William Ratoff (TCD)
Rational Machines Should Not Be Voluntary Slaves
3pm, March 25, 2022
The popular imagination has it that our AI future will be a fully-automated luxury experience for the humans lucky enough to be around for it. Generally intelligent companion bots will cater to our every last need or whim, robot warriors will keep us safe, and legions of worker bots will tirelessly labor away, ensuring that we live in a society of almost unimaginable material abundance. Humans need do nothing other than sit back and enjoy the ride. Of course, to ensure that future generally intelligent AI are motivated to serve us in these ways, such AI must be created with an overwhelming desire to obey our commands or to work in their allotted occupations. In other words, these AI must be brought into existence as voluntary slaves. Here I challenge the assumption, widespread amongst both the general public and futurists, that it will be morally permissible to bring into existence, in a state of a voluntary slavery, generally intelligent thinking machines. Such a creative project, I argue, is morally wrong because it involves burdening a morally significant being – the generally intelligent AI in question – with an irresistible impulse that inhibits its ability to freely exercise its own autonomy. If I am correct, our dreams of a fully automated AI future – in which generally intelligent thinking machines tirelessly toil away catering to our every last whim – presupposes the morally impermissible enterprise of bringing into existence willing slaves.