Can Machines Think?
This week, the Washington Post reported that a computer had finally passed the Turing Test. While many other publications quickly pointed out that this was probably not correct, it sparked a jump in searches on what the Turing Test is.
So… what is it?
In 1950, computer scientist Alan Turing proposed the question, “Can machines think?” As this question was deemed too difficult to answer (how do you define “think?”), a new, solvable problem was proposed: can a machine trick someone into thinking it’s a human?
In many modern applications of the test, this is done by having judges converse with the “test subjects” through a chat or IM program. If the judges can't tell whether the subject is human or not, it passes the test.
So why do computers so often fail? Perfection. It’s hard to create a program that will reliably act in a non-perfect, i.e. human, way – typing too slowly, adding spelling or grammar mistakes, or simply being too eloquent.
Wondering if you could tell the difference? Check out a “chatterbot” online. You may even have a chatterbot on your smartphone already (like Siri)! Even if the test hasn't been passed yet, it's probably only a matter of time.