The Turing Test

July 17, 2008 at 12:44 AM | Posted in history | Leave a comment

In 1950, Alan Turing, a computer science pioneer, suggested a simple test to assess the thinking power of machines.  A human would converse with either a program or another human through a computer interface and would try to guess which it was speaking to.  Programs consistently capable of fooling its judges into thinking they were human could be said to emulate intelligence.  Turing figured that trying to measure whether programs really thought would be very difficult — after all, defining creativity or intelligence has been the source of philosophical debate for centuries.  He predicted, though, that within a few decades, computers would be good enough at his test that arguments over whether they could think would become irrelevant.

A quick look at a hilarious transcript from the Loebner Prize competition shows that his prediction hasn’t come true.  Each year, chat bots like George, pictured above, are entered into this competition and the most convincing one wins a cash prize.

These bots may not even come close to passing as human, but they sure are entertaining to talk to!  Try the following:

  • Eliza – one of the first programs of its kind, written in the 60s
  • A.L.I.C.E – winner of the 2000, 2001 and 2004 Loebner Prize
  • Jabberwacky – winner of the 2005 and 2006 Loebner Prize

Jabberwacky and A.L.I.C.E both learn from their conversations with different people but Eliza only looks for key words and responds with stock sentences.  Loops, conditionals and a way to read user input are enough to replicate most of its behavior — cool!

Create a free website or blog at
Entries and comments feeds.