PEOPLE THINK, CREATURES?
BUT DO DIGITAL
Apart from the merits or deficiencies of Penrose’s approach, it illus- trates one of two main corollaries that accompany theories of con- sciousness; namely, that machines can never think or be conscious in the way that people are—accompanied, of course, by the conflicting belief that machine consciousness is possible. In the early days of AI, the answer seemed simple.The pioneering AI researchers considered thinking to be the processing of information, which is, in turn, the manipulation of symbols; hence, minds are simply systems for process- ing symbols.As it happens, our own minds are based on brains made of neurons, but the physical nature of the processor is unimportant. Thus, whether the “brain” consists of billions of living nerve cells, a stack of silicon chips, or for that matter, one of Isaac Asimov’s positronic units, the important thing is that symbols are meaningfully manipulated. When that happens, thinking, and perhaps even con- sciousness, occurs.
This view is often called, semijocularly,GOFAI—“good old-fash- ioned artificial intelligence”—and is now recognized as falling short of a complete approach to machine intelligence. Decades ago,as com- puter programs began to manipulate symbols in meaningful ways such as carrying out mathematical proofs, proponents of GOFAI felt we were well on our way toward full AI. But as understanding grew, we came to realize that GOFAI omits some aspects of cognition—for instance, the sensory experience of smell—which might not be repre- sented by words or other symbols inside our minds.
Today, with AI and cognitive science far more advanced,and theo- ries of consciousness abounding, there is ammunition for those who believe that machines can think and for those who don’t. In Daniel Dennett’s view, a human mind that is thinking is running what amounts to a computer program that processes information. To Dennett, this scenario opens the door to machine thought. He claims that “a computer that actually passed the Turing test would be a thinker in every theoretically interesting sense,” and adds “I do think it’s possible to program self-consciousness into a computer.” But