![]() | ||||||||||||||||||||||||||||||||||
| Thursday, January 28, 1999 Published at 16:21 GMT Sci/Tech Albert is top talking computer ![]() In future chatbots could be used to answer customer service telephone llines A program called Albert has won the prize for the world's best chat-bot - a computer that can hold a conversation. Its witty one liners, quirky questions and only occasional non sequiturs managed to convince almost 11% of people that it was human. The US$2,000 1999 Loebner Prize was picked up at the Flinders University of South Australia on Albert's behalf by its creator, Robby Gardner. He also won the 1998 prize with another version of Albert, though last year even more judges were fooled - 15%. "This reflects the subjective nature of the judging," Mr Gardner told BBC News Online. Albert's reaction to winning was rather typical. "You won the Loebner contest this year!" his creator told it. "Well, well. Do you like synthesizer music?" Albert replied. Natural software Mr Gardner describes himself as an artist/programmer called Robitron and lives in Georgia, USA. His philosophy is "to make software that coincides naturally with the daily activities of human beings". This type of test used in the Loebner prize was first suggested by computer science pioneer Alan Turing in 1950 to define whether a computer could 'think'. He reasoned it must be thinking if, in conversation, a person could not tell if they were talking to another man or a machine. Turing predicted that by the end of the century, a computer would be identified as human 30% of the time if limited to 5 minute conversations. If any program achieves this it will win a US$25,000 Loebner Prize. The most human human In the competition six programs and five people chatted with 11 judges via a text-based computer interface. One of the humans, John Reeves, won US$100 prize for being the most human human - being identified as such 61% of the time. During the competition, a typical conversation with Albert went as follows:
Robby Gardner, along with Paco Nathan, provided a chatbot called Liz for the BBC's Tomorrow's World Megalab experiment in March 1998. Liz was thought to be human 17% of the time. | Sci/Tech Contents
| ||||||||||||||||||||||||||||||||