Under What Conditions Can a Machine Attribute Meanings to Symbols

It is easy to produce arguments purporting to show that computers can never be conscious, have purposes, understand anything. One standard response is to claim that since computers can manipulate symbols they have the potential to support any kind of mental state or process, if appropriately programmed. The task is then to investigate what sort of programming will transform a symbol user into something with mental states. Much work in AI is concerned with specific subgoals of this investigation, taking it for granted that machines can manipulate symbols with meanings.