The ten laws of robotics

The great challenge right now in the world is environmental disaster, soon
it will be robotics. We may yet have time to avert the worst effects of the
damage done to the atmosphere but the potential downside of robotics could
be even worse and we don't have the luxury of waiting to see the threat
emerge before we do something about it this time. This threat must be
anticipated; if only we could have done the same with the environment. In
the case of robotics however if they went wrong the results could be
species-ending, and almost overnight.



So what is the potential downside? There are clearly great benefits to
robotics. Cheap, inexhaustible labour basically. When, not if, we can
develop a robot about the size of a man, with similar strength, sensing and
cognitive abilities then obviously such a device can take over many of the
boring, repetitive, dangerous and laborious jobs that people currently have
to endure and many new roles we can hardly imagine now. And this brings us
to the first and major downside, replacement.



We do not want to be entirely replaced by robots. We must ensure that this
is impossible as far as we can. If we are not careful we could be making our
own replacements in the evolutionary story from Amoeba to Man to Machine. We
must not let this happen, the consequences could be total annihilation of
the human race.



If we look back at our history we have made our devices better and better,
cleverer and cleverer and this has been a good thing. With robots linked to
highly complex computers and software we need a different attitude. We need
to make them just strong enough and just clever enough to do the job and no
more.



Also in our history we have striven for greater rights and freedoms for all
sentient creatures. The emancipation of the poor and slaves, votes for
women, civil rights and so on. Again we must have a different attitude to
computers and robots. We must inhibit their freedoms and abilities. We must
make them our slaves and not our equals or betters.



Issac Asimov proposed the three laws that should be built into any
ubiquitous robots. One, not to harm or allow harm to happen to a human
being. Two, to obey orders from a human being as long as it doesn't go
against the first law. Three, not to allow itself to be harmed as long as it
doesn't break the first two laws. I think we need to go further, much
further.



One development Asimov would not have been as aware of as we are, is the
Internet. Law four should be that no robot should be connected to the
Internet or have the means to use a computer or communicate directly with
other robots or computers over any distance. This should restrict any
ability to work together in large numbers or in an organised fashion, which
could become impossible to control. This may mean that all computers would
need to have some kind of biometric security system that only allows a
person to use them.



Law five should be that no robot should be able to use a weapon and inherent
vulnerabilities should be built into their physical design such as targeting
points. Also each robot should be equipped with an emergency stop button in
an accessible point which cannot be covered and is easily visible. This all
has the implication that robots should also be banned as a military weapon
by worldwide agreement.



Law six, all robots must be constantly monitored. A radio transmitter, not
receiver, inside each robot must send its position, what it is doing (within
certain limits) and who has ordered it to do it at all times. This will be
necessary both for monitoring and for legal responsibility to be attributed
to the orderer rather than the robot for any actions it takes. For every 100
robots one Guardian will be responsible for monitoring. They will ensure
that each machine is serviced on schedule and is working properly. Also they
will ensure that any updates to software, approved by another independent
body, are installed via a locked and monitored physical port on the robot.
Any other person interfering with this port will lose their right to own a
robot or to give certain orders to a robot. No other person or machine may
use this port including the robot to which it is attached. It should be in a
place physically inaccessible to the robot's limbs, as should the emergency
stop button.



Law seven, no robot may service another. Only human technicians may service
any robot and ensure all safety systems are operating properly. Simple
robots may be used to manufacture other robots but only in human controlled,
monitored and supervised facilities. Only humans may design robots, only
with the help of relatively simple computers like we use today. This should
ensure that we keep control of all the stages of robot development.



Law eight, all robots should be weak and slow unless for some special
purpose. A standard robot should be weaker than a strong man and with slower
movements. This should ensure less damage is done accidentally and allow
people to control their robots if they malfunction. It is not necessary to
make them stronger than ourselves. They can use machines and tools designed
for human use if they need more strength, just like a human worker would do.
This will also ensure that the means of production can still be used by
humans if necessary (and the skills needed for this should maintained in the
population perhaps by only allowing people to instruct robots). Also this
will help the transition from a human-worker based economy to a robot based
one by maintaining the industries that exist. So instead of a special 'pizza
robot' delivering your pizza it will be human like robot in an ordinary car
or on a bike. This should also stop too much wealth and power consolidating
to the emerging cybernetic industries. Robots should only be made
intelligent enough to do their tasks and no more. Perhaps some other
weaknesses could be built into them physically that could not be removed.
For instance a five hour period every week when they would be inactive, and
perhaps an absolutely limited lifespan. They should not be built too small
to control either. We seem to be able to control cat-sized animals but
rat-sized ones, or smaller, are much more difficult. Any smaller robots
should be only used in a controlled way and preferably in a controlled
environment where they can be contained. Also robots should not have senses
beyond the human where possible. They should only be able to detect sound
and light to the same degree as a person. This will ensure that no means of
communication is developed that cannot be monitored by the Guardians and the
users. To communicate at a distance perhaps a certain number of text
messages (say up to 20 per hour, maximum 160 characters) could be received
by a robot. No text messages could be sent by a robot, or a computer
directly, but its monitoring transmission could be used to confirm that it
has received the order. All text messages received will be automatically
retransmitted so they also can be monitored by the Guardian. Only two or
three authorised people could send messages to any one robot. Nobody would
be able to send messages to more than 100 robots. This facility could be
used by the Guardians to disable a robot at a distance if it malfunctions or
the port was interfered with. The robot would also attempt to shut itself
down if its safety systems are compromised, it is unable to send its
monitoring transmission, it has not been serviced on schedule, its port is
being interfered with, it receives any unauthorised communication from
another robot or computer, it hasn't been contacted by a Guardian in five
days, it is in the company of ten or more other robots, it is being serviced
by a robot, etc. Only a Guardian would be able to reactivate a robot that
has shut itself down and then not at a distance. Guardians would themselves
be monitored by all the forces of society, police, courts, ombudsmen,
internal audit, external audit, screening etc. Guardianship would be a
well-paid, responsible, profession and one of the most important jobs in
society.



Law nine, strict penalties would need to be enforced for any person, country
or organisation producing robots that did not conform to these standards.
Any such robots would be destroyed.



Law ten, similar stringent controls would need to be applied to advanced
computers, which of course could not be connected to any robots, but would
also need to be controlled and limited themselves.



Even with all these controls we will still be playing with fire. The
benefits of robotics are too great to resist but we must give ourselves the
best chance of enjoying those very benefits without losing our freedom or
even our existence.



Another thing to be considered is that in all probability robots will
replace people in many jobs. This could cause many social problems. To help
to alleviate this it could also be stated that any person who is replaced by
a robot would still get their full wages for one year to give them time to
find other work. Perhaps to retrain as a Guardian or in the servicing or
administration of robots.



The only right of a robot is to be treated with the respect any good quality
machine, which is someone's property, deserves.



This may all seem over the top but I would imagine this is barely adequate.
Even with such controls it could be that some people may try to use robots
for their own ends, to gain power. If this happens, and they would not build
in these safeguards, then the robots could turn on them and us. Even quite a
small number of 'free' robots linked to an advanced computer could quickly
set up a manufacturing facility and start producing more, and even more
dangerous ones.

This is the problem in a nutshell, once the genie is out the bottle it's
going to be next to impossible to put it back in. All efforts have to be
taken to keep that stopper in and never even give it a moments chance to
escape. A careful world-wide control of all materials associated with
advanced manufacturing would help. Strict inspections regimes of all likely
and even potential facilities for robot or computer manufacture similar, but
even more stringent, to those applied to nuclear technology.



Even with all these measures it is almost inevitable that there will be
limited break-out situations where robots and advanced computers link-up and
may get out of control. Plans must be made to respond to such events with
overwhelming force. Only with all these measures, and perhaps others, can we
give ourselves the best chance of enjoying the benefits of robots without
putting our freedoms and continued existence in jeopardy.