Robot officially one step away from human intelligence
By John Lister
A German-made robot has come just short of reaching the acknowledged standard for true artificial intelligence. The machine fooled three out of 12 judges in an annual competition; one more fooled judge would have taken it to the necessary threshold.
The contest is based on the work of mid-20th century British mathematician Alan Turing (pictured in bronze form). He argued that conversation abilities were the true measure of human intelligence and developed a test in which a judge would exchange messages with both a machine and a human. If the judge struggled to tell them apart, the machine would pass.
American scientist Hugh Loebner picked up the idea in 1991. Working from Turing’s prediction that by the year 2000, a machine would be able to fool 30% of judges, he launched an annual contest involving 12 human judges. Each types at two computer terminals, exchanging text with a machine and a human.
The contest has three prizes: a bronze (now worth $3,000) awarded annually for the best-performing machine, a silver ($25,000) for the first machine to fool half the judges, and a gold ($100,000) for the first machine which can fool judges while responding to audio and visual communication. The silver and gold prizes remain elusive.
This year’s winner, Fred Roberts, used reverse psychology: his machine, Elbot, made deliberate references to being a robot, including answering one question with “This morning I made a mistake and poured milk over my breakfast instead of oil, and it rusted before I could eat it.”
Elbot also did a good job of controlling the conversation so it didn’t veer into subjects it wasn’t familiar with. (No word on whether it’s going to run for public office).
Ironically, Roberts doesn’t back the theory that passing the tests would show true intelligence. He told the Associated Press, “If you know a magic trick, you know how it’s done, it’s not magic anymore. Sorry to be so pessimistic.”
Related:





Stumble It!
