In 1950, Alan Turing wanted to answer the question, “Can machines think?” However, due to the difficulties involved in defining exactly what constitutes “thinking,” he chose to answer a closely related question: “Are there imaginable digital computers which would do well in the imitation game?” This question, which is known popularly as the “Turing Test,” has become the most famous test in the field of artificial intelligence. Recently, a piece of software known as Cleverbot took the “Turing Test.” How did it fare?
What was the Turing Test?
The “Turing Test” was first discussed in Alan Turing’s paper, Computing Machinery and Intelligence. The basic idea is to measure the ability of a machine to mimic intelligence. If a human engages in separate text-only conversations with another human and a machine and is unable to tell which is which, the machine passes the test.
Recently, a piece of software known as Cleverbot was put to the Turing Test. Thirty volunteers at the Techniche festival in India held on-line conversations for four-minutes. Half of the participants spoke to a human, half spoke to Cleverbot. The conversations were displayed for an audience. Afterward, the volunteers and the audience voted on whether the responses were spoken by Cleverbot or by a human. Out of 1,334 votes, Cleverbot was believed to be human 59.3% of the time. Sadly, the actual humans only rated 63.3%.
“The world had better decide rather than me – it’s either passed or it’s come very close in this particular test.” Rollo Carpenter, Cleverbot Developer
Cleverbot doesn’t “think,” at least not in the way we would define the word. Instead, it bases its answers on records of its previous conversations. The version used in the test runs 42 searches before furnishing an answer. Still, since Cleverbot fooled people more than half the time, it arguably passed the Turing Test. So, does that mean that artificial intelligence has passed some sort of threshold?
Guerrilla Explorer’s Analysis
The answer is a definite “sort of.” Although Cleverbot managed to fool the participants and spectators, this represents a very small segment of the population. More importantly, there’s obviously a large difference between how Cleverbot operates and the way a human thinks. Still, Cleverbot does learn from previous conversations and this is obviously a form of intelligence.
If you’d like to give a “less intelligent” version of Cleverbot a try, check it out here. While the real machine chose from 42 separate answers, this version only gets three choices. Here’s a snapshot of my own conversation with Cleverbot. You can decide for yourself how well it did.
Cleverbot: My favorite color is blue.
Cleverbot: Because purple is the color of royalty.
User: But you said blue before.
Cleverbot: That wasn’t me. I was someone else that you were talking to.
Well, nobody’s perfect.