> A self-driving all-terrain vehicle is harder than a self-driving
> automobile for ordinary civilian use.
>
> The thing I want to see the self-driving car do is choose a suitable
> parking space, slow down, stop, and parallel park.
>
>
>
>
> -- Steven L.

I wouldn't mind if it were limited to 5-10 mph, it could still drop me off
and pick me up at the bus stop for daily commutes and park itself at home or
if I drove somewhere and got out at the front door it could park itself at
and pick me up from the local parking lot or find a nearby parking space.
Natch it could use narrower parking spaces since I wouldn't need to get in
the car while it was parked.

-- Ken from Chicago

[David Harmon]

On Fri, 30 Mar 2012 22:26:42 -0700 in rec.arts.sf.written, Dimensional
Traveler <dtr...@sonic.net> wrote,
>I surprisingly high percentage. I once had a map/route calculation
>software send me on a 100 mile roundabout because the map file had a
>five foot gap in a road where it crossed a railroad track.

A couple of blocks from me, Microsoft Streets & Trips will routinely
route me across the railroad track where in reality there is a serious
gap in the road.

[Gene Wirchenko]

On Fri, 30 Mar 2012 23:22:35 -0400, d...@gatekeeper.vic.com (David
DeLaney) wrote:

>Ken from Chicago <kwicker1...@comcast.net> wrote:

[snip]

>>This project doesn't prevent space exploration projects. There are several
>>countries working on that, not to mention several companies, such as Virgin
>>Galactic and SpaceX (which is prepping a launch to the ISS for the end of
>>April).
>
>I am glad to hear that we have multiple countries working on preventing
>space exploration projects! This will lower the probability of one set
>of Grim Meathook Futures where endless wealth actually rains from the sky
>and space stations can be set up to shuffle dissidents and terrorists off
>to so that Earth can settle into its middle-age comfortably!
>
>Dave, now we just have to make sure the FTL projects don't come up with
> anything useful either

Why stop with preventing just two things? Why, the sky's the
limit!

Sincerely,

Gene "A ounce of prevention..." Wirchenko

[Sea Wasp (Ryk E. Spoor)]

On 4/1/12 11:39 AM, Joseph Nebus wrote:
> In<jl8f89$qgn$1...@dont-email.me> "Sea Wasp (Ryk E. Spoor)"<sea...@sgeinc.invalid.com> writes:
>
>> On 3/31/12 10:26 PM, Dimensional Traveler wrote:
>>> On 3/31/2012 11:32 AM, Sea Wasp (Ryk E. Spoor) wrote:
>>>> It's live interaction with other, unpredictable moving vehicles in a
>>>> non-controlled world that's the problem.
>>>>
>>> So you're saying it has to be ALL robocars or none at all. ;)
>>>
>>> ('Cause if they're all robocars then you can just have them talk to each
>>> other via blutooth so they can form a consensus about how late to work
>>> we all are going to be.)
>>>
>
>> All robocars would make it easier, but even that wouldn't deal with the
>> unpredictable. How does a robocar know that some kid's about to run out
>> in front of it? A human being can SEE the movement and understand what
>> it means. How does a robocar look ahead and see a tree coming down, and
>> know it has to brake NOW because the other cars ahead of it CAN'T stop
>> in time?
>
>> That's actual *understanding* of a scene. None of the robocar research
>> is close to cracking THAT problem.
>
> And if there's anything that a half-century's experience with
> computer programming proves, it's that computers can never be programmed
> to make the kinds of decisions that trained humans can.
>

No, it doesn't prove that. But it does prove that it's a very hard
field. And your sarcasm tends to prove that you do not understand the
problem, or are ignoring large parts of it.

It also depends on what you mean by "make decisions". Expert systems
can be made to do things like diagnosis quite well; where I work, we've
had considerable success in doing just that to allow a computer to
decide whether (for example) a wheel bearing is good or bad. But that's
not the same kind of problem as we are dealing with in perception and
understanding.

That is, if you had some way to input to the computer "There is a child
in X location, moving at Y speed in Z direction, and you are in a car in
A location, moving at B speed in C direction", you could -- relatively
easily -- instantiate rules whereby it could make decisions about
whether to brake, speed up, turn, etc., to avoid hitting the child.

The problem is that GETTING that information is exceedingly hard.
People have been working on the "smart video" side of things for many
years, and thus far it's basically demonstrated that we're very good at
seeing and understanding stuff, and computers aren't, because humans
have huge built-in processing for that purpose, and we actually
understand the world around us. We know what "child" means, and we know
why it is that we want to avoid hitting a child, and what priorities we
should assign to that, and we don't even THINK about it for the most part.

Where I work, we are currently working on an application for a robot to
operate something on a vehicle moving past the robot. This is an
operation we know human beings do very routinely, thousands of times per
day. We have plenty of data on how the human beings do it, we have a
quite constrained environment in which to try to do this same operation,
and it is STILL one hell of a challenge.

The challenge of smart-driving cars is several orders of magnitude greater.

[Sea Wasp (Ryk E. Spoor)]

Yes, and it sounds very easy when you describe it that way. It's
actually one of the most difficult problems in AI, to UNDERSTAND a scene
like that reliably. We have no problem at all seeing a figure in front
of us and recognizing that it is a policeman gesturing for us to stop or
turn. We know what a policeman is, we know what sort of things a
policeman will expect of us, and so on.

A computer doesn't know these things, and getting a computer to
reliably recognize even such apparently simple things as "that's a human
being, and that's a dog" is a major achievement -- and often it's
considered a success when such a system manages 95% reliability. Just
how reliable would you consider a person if they confused a human with a
dog 1 out of 20 times they looked? We're close to 100% accurate, even in
some pretty difficult circumstances, and if you're talking about
driving, you can't afford 1 out of 20 failures. Maybe 1 out of a million.

[Your Name]

In article <jl9ssd$t2t$1...@reader1.panix.com>, nebusj-@-rpi-.edu (Joseph

Nebus) wrote:
> In <jl8f89$qgn$1...@dont-email.me> "Sea Wasp (Ryk E. Spoor)"
> <sea...@sgeinc.invalid.com> writes:
> >
> >On 3/31/12 10:26 PM, Dimensional Traveler wrote:
> >> On 3/31/2012 11:32 AM, Sea Wasp (Ryk E. Spoor) wrote:
> >>> It's live interaction with other, unpredictable moving vehicles in a
> >>> non-controlled world that's the problem.
> >>>
> >> So you're saying it has to be ALL robocars or none at all. ;)
> >>
> >> ('Cause if they're all robocars then you can just have them talk to each
> >> other via blutooth so they can form a consensus about how late to work
> >> we all are going to be.)
>
> >All robocars would make it easier, but even that wouldn't deal with the
> >unpredictable. How does a robocar know that some kid's about to run out
> >in front of it? A human being can SEE the movement and understand what
> >it means. How does a robocar look ahead and see a tree coming down, and
> >know it has to brake NOW because the other cars ahead of it CAN'T stop
> >in time?
> >
> >That's actual *understanding* of a scene. None of the robocar research
> >is close to cracking THAT problem.
>
> And if there's anything that a half-century's experience with
> computer programming proves, it's that computers can never be programmed
> to make the kinds of decisions that trained humans can.

Half a century?? It only takes five minutes playing any game with
so-called "AI" characters to learn that fact. ;-)

[Your Name]

In article <jla4rf$666$1...@dont-email.me>, "Ken from Chicago"

<ken...@ymail.com> wrote:
> "Steven L." <sdli...@earthlink.net> wrote in message
> news:yt6dndZQWaz5hOrS...@earthlink.com...
>
> <snip>
>
> > What would a self-driving car do when a road is temporarily obstructed due
> > to construction work, and a police officer is detouring traffic manually?
> > Just plow right into him?
>

> The same thing they'd do if someone runs out into the street or if a car is
> simply parked in the middle of the street--detour.
>
> It's not like this is some insurmountable or even unimaginable problem.
> Basic video cameras could detect the police officer, or construction worker,
> or just a regular person in the middle of the street and scan their
> gesturing in a certain direction and go in the direction, recalculating the
> path back to its destination.

That should make life extremely easy for people like carjackers, armoured
car robbers, etc. ;-)

It's impossible to try to get a car to recognise and react sensibly to an
endless variety of situations, no two of which will ever be exactly the
same.

There's also the problem of more and more electronic garbage being added
to the car - both in terms of battery technology and all the added
gimmickry to get it to drive itself simply adds complexity with more
things to go wrong and more things blocking the ability to actually drive
the car (e.g. the computer won't let me change into second, but there's
nothing physically wrong with the working of the gears).

[Dimensional Traveler]

There is also the matter of computer don't "learn" the same way humans
do. I recall hearing about one test of, IIRC, an early neural net
learning computer. They showed it pictures of tanks in a field and of
the same field empty, trying to teach it to recognize tanks. What the
computer learned apparently was that tanks are a kind of mushroom since
it turned out that all the empty field pictures were taken during sunny
days and all the full field pictures were taken during twilight or
overcast days. :\

[Sea Wasp (Ryk E. Spoor)]

On 4/1/12 6:00 PM, Dimensional Traveler wrote:

> There is also the matter of computer don't "learn" the same way humans
> do. I recall hearing about one test of, IIRC, an early neural net
> learning computer. They showed it pictures of tanks in a field and of
> the same field empty, trying to teach it to recognize tanks. What the
> computer learned apparently was that tanks are a kind of mushroom since
> it turned out that all the empty field pictures were taken during sunny
> days and all the full field pictures were taken during twilight or
> overcast days. :\

Neural-net approaches are very powerful, but the problem you point out
is one of their major weaknesses, which is that they are, effectively, a
black box. Properly trained, they can work well, but you don't actually
KNOW what it is that they're recognizing, so you can't tell if they're
working well because they have actually constructed a mechanism to
recognize the targets in question, or if they've trained on some
coincidental pattern that you've put into the training data -- often, a
pattern that you as a human wouldn't see specifically because you
understand the image and know what's irrelevant.

[Derek Lyons]

David Dyer-Bennet <dd...@dd-b.net> wrote:

>fair...@gmail.com (Derek Lyons) writes:
>
>> David Dyer-Bennet <dd...@dd-b.net> wrote:
>>
>>>fair...@gmail.com (Derek Lyons) writes:
>>>
>>>> "Ken from Chicago" <kwicker1...@comcast.net> wrote:
>>>>
>>>>>"Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote in message
>>>>>
>>>>>> That's a severely constrained test under very controlled conditions --
>>>>>
>>>>>It's still on a public street, which further along than I suspected.
>>>>
>>>> It's on a public street - but it's not out in public. The streets
>>>> appear to have been closed and cleared in advance of the "test".
>>>
>>>*This* test, perhaps; but they've been on public streets with traffic in
>>>earlier work (with a co-pilot). This test was special in having no
>>>licensed driver on board. Regardless of the state of the car, they
>>>*HAD TO* close the streets and get police cooperation to be legal.
>>
>> Right - they *had to*. There was JUST NO WAY a four passenger car
>> could have a co-pilot onboard. (tl;dr version: Bullshit.)
>
>Think it through. Of course there *could have* been a co-polit on
>board. But then it wouldn't have been newsworthy.
>
>This was a publicity stunt -- a blind man in a self-driving car. If you
>have a co-pilot there, it becomes dull -- just the same thing they've
>been doing for a couple of years now.

Yeah, like a publicity stunt involves telling 100% of the truth. it
would *never* occur to anyone to put a co-pilot onboard and not tell
anyone.

Sheesh. Are you for real?

D.
--
Touch-twice life. Eat. Drink. Laugh.

http://derekl1963.livejournal.com/

-Resolved: To be more temperate in my postings.
Oct 5th, 2004 JDL

[Derek Lyons]

"Ken from Chicago" <kwicker1...@comcast.net> wrote:

>"Gutless Umbrella Carrying Sissy" <taus...@gmail.com> wrote in message
>news:XnsA0269B2A840...@69.16.186.7...
>> "Ken from Chicago" <kwicker1...@comcast.net> wrote in
>> news:jl3hd2$tg5$1...@dont-email.me:
>>
>>> "Gutless Umbrella Carrying Sissy" <taus...@gmail.com> wrote in
>>> message news:XnsA025AC15865...@69.16.186.7...
>
><snip>
>
>>>> Unless, of course, something happens that the computer doesn't
>>>> expect, and can't cope with. What are the odds of that, on a
>>>> random public street?
>>>
>>> So it speeds up, swerves, stops or reverses direction.
>>
>> As it is programmed to do. Are you prepared to bet your life -
>> literally - that it will _always_ have the proper response?
>
>How does that differ from now trusting that brakes, steering columns,
>accelerators will always have the proper response? Or that of EVERY other
>vehicle around us? Trusting the machines genie is already out of the bottle.

You honestly don't see the difference between the items you list and a
many orders of magnitude more complex bucket of software and
actuators?

Let's talk about some property in Florida. There's no difference
between dry land and swamp anyhow.

[Your Name]

In article <4f78d012$0$49350$742e...@news.sonic.net>, Dimensional

Traveler <dtr...@sonic.net> wrote:
>
> There is also the matter of computer don't "learn" the same way humans
> do. I recall hearing about one test of, IIRC, an early neural net
> learning computer. They showed it pictures of tanks in a field and of
> the same field empty, trying to teach it to recognize tanks. What the
> computer learned apparently was that tanks are a kind of mushroom since
> it turned out that all the empty field pictures were taken during sunny
> days and all the full field pictures were taken during twilight or
> overcast days. :\

There was a TV show I saw a few months ago with a robot that did learn and
could recognise similar objects by being shown them and told their names.
It did very well for a robot / computer, but of course did make mistakes
too, and that was in a controleed situation with specially selected
objects. Even at the rate technology is changing, it's still many many
decades before this is actually a remotely usable and useful feature.

[David DeLaney]

Sea Wasp (Ryk E. Spoor) <sea...@sgeinc.invalid.com> wrote:
>On 3/31/12 10:26 PM, Dimensional Traveler wrote:
>> So you're saying it has to be ALL robocars or none at all. ;)
>>
>> ('Cause if they're all robocars then you can just have them talk to each
>> other via blutooth so they can form a consensus about how late to work
>> we all are going to be.)
>
> All robocars would make it easier, but even that wouldn't deal with the
>unpredictable. How does a robocar know that some kid's about to run out
>in front of it? A human being can SEE the movement and understand what
>it means. How does a robocar look ahead and see a tree coming down, and
>know it has to brake NOW because the other cars ahead of it CAN'T stop
>in time?

Well, duh; you need robokids and cybertrees.

Dave "and nanocommunication between it all so that it's all busy editing
Wikipedia instead of taking over the world or anything" DeLaney

[Ken from Chicago]

"Sea Wasp (Ryk E. Spoor)" <sea...@sgeinc.invalid.com> wrote in message
news:jl55ig$ood$1...@dont-email.me...
> On 3/30/12 4:18 PM, Leif Roar Moldskred wrote:
>> In rec.arts.sf.written "Sea Wasp (Ryk E.
>> Spoor)"<sea...@sgeinc.invalid.com> wrote:
>>>
>>> Driving as in "following along the street" and obeying
>>> reasonable local
>>> laws is a fine achievement, but it's not tackling the hard problems.
>>
>> I'd expect that some of the hard problems with AI control of machinery
>> is to ensure they behave well when things go pear-shaped, right?
>
> Yes. Following a known route and following rules of known conduct are not
> trivial, but are well within the areas that computers excel.
>
> Decisionmaking in unbounded conditions they are *terrible* at. Some humans
> are very good at it, most of us are okay, a few are terrible.

Except driving isn't unbounded.

Your boundaries include at least a pair or more of moving perimeters
surrounding a vehicle. Anything that breaches that outer perimeter is
analyzed for direction, velocity, size and mass. Anything that breaches the
inner perimeter would be placed on high priority for avoidance, be it
speeding up, swerving, slowing down or stopping. Between the two perimeters
a robocar would have an idea of its immediate surroundings, situational
awareness.

The key is speed. How fast can it detect an obstacle, or obstacles, and how
fast can it react? If it is fast enough then the rest is gravy.

For example, with enough speed it can identify nearby road signs and
addresses to confirm its location in comparison with GPS tracking and where
there are differences, it can decide how to react (e.g., if a street sign
differs from the street it expects, is it an honorary street sign, very
common in Chicago city, or has the sign been altered by vandals where
previous and subsequent street signs match, or is the GPS route wrong and it
has to scan local addresses to get its bearings on its location, or even
pause to ask local pedestrians for directions, much as a human driver would
do), while recording the data on its own internal map for future trips. Also
with its owner's permission, it can opt to anonymously submit discrepancies
with reality to the Cloud so GPS tracking can be updated.

> Unfortunately, those are exactly the conditions your robocar needs to
> perform BEST in. As in, they'd better be AT LEAST as good as the best
> human driver, and probably better, or they will be blamed for any accident
> they get involved in, which will set back the deployment of such vehicles
> significantly.

Plus said deployment can be in stages, further boundaries, limiting use of a
robocar's auto-drive to highways where following a known route or the
highway is much easier, there's little to no cross traffic, and things are
more predictable. Another restriction might be as an auto-valet, restricting
the car to 5-10 mph to basically park itself after dropping you off at the
front door or coming to pick you up at the front door from nearby parking
spaces, thus minimizing the damage it could do even if it hit someone.

> Weather is a different challenge in many ways. You could drastically
> improve human performance in bad weather with the addition of better
> sensors and displays. The computer has a major advantage over human there
> in that a computer doesn't care about the particular perception mode used;
> on the other hand, the REASON for that is one of the computer's primary
> weaknesses; it has no understanding of what it sees. That is, it may have
> the data that shows the existence of a curb, and of cars ahead of it, and
> have rules on what to do around those things, but it actually doesn't know
> WHAT those things are, which means it cannot make judgments based on what
> the things are.

You're saying "That is, it may have the data that shows the existence of a
curb, and of cars ahead of it, and have rules on what to do around those
things" does NOT contradict "but it actually doesn't know WHAT those things
are, which means it cannot make judgments based on what the things are"?

Does it matter?

If it has rules on how to translate "Buenos dias, mi amigo" to "Good
morning, mine friend", does it matter if the computer doesn't understand
intrinsically *what* those words mean as long as it can translate those
words fluently--especially if it has a sufficiently detailed thesaurus to
paraphrase so that "De nada.", in response to "Gracias." isn't translated as
"It's nothing." but translated as "You're welcome." or "No problem."?

If a computer can simulate intelligence sufficiently enough, would most
people care philosophically whether it's *actually* intelligent?

Given sufficient speed, a large enough database, and the ability to learn
the kind of response an owner would want (e.g., one owner prefers swerving
or changing lanes to avoid obstacles, while Miss Daisy prefers the car to
stay in the lane and slow down or stop altogether, while another prefers the
robocar to decide based on situational awareness of surrounding traffic,
etc.), a robocar would fit the bill of chauffeur for most people.

Natch there'd always be a percentage of people that would never trust them,
just as there's always early adopters willing to take a risk of being "cut"
to be on the bleeding edge. But then human traffic isn't like being in a
grand central arena--not always.

> --
> Sea Wasp
> /^\
> ;;; Website: http://www.grandcentralarena.com
> Blog: http://seawasp.livejournal.com

-- Ken from Chicago

[Edward A. Falk]

In article <XnsA025AC15865...@69.16.186.7>,

Gutless Umbrella Carrying Sissy <taus...@gmail.com> wrote:
>>
>Unless, of course, something happens that the computer doesn't
>expect, and can't cope with. What are the odds of that, on a random
>public street?

In the video, you can see there's an operator in the right seat.

>Plus, they're still in the alpha stage anyway.

Yep. It will be a while before they're as safe, on average, as a human
driver. It will be longer still before enough statistics have been
collected to convince regulators of this. And even longer still before
the public accepts it.

But I have hope, and every advance is very exciting.

--
-Ed Falk, fa...@despams.r.us.com
http://thespamdiaries.blogspot.com/