Spend a few moments with a chatbot, and it quickly becomes clear that the program is a bot and not a human being. Still, a program that is indistinguishable to a human from another person instead of a computer has been a longstanding dream of researchers, dating all the way back to the creation of the first digital computers.
Mathematician Alan Turing (1912-1954) devised a test to determine if a machine really could pass for a human in a 1950 paper, “Computing Machinery and Intelligence.”
In the paper, he poses a question: “Can machines think?”
Turing proposed a way to answer this: a version of a party game known as the “imitation game” where a person tries to determine which person is which based on answers to questions.
In this game, a human would be in a room, while a computer and a human being would be in other rooms. The human and computer would answer questions, and the player would try to judge whether the responses came from a human or computer. To make things fair, the responses would be based on text only.
If the player couldn’t tell the human responses apart from the computer, then machines could be truly deemed intelligent. This test has become known as the “Turing test.”
Some AI programs might have this effect. Joseph Weizenbaum’s ELIZA program, which is modeled on Rogerian psychotherapy, has caused some people to believe that they were talking to a human.
ELIZA is a relatively simple program that its creator said had almost no knowledge about the real world.
It appears that some programs can pass for human beings, but most of these have relatively narrow fields of expertise. Researchers believe that they have a long way to go before machines can qualify as truly intelligent. Some major researchers, such as Gary Kurzweil, have offered prizes for people who create programs that really can pass for human beings.
