ABSTRACT
Could machines ever be conscious? Ways to find out include reverse-engineering human machines or artificial ones. A brief history of artificial intelligence (AI) leads from early automata through calculating machines and ‘Good Old-Fashioned AI’ to connectionism, embodied cognition, and swarm robotics. Turing asked ‘Can machines think?’ Attempts to pass his famous test are reviewed, including chess-playing machines and programs that can converse with humans. Strong AI posits that machines can really think while weak AI claims that they can only simulate thinking. A similar distinction applies to consciousness (strong AC versus weak AC), but how can we test whether a machine is really conscious? Some dualist and religious arguments claim that only biological systems can be conscious. Searle’s ‘Chinese Room’ thought experiment purports to refute strong AI; its many criticisms and variants are reviewed. Others argue that conscious machines are perfectly possible and may already exist. Among strategies discussed are building machines that model attention, self, and the external world; machines with global workspace architecture; and social robots which people react to as though they are conscious. A final approach is to build a machine that believes it is conscious even though this is an illusion.
