What Is Turing Test? • A person and a computer, being separated in two rooms, answer the tester’s questions on-line. • If the interrogator cannot tell whether he/she is talking to is a person or a computer, then the computer is viewed to be as intelligent as the person.
What Is Turing Test For? • To check whether a machine is as intelligent as humans.
Why Turing Test? • One needs a benchmark for machine intelligence. • It’s difficult to measure internal intelligence mechanism – because we do not quite understand it yet, and different persons may have different mechanisms. • “External performance” is a objective, measurable and consistent benchmark.
Feature of Turing Test • It gives an objective notion of intelligence. • It tests only intelligence of “brain without body”, whose inputs are from the keyboard whose outputs are sentences on screen. • Its judgment is based on the reactions of the machine rather than its pro-actions.
Test Intelligence or Test Consciousness? • Turing test was designed to test machine intelligence. • Its principle can be used to test machine consciousnesses such as emotions and self-awareness.
Intelligence vs. Consciousness • Different between intelligence and consciousness? • Higher level consciousnesses pre-require higher level intelligence, though it is controversial whether lower level consciousnesses must be associated with certain level of intelligence.
Is a Machine Passed Turing Test Intelligent? • Its “art intelligence” and “sport intelligence” are not tested. • It passes the “intelligence of thinking”. If a human-like robot showed such intelligence, then we would not distinguish it from a real person in terms of intelligence of thinking. • Some insist that intelligence or not should be determined by its internal mechanism …
Ambiguities in Objectives of Turing Test • Intelligence – defined in terms of “results” or in terms of “process”? • Understanding? • Semantics vs. symbolic?
Searle’s Chinese Room • A person knowing nothing about Chinese is locked in a room with a rule book about Chinese symbols. • He answers questions in Chinese given to him from the outside according to the instructions in the rule book.
What Searle’s Chinese Room Shows • From the stand point of a person outside of the room, the person in the room understands Chinese well since he answered the questions in Chinese perfectly. • But the person inside the room knows not a single word of Chinese!
Searle’s Arguments • Programs are entirely syntactical. • Minds have a semantics. • Syntax is not same as, or sufficient for, semantics. Therefore, computer programs are not minds
J. Wang Flaws in Searle’s Arguments • He is assuming what he is proving: • He is proving “computer programs are not mind, i.e., they are entirely syntactical”; • He assume “programs are entirely syntactical”. • Lack of definition of “semantics”. • Lack of rigorous proofs of: • Programs cannot be semantics • Programs are entirely syntactical.
What Searle Wants to Say • A computer may look intelligent, understanding, touching, and even emotional, but it achieved them by following digital symbol processing rules, which are the simulations of intelligence, understanding, touching, and emotions. • Simulation of an explosion is never an explosion.
J. Wang Example of E.T. • Suppose that E.T. is locked in the Searle’s Chinese Room, and he/she (it?) does the translation work perfectly. • Would we hesitate to say, “Hey, E.T. understands Chinese!” • Would we care how E.T. does the translation?
Strong AI • Strong AI asserts that human’s mind (intelligence, consciousness, emotion, spirit and soul) can be programmed on computers. • Advocators include R. Kurzweil, M. Minsky, H. Moravec, B. Joy, J. Hall, W. Bainbridge, T. Kaczynski • Opponents / skeptics include J. Searle, J. Hawkins, R. Penrose, D. Hofstadter.
Anyway Better Than Turing Test? • Despite of the criticism on Turing Test, no one has so far has put forward an alternative way of testing machine’s intelligence / consciousness, which is as (or more) acceptable, practical and effective as (than) Turing Test.