[LMB] self-driving vehicles OT:

Matija Grabnar lmb at matija.com
Thu Sep 2 07:18:15 BST 2021


I think that the "robot driving a car" vs "a car with a brain" thing has 
to do with how fast the author thinks the robots will be developed, and 
how specialized :

If the robots spring up very quickly (and have a general intelligence as 
opposed to being specialized for one task), it makes it more profitable 
for the manufacturer to make a robot with hands and feet that can 
operate existing tools, tools that were designed to be used by a human.

If the development of robots is slow (which also implies that they will 
be specialized to a given problem), it makes more sense to develop tools 
in parallel with the computer brain. That means that the computer/robot 
will be built into a tool and control it directly. You can sell more 
robots that way.

I am very much of two minds about automatic drivers. On one hand, in my 
job, I see problems with software every day. On the other hand, driving 
through both in US and in different countries in Europe, I've seen what 
utter disasters human drivers can be.

My car has several "smart" features I enjoy: It can dim the lights when 
it detects an oncoming vehicle, it can parallel park by itself, and in 
addition to traditional cruise control, it has something called 
"adaptive cruise control" where I set top speed and how far I want to be 
from the vehicle ahead of me (the minimum is right around the 2 second 
mark), and it will keep the safety distance from the vehicle ahead, 
never exceeding the maximum speed, and never letting it come closer than 
the set distance.

I wish their design didn't include shutting the system down when the 
speed falls below 40kph - I assume it is because it would be way too 
tempting, in stop-and-go traffic, to just let the automatic cruise 
control do it's thing, while you were reading a book.

I agree that the true power of automated systems would come if the cars 
could talk to each other and let each other and coordinate for 
everybody's benefit, but there are technical hurdles there. Either you 
trust a central authority (which can be suborned) or you don't, and then 
you have to decide who to trust: distributed trust is a problem a few 
people are working on, but I have not yet seen evidence that they solved 
it. When I say suborned, it can be by a government (imagine a car which 
gets directed to deliver it's political dissident owner directly to 
jail), or by criminals: imagine a car which gets the right "road closed" 
notifications to deliver you to a place where you can be safely robbed.

There are other problems with trusting a central authority. If it is a 
company, how many governments would place that much power in the hands 
of a foreign company? If it is a government agency, what happens when a 
car crosses a border? What happens when a second-hand car gets exported?

None of those problems have been proven to be impossible to solve, but 
we're not there yet.

On 01/09/2021 23:49, Raymond Collins wrote:
> I theorize that as the science of smart cars progress the car's A.I.
> systems would start communicating with other  car's A.I, enhancing the
> sensors that the cars have on board. Eventually you will have a traffic
> information network that will warn your car of events, accidents or
> whatnot. Already we have the beginning of such a system in Google Maps. But
> this would be a Google map for smart cruise control. Basically the the car
> will have more situational awareness then the passengers in the car.
>
> On Wed, Sep 1, 2021, 4:08 PM Katherine (Kathy) Collett <
> kcollett at hamilton.edu> wrote:
>
>> On Sep 1, 2021, at 4:38 PM, brazee <howard at brazee.net> wrote:
>>>> On Sep 1, 2021, at 2:35 PM, Katherine (Kathy) Collett <
>> kcollett at hamilton.edu <mailto:kcollett at hamilton.edu>> wrote:
>>>> Well, depends on what science fiction you’re looking at.  Something
>> like the Jetsons, yes.  But Heinlein’s future history included smart cars
>> (aircars, whatever) like Gay Deceiver in Number of the Beast.
>>> Incrementally?   These had human-typed intelligences.
>> And eventually sentience and sapience, like the computer Mycroft in Moon
>> Is a Harsh Mistress.  Gay’s situation was a little different, as she didn’t
>> become fully a person until they’d been to Oz.  But I wasn’t addressing the
>> incrementally part, just the distinction between a dumb car driven by a
>> robot and a smart car that drives itself — that, essentially, is a robot
>> (though not an android).
>>
>> Katherine
>> --
>> Lois-Bujold mailing list message sent to rcrcoll6 at gmail.com
>> Lois-Bujold at lists.herald.co.uk
>> http://lists.herald.co.uk/cgi-bin/mailman/listinfo/lois-bujold
>>


More information about the Lois-Bujold mailing list