AW: [aiethics] Thought experiment: Punishment of robots...

[Christian DroЯmann]

Alicebot AI Ethics Committee - http://www.alicebot.org



> Sorry, software can't be killed.  Think of ELIZA.  Her creator
> tried to stop
> her.  Yet once the genie
> was out of the bottle she became the most popular AI program of all =
time,
> until maybe ALICE.

But ELIZA was rather harmless...

> How could you erase all of the millions of backup copies of a piece of
> software?  Maybe some
> kind of virus could try to track them down, but could it really
> destroy all
> the backup copies on
> CD and whatever else we have in the future?

Question: Would there be any need to do so?
IMHO Once AI develops a type of conscience, every single copy of a =
certain
program becomes an individual, as all of the copies run in different
environments with different input...I would consider a particular =
program as
a species of AI programs and all of the copies would belong to this =
species,
but would develop different personalities.
Just because one of the copies turned out to be evil does not mean that =
the
whole species is...
Example: If a shepherd's dog got mad and bit your son, would you say =
that
the whole species should be killed or just that one particular dog? =
Sure,
all shepherd's dogs are equal to some extent and they all COULD bite a
human, but many of them just don't...

> I think the stupidest Star Trek plots revolve around "killing" =
Commander
> Data or The Doctor.
> Turning off one copy of an AI is hardly going to stop it.
> Don't they do backups in the 23rd century?

I don't agree with you here for the reasons stated above..
The PERSON Commander Data CAN be killed, he is an android and he is one =
of a
kind...the doctor is a program, so killing him would be more difficult =
as
backup copies could exist...but if the whole ship is destroyed the =
series of
holodocs just like the doctor will continue to exist, but THE doctor can =
be
considered "dead"...

Christian