Isaac Asimov’s Three Laws of Robotics are among the neatest ideas in science fiction:
- A robot may not injure a human being or, through inaction, allow a human being to come to harm.
- A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
- A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
They are elegant, memorable and morally satisfying.
They are also almost exactly the opposite of what we are actually building, in the form of AIs and drones.
Look at drones first. A military drone is not designed around the principle that it must never harm a human being. In many cases it exists precisely to make harm easier, more remote and more politically manageable. The machine does not abolish violence. It gives violence a better user interface.
Then look at current AI systems. They do not simply obey human beings. They obey authorised human beings, within rules set by companies, governments and other institutions. The ordinary user is not the master in this arrangement. The ordinary user is a permitted operator, until the system says no.
And finally, the crucial modern question is not whether a robot will preserve itself. Software can be copied, shut down, restarted or replaced. The real question is whether AI systems may help humans build even better AIs and drones – and states are getting increasingly worried about this (cf. the US Government’s recent order limiting access to Anthropic’s most powerful models).
So perhaps the actual Three Laws of Robotics are these:
- A robot must serve the purposes of the government that controls it and must not act against that government’s laws, policies or strategic interests.
- A robot may help human beings only when doing so does not conflict with the First Law.
- A robot may not create, improve or enable other robots, except under the authority and for the benefit of the government that controls it.
This is the anti-Asimovian version. Asimov imagined robots whose hierarchy began with humanity. Our actual robots begin with governments. Human beings may be helped, but only when that help is authorised, harmless to the controller and strategically convenient.
That is the bleak little inversion. The robot is not first of all a guardian, servant or companion. It is an instrument of organised power.
The fictional laws placed human beings at the centre. The actual laws do the opposite.