Nothing has ever truly passed the Turing test.
A few chat bot have fooled some rather simply minded and unexpecting people, but nothing has ever fooled 30% of judges during a 5-minute test, like Turing proposed.
Never even came close.
Actually, they have. Roger Schank designed a program that succeeded in understanding simple stories. Roger Penrose provides an example in
The Emperor's New Mind, pg. 18.
Basically, a computer is fed a story such as, "A man went into a store and ordered a hamburger, enjoyed it, paid the bill and left." The computer is asked whether the man ate the hamburger. It would then pass out yes or no based on the story. In essence, this machine passed a Turing Test.
Of course, the counterargument is that the machine had no
concept of eating, or of hamburgers, but rather that the computer doesn't strictly need to understand these concepts in order to process yes or no.
Imagine if the story was written in Chinese. You could have an algorithm that if you saw a character that looked like <insert chinese character here>, you would say yes. Thus, you understood the Chinese story without understanding it at all.
