The Turing Test and Chatbots

(1) Introduction

This conversation is between a real human and a computer program called Mitsuku.
Programs like Mitsuku are called “chatbots.” Mitsuku was the winner of the Loebner Prize in 2013, an annual competition to find a program that can pass the Turing Test.
The test, designed to measure artificial intelligence, was invented by British mathematician Alan Turing.

(2) The Origin of the Turing Test

Fifty or sixty years ago, computers were very simple machines, but even then scientists believed that one day machines would be able to “think” and would probably be as intelligent as humans.
The question was, “How can we measure the intelligence of a machine?”

Alan Turing, who helped break the German Enigma code during the Second World War, came up with an answer.
He said we should ask, “Can this computer talk? Can it have a conversation like a human?”

If it can, then it is intelligent and it can think.
In a Turing Test, judges sit at a screen and have a chatroom conversation with the chatbot program.
They don’t know if they are chatting with another person or with a chatbot.
After exchanging messages for five minutes, the judge decides whether they are chatting with a human or a machine.

(3) Predictions and the Loebner Prize

Turing predicted that by the year 2000, the average person “will not have more than a 70% chance of making the right identification.”
In other words, computers would trick the judges 30% of the time.

An American called Hugh Loebner was fascinated by Turing’s idea, and in the early 1990s he offered a prize of $100,000 to the creator of the first chatbot to pass the Turing Test.
To win the prize, a chatbot must convince at least 30% of the judges that it is human.

Many chatbots have entered the competition, but so far no chatbot has won the big money prize.
Mitsuku, however, came very close.

(4) Mitsuku and Human Conversation

In the conversation at the beginning of this article, Mitsuku is A and the real human is B.
You can tell that A is not human because at one point in the conversation, A says, “Humans are not too intelligent at times.”
Although Mitsuku failed to win the $100,000, it probably won’t be long before a chatbot is able to fool the Loebner judges into thinking it is a real person.

(5) Is the Turing Test a Good Measure?

But is the Turing Test a good way to decide if a machine is intelligent?
Critics argue that the chatbots in the competition are merely imitating humans.
Humans are the only animals on Earth that can speak, and that’s why Turing chose to focus on it.

However, critics say that machines do things we can’t do.
For example, Google can search hundreds of millions of websites for a single word in seconds, or a NASA computer can control a rocket from Earth to Jupiter.
Even smartphone apps can do extremely impressive things — often more useful and interesting than a chatbot’s answers.

A chatbot, they say, is really nothing more than a successful liar.

(6) Humans and Machines

Fans of the Turing Test, on the other hand, believe that humans are themselves machines — just far more complex ones.
As philosopher and scientist Daniel Dennett said in a recent interview:

“It’s not impossible to have a conscious robot. You’re looking at one.