Artificial intelligence and humanity
The power of being human
May 5th 2011 | from the print edition
The Most Human Human: What Talking With Computers Teaches Us About What It Means To Be Alive. By Brian Christian. Doubleday; 303 pages; $27.95. Viking; £18.99. Buy from Amazon. com, Amazon. co. uk
WHAT does it mean to think? The question has bothered philosophers for millennia and computer scientists for decades. In 1950 Alan Turing, who pioneered artificial intelligence, devised a test to determine how well a machine can mimic mankind in an attempt to address whether computers can think. He argued that if a computer could fool a person into thinking that he was corresponding with another person, then it would be impossible for that person to tell whether or not the computer was thinking. Human judges have since been asked to distinguish between two correspondents, one man, one machine, in a competition held every year for the past two decades. Two years ago Brian Christian landed the task of persuading these judges he was not a computer. His account of the experience is entertaining and informative.
Human beings like to think of themselves as possessing unique capabilities such as being able to use tools, employ a language with syntactic rules and process complex mathematical equations. Aristotle argued that only people could reason. Yet the development of ever more powerful and complex computers has chipped away at these claims: the invention of logic gates, for example, allowed machines to make deductions and when IBM’s Deep Blue computer defeated a chess world champion, Garry Kasparov, in 1997, mankind seemed to lose ground to machine.
Mr Christian is optimistic about humanity. He points out that computers have areas in which they are competent and others in which they are not. A computer can guide a missile, he notes, but not ride a bicycle. People may be replaced by robots on farms, in
factories, call centres and laboratories but machines can only do work that lends itself to automation. He describes the rise of artificial intelligence as “maggot therapy: it consumes only those portions [of the job market] that are no longer human, restoring us to health.”
Likewise the game played by Deep Blue serves to identify the whereabouts of creativity in chess, he argues. A computer can win through brute force by taking opening moves that lead to a game it has already played out and won. The game becomes interesting only when it treads on new territory and the computer is forced to recalculate in response to an original placing of pieces rather than continue down a well-worn path.
To prepare for his role in the Turing test held in 2009, in which contestants compete for the Loebner prize for the most convincingly human computer and the most convincingly human human, Mr Christian analyses how computers fool people into thinking they are human and what persuades people that they are indeed talking to a machine. He recounts a tale told by Robert Epstein, who co-founded the prize with Hugh Loebner. Mr Epstein was deceived into thinking he was exchanging long letters online with a woman over a period of four months when he was, in fact, talking to a chatbot. Mr Christian identifies what distinguishes a computer from a real person: a machine that analyses vast numbers of previous conversations to identify a suitable response becomes inconsistent, for example, because its identity is actually that of many individuals.