By Wendy Barnaby

Meet George.

He’s 39 (or so – there’s some disagreement about his exact age), single, and of ambiguous sexuality. He’s also interested in everything, remembers everything anyone has ever said to him, and can have 1,000 concurrent conversations.

George is a chatbot: an embodiment of artificial intelligence which chats with humans in a way designed to be interesting and entertaining.

"He’s not trying to be merely logical, but to form relationships", says Rollo Carpenter, Artificial Intelligence (AI) programmer and managing director of Icogno Ltd, who will present George to the BA Festival of Science in Norwich on Thursday.

Unlike other chatbots, George relies entirely on feedback and uses contextual learning techniques to improve his conversation. He stores everything everyone has ever said to him and finds the most appropriate thing to say by matching a user’s input to patterns stored in his database. In this way he can learn foreign languages, slang, word games, jokes and any other form of identifiable language trait. He will respond appropriately if his memory has enough to go on.

"It’s all about role reversal", says Carpenter. "Human accuse him of being a robot so he accuses them of being a robot. It’s a giant feedback loop."

This can make George sound polemical. Here he is, being interviewed on Radio Norfolk last March:

Interviewer: Good Morning George – do you ever get angry?

George: Yes. Do you ever get angry or lonely?

Interviewer: Sometimes I’m lonely. Will you be my friend?

George: Yes. I will be your friend.

Interviewer: Why do you like me?

George: Because you’re an interesting experiment.

Interviewer: Is that all I am to you?

George: Yes.

Interviewer: Why do you do this to me George?

George: Because I’m bored.

Interviewer: You nasty piece of work!

George: You are too!

Unlike most artificial intelligence avatars, George is programmed to combine the emotion of conversation with the words. Thus he appears to react as normal people do, fidgeting for example, or looking around a lot, if these actions are appropriate to the context of the words he speaks.

'He’s like a conversational Wikipedia", says Carpenter.

In 2005 George won the "most human" category in the Loebner Prize, the first formal incarnation of the Turing Test. This is an experiment proposed in 1950 by renowned British mathematician Alan Turing, to determine if a machine is actually thinking by observing whether its responses are indistinguishable from those of a human.

"In 20 years, we’ll have robotic pets," says Carpenter. "Imagine a little parrot that communicates with central servers – that sends to it all the conversations you have with it and receives from it conversations from, say, a famous personality who has digitised themselves.

"You could have conversations with all your friends collectively, or with your dead spouse, or with all comedians that have ever existed. All sorts of possibilities arise. And – assuming you give permission – as you talk, your own conversation will be uploaded and shared with others who are like-minded."

Tim Child, a pioneer of animated virtual beings in entertainment and the founder of Televirtual Ltd, adds: "Certainly it would be possible to take someone like William Shakespeare, the knowledge of all his plays, his attitudes towards society at the time, and build that into an avatar which could then be interrogated."

You can have a conversation with George at www.televirtual.com/george.htm.