Person or computer: could you pass the Turing Test?

This year marks the 100th anniversary of the birth of famed British mathematician Alan Turing, who designed a test in which a human converses with another human and a computer program, and then attempts to determine which is the human and which is the computer.

As mentioned already on The Conversation and other websites, this year marks the 100th anniversary of the birth of famed British mathematician Alan Turing.

According to some, computer intelligence is on course to match human intelligence by 2045.
(Credit: Sybren A Stüvel)

The outline of his remarkable life and sad ending has by now become fairly well known.

Turing laid numerous foundation stones of modern computing, ranging from the deepest mathematical nature of computing (using what are now called Turing machines, he provided the modern approach to incompleteness (PDF) and undecidability) to specific issues of practical design; he also contributed to mathematical biology (morphology) and much else.

At the same time, he played a key role in the British government's breaking of the German Enigma code at the now-fabled, but then ultra-secret, Bletchley Park, thus arguably accelerating the end of the second world war.

Alan Turing
(Credit: Revolweb)

Turing was many other things: a world-class marathon runner, a troubled homosexual and an atheist who famously said:

"The universe is a differential equation. Religion is an initial condition."

His achievements are perhaps most succinctly summarised by Harvard scholar Steven Pinker, who declared:

"It would be an exaggeration to say that the British mathematician Alan Turing explained the nature of logical and mathematical reasoning, invented the digital computer, solved the mind-body problem and saved Western civilisation. But it would not be much of an exaggeration."

One of Turing's many signal contributions was a 1950 article that defined what is now known as the Turing Test.

In it, he proposed a test in which a human "converses" with two entities — one human and one computer program — over a text-only channel (ie, a computer keyboard/screen), and then attempts to determine which is the human and which is the computer.

(Credit: Indieink)

If after, say, five minutes of testing, the majority of human interrogators are unable to determine which is which, Turing said that we could claim the computer system has achieved a certain level of intelligence.

Turing's article even anticipated several possible objections to his test, including mathematical and philosophical objections, which continue to be debated to the present day.

Some potential questions might not be "fair" to a computer. And we all have human acquaintances who might be judged "computer" in such a test.

In the decades since 1950, when Turing proposed the test, it has been widely influential in directing progress in the computing field in general, and in artificial intelligence in particular.

Some early attempts at Turing Test programs pointed out both the promise and the perils of this enterprise. In 1966, computer scientist Joseph Weizenbaum created a program, known as ELIZA, which identified keywords in text typed by a human, and then responded with some sort of clever but enquiring response, in the style of a psychologist interviewing a patient.

ELIZA in action.
(GNU Emacs ELIZA example image by Ysangkok, LGPL v2.0)

Although some subjects were genuinely surprised to discover the "psychologist" was a computer, to more sceptical testers, its weaknesses quickly became evident.

The present authors do remember enjoying playing with it when personal computers first allowed for relaxed therapy sessions.

Progress in the field languished somewhat during the '70s and '80s, but since 1991, there has been an annual Loebner Prize in artificial intelligence, in which ELIZA's children — now called "chat bots" — compete to pass the Turing test.

Two recent advances have dramatically enhanced interest:

  1. The ready availability of many terabytes of data, from technical documents on every conceivable topic to the growing personal databases of "lifeloggers"

  2. Sophisticated statistical (computational and mathematical) techniques for organising and classifying this data.

This technology was perhaps brought to the public eye most effectively with the recent defeat of two champion human contestants on the American quiz show Jeopardy by an IBM-developed computer system known as Watson. The video below is illustrative.

Watson is now rapidly moving into specialisations for medicine and voice recognition, among other things. IBM clearly views Apple's "intelligent personal assistant" Siri, available with the iPhone 4S, as a main target of competition.

Meanwhile, Google and AT&T are working on similar systems, according to a recent UK report.

Among other things, Watson-type technology offers amazing opportunities as an intelligent assistant for mathematical research and other specialist fields — both for use by specialists and outsiders.

So far, no computer system has passed the Turing test, according to the strict rules of the Loebner Prize competition, but they are getting close.

The 2010 and 2011 competitions were won by a chat-bot computer system known as "CHAT-L", made by artificial-intelligence programmer Bruce Wilcox.

In 2010, this program actually fooled one of the four human judges into thinking it was human.

(Credit: Photo Extremist)

All of this raises the question of whether a computer system that finally passes the Turing test is really "conscious" or "human" in any sense.

These issues were summarised by the University of Bourgogne's Robert M French in a recent Science article:

All of this brings us squarely back to the question first posed by Turing at the dawn of the computer age, one that has generated a flood of philosophical and scientific commentary ever since.

No one would argue that computer-simulated chess playing, regardless of how it is achieved, is not chess playing. Is there something fundamentally different about computer-simulated intelligence?

French is among the more pessimistic observers. Others, such as the American futurist Ray Kurzweil, are much more expansive.

He predicts that in roughly the year 2045, machine intelligence will match, and then transcend, human intelligence, resulting in a dizzying advance of technology that we can only dimly foresee at the present time — a vision outlined in his book The Singularity is Near.

Only time will tell when Turing's vision will be achieved. But civilisation will never be the same once it is.

Further reading:

A version of this article first appeared on Math Drudge.

This article was originally published at The Conversation. Read the original article.


Join the discussion

Conversation powered by Livefyre

Don't Miss
Hot Products
Trending on CNET


Mac running slow?

Boost your computer with these five useful tips that will clean up the clutter.