|
The Three Laws of Robotics were an imperfect set of rules at best.
Though Asimov did foresee how the laws would introduce
complications, they were never more than that. Asimov had an
incredible faith in technology and never really grasped how something
might be dangerous in spite of its safeguards, that safeguards can
fail spectacularly, or that an an extremely complicated machine might
make safeguards at best a way of decreasing the probability of
disaster rather than preventing it.
Asimov claimed that he and Campbell came up with the three laws
because robots are machines and man would naturally design them with
safeguards. And yet, Asimov had great difficulty thinking of
robots as machines. From the get go, he attributed emotions,
motives, and consciousness to his creations. Sentience is not
presented as something that a robot might one day attain, but as an
inherent property. It's plain that
Asimov had very little understanding of how computers actually work,
even in terms of the primitive machines of '40s and '50s. He
tended to explain his robots in terms of analogue machines with
"behaviour" resulting from various "potentials" in the robot's
circuitry, or by drawing analogies with human psychology.
If we were talking motor cars this would be good old fashioned
anthropomorphism.
 If you think about it,
the three laws are as big a heap of budgie doo doo as you can find. The
idea behind the laws is interesting, but in terms of real machines
they make no sense. Take the first law, for example. It
says, "A robot may not injure a human being, or, through
inaction, allow a human being to come to harm." Feed that into
a real robot and it would be like trying to run a car on water.
This law, like the others, is a collection of abstract concepts and
moral imperatives that are meaningless from an engineer's point of
view. No robot could ever understand the concept of "human
being" or "harm" or "action" or "inaction." Heck, it couldn't
understand "may." Instead, a real robot would have to be told
that this set of inputs in these circumstances correspond to the
definition of human in this situation. If such a set of
conditions as indicate a human at location X are true, then this
series of actions Y may not intersect at location X and this
series of actions Z must be performed in relation to location X to
achieve condition A. Or something like that. In other
words, its good old fashioned programming with all the outcomes
predicted and their eventualities accounted for.
As for the second law: "Robots must obey?" Machines obey
anything you tell them to do. That's the nature of machines and
often what makes them so dangerous. No "law" is required to
enforce this. And as for preventing this obedience from causing
harm, well, the first law is really nothing more than a subroutine "If
this, then stop," or whatever. Hardly the Golden Rule.
|
