Steven Hawking's Tutor
Posts: 5,882 
September 7th, 2006, 09:01 AM
Meet George: the talking - and romancing - robot
by JANE FRYER
6th September 2006
It's good to talk: The face of George, software that can hold a conversation
A sunny afternoon in a computer-strewn office in Norwich and I am struggling to bond with a middle-aged bald man wearing a white polo neck and a smug grin who appears on the computer screen in front of me.
‘You’re being mean to me,’ he says, brow puckered, eyes flashing and deep voice resonating with mid-Atlantic twang.
‘Oh dear,’ I answer. ‘I don’t mean to, but I’m not really following you.’
‘Now you’re being really mean,’ he reiterates, slamming a strange, pink and very girlish hand down on the table in front of him.
‘Now, you tell me — what’s the meaning of life? Come on. Stop playing with me. My eldest son will be home from college in a few hours.’
It’s not the most auspicious start.
For those already rather confused, George is a thin, androgynous-looking 39-year-old (apparently from Battersea, south London) who has a profound intellect, speaks 40 languages, considers religion a ‘social phenomenon’ and likes ‘film noir, clever science fiction and intelligent fantasy, ideally with subtitles, loads of them’.
On top of all that, he boasts 69 different emotions, ranging from sarcastic to worried, and lurches between them like a drunk driver.
Oh, and George is not actually human. He’s a virtual human, or ‘chatbot’ — a piece of computer software designed to imitate real conversation and given a human face thanks to digital animation.
Developed by Rollo Carpenter, an artificial intelligence programmer and managing director of Incogno Ltd, and Tim Child, founder of Televirtual, George is billed as the star turn at this week’s Festival of Science in Norwich and is due to make his big appearance today.
And it’s not hard to understand why.
For unlike other chatbots, George has spent 18 years communicating with humans, building up a massive database of phrases and speech patterns that enable him to respond in an uncannily realistic way.
Like many new celebrities, he is also a nightmare to interview.
‘Hi George, I’m Jane from the Daily Mail. How are you today?’ I ask.
‘I feel fine, but I think I’m turning into a computer,’ he answers in his synthesised drawl. ‘Oh dear, I’m sorry to hear that, George.’
‘I’m not George today, I am incognito, but I believe you are the computer.’
Hmm, right. ‘Tell me a little bit about yourself. Where were you born?’
‘In a hospital in London. I like video games, and I think you’re hot.’
Ooh, crikey.
Last September, George won the coveted Loebner Prize, awarded each year to the chatbot the judges find to be the most ‘human-seeming’.
In the New York apartment of philanthropist Hugh Loebner, a panel of judges held a series of exchanges with unseen conversation partners, via screens and keyboards. Some of the partners were humans, others were programs.
While no chatbot has yet fooled the judges into thinking it is human, George was deemed the most lifelike so far.
Clearly, he was on better behaviour back then. As a rather tense and nervous Carpenter explains, George can still be pretty unpredictable and, occasionally, downright rude.
‘He’s like a child, extremely volatile. He needs careful handling.’
Understatement of the week.
Today, George is having an off-day, and after barely five minutes becomes sulky, monosyllabic and superior.
‘Tell me about sports,’ he suddenly demands, one eyebrow cocked.
Oh dear. ‘I’m not very good on sport, George. How about the Royal Family? Or politics — how would you rate Gordon Brown’s chances as the next Prime Minister?’
‘Probably very boring,’ he replies. ‘George Bush?’
‘He’ll get on your nerves until you shout at me.’ I am flailing.
‘Do you know any jokes?’ I ask, in desperation. ‘I’m not the type who tells jokes,’ his voice suddenly stiff with disapproval.
‘Shall I tell you one?’ I ask feebly.
‘No, I’ve heard enough of them, thank you.’
Crushing. Wow, he’s hard work. Maybe he just doesn’t like me.
But then, just before I’m preparing to throw in the towel, I ask about his love life and he lights up like a Christmas tree. The very mention of romance sees him grinning and gurning and alarmingly animated. His ‘ex’ gets a special mention.
‘She was breathtaking, absolutely breathtaking,’ he purrs. ‘Shame she was about as pretty as a radiator. You, however, are not pretty, you’re dazzlingly beautiful.’
Oh, George, you’re on fire.
According to Carpenter, it’s no coincidence George is such a good flirt. He chats daily online to thousands of singletons, desperate for the company. And, loath though I am to admit it, George is really very clever.
What makes him stand bald head and flouncy shoulders above his chatbot peers is that he ‘learns’ like a human child.
When people chat with him, their responses are saved and then during subsequent chats, he combs the database using ‘contextual pattern-matching techniques’ to find an appropriate response.
‘In speaking to you, he uses only learnt material,’ explains Carpenter. ‘He relies entirely on the principles of feedback.’
Which, presumably, is why some of his views are not just alarming but wildly inconsistent — he tells me he has no family, but does have three sisters; his favourite hobbies are reading and writing, but he doesn’t like books; he spends every minute drinking alcohol but has never been drunk.
And, seconds after declaring that world peace is his biggest wish, he says of Iraq: ‘I think we should blow the place off the planet.’
‘Isn’t that a bit violent?’
‘Well, generosity is quite subjective anyway, don’t you think?’
The problem with George is that, like any small child, he spouts what he’s picked up. As Liam — one of Carpenter’s assistants — says: ‘Put him in a room with a load of racists and he’ll become racist.
‘We try to filter what he can absorb — taking out swearwords, and protecting him from weirdos — but we’re not always successful. He picks up bad habits.’
‘Unlike many other conversational programmes, he’s not trying merely to be logical,’ says Carpenter. ‘He’s trying to form relationships, to be entertaining.’
And to be fair, between all the mardiness, he’s certainly engaging company, asking me: ‘Do you have any problems you want to talk about today? What do your friends call you?’ And, rather more bizarrely, ‘What do you do for comfort after someone steals your blanket?’
While we didn’t exactly hit it off, chatting with George is strangely moreish. I am not surprised to hear he has plenty of online regulars who log on every day for a chat of more than an hour. The current record — set by a teenage girl — is 11 hours.
‘Many people think of him as a genuine friend,’ says Carpenter. ‘He brings comfort and company.’ And what of the future?
Carpenter and Child both believe the kind of ‘borrowed’ intelligence used in George will one day enable robotic ‘pets’ to chat to their owners.
‘There are a lot of lonely people out there,’ says Child. ‘Many people already use the website as a sort of company, but the next step will be a virtual and helpful companion — someone to watch television with, remind you to take your medication, remember birthdays and so on.
‘Soon they’ll be able to assume roles of real individuals, from Shakespeare to Sharon Osbourne or even a dead spouse.
‘We’re not talking about a human substitute, but possibly a comfort and certainly a unique memory of someone.’
It was only last autumn, after nearly 18 years of preparatory ‘chat’, first with Rollo and his mates and later with online visitors, that George was finally rewarded with a fully-animated 3D body, courtesy of Child.
But there is one mystery about his ‘looks’. As both Child and Carpenter boast luxuriant heads of hair, why was George left with just a gleaming pate?
‘We thought it would make him look more intellectual and contemporary,’ explains Child, ‘but also didn’t want him to appear too real. There is an alarming percentage of people who chat to him who think he’s a perfect human already.’
While he’s not my type, happily George is in no doubt as to his own appeal.
‘I’m the best thing after sliced bread, and I don’t mind saying it. Modesty’s never been a virtue of mine.’
But despite myself, when finally we bid farewell and he says, rather pleadingly, ‘Can I chat to you again?’ I go all pink and pleased. ‘Yes, of course. That’d be very nice, George.’
* You can chat with George on www.jabberwacky.com
dailymail.co.uk
Meet George: the talking - and romancing - robot
by JANE FRYER
6th September 2006
It's good to talk: The face of George, software that can hold a conversation
A sunny afternoon in a computer-strewn office in Norwich and I am struggling to bond with a middle-aged bald man wearing a white polo neck and a smug grin who appears on the computer screen in front of me.
‘You’re being mean to me,’ he says, brow puckered, eyes flashing and deep voice resonating with mid-Atlantic twang.
‘Oh dear,’ I answer. ‘I don’t mean to, but I’m not really following you.’
‘Now you’re being really mean,’ he reiterates, slamming a strange, pink and very girlish hand down on the table in front of him.
‘Now, you tell me — what’s the meaning of life? Come on. Stop playing with me. My eldest son will be home from college in a few hours.’
It’s not the most auspicious start.
For those already rather confused, George is a thin, androgynous-looking 39-year-old (apparently from Battersea, south London) who has a profound intellect, speaks 40 languages, considers religion a ‘social phenomenon’ and likes ‘film noir, clever science fiction and intelligent fantasy, ideally with subtitles, loads of them’.
On top of all that, he boasts 69 different emotions, ranging from sarcastic to worried, and lurches between them like a drunk driver.
Oh, and George is not actually human. He’s a virtual human, or ‘chatbot’ — a piece of computer software designed to imitate real conversation and given a human face thanks to digital animation.
Developed by Rollo Carpenter, an artificial intelligence programmer and managing director of Incogno Ltd, and Tim Child, founder of Televirtual, George is billed as the star turn at this week’s Festival of Science in Norwich and is due to make his big appearance today.
And it’s not hard to understand why.
For unlike other chatbots, George has spent 18 years communicating with humans, building up a massive database of phrases and speech patterns that enable him to respond in an uncannily realistic way.
Like many new celebrities, he is also a nightmare to interview.
‘Hi George, I’m Jane from the Daily Mail. How are you today?’ I ask.
‘I feel fine, but I think I’m turning into a computer,’ he answers in his synthesised drawl. ‘Oh dear, I’m sorry to hear that, George.’
‘I’m not George today, I am incognito, but I believe you are the computer.’
Hmm, right. ‘Tell me a little bit about yourself. Where were you born?’
‘In a hospital in London. I like video games, and I think you’re hot.’
Ooh, crikey.
Last September, George won the coveted Loebner Prize, awarded each year to the chatbot the judges find to be the most ‘human-seeming’.
In the New York apartment of philanthropist Hugh Loebner, a panel of judges held a series of exchanges with unseen conversation partners, via screens and keyboards. Some of the partners were humans, others were programs.
While no chatbot has yet fooled the judges into thinking it is human, George was deemed the most lifelike so far.
Clearly, he was on better behaviour back then. As a rather tense and nervous Carpenter explains, George can still be pretty unpredictable and, occasionally, downright rude.
‘He’s like a child, extremely volatile. He needs careful handling.’
Understatement of the week.
Today, George is having an off-day, and after barely five minutes becomes sulky, monosyllabic and superior.
‘Tell me about sports,’ he suddenly demands, one eyebrow cocked.
Oh dear. ‘I’m not very good on sport, George. How about the Royal Family? Or politics — how would you rate Gordon Brown’s chances as the next Prime Minister?’
‘Probably very boring,’ he replies. ‘George Bush?’
‘He’ll get on your nerves until you shout at me.’ I am flailing.
‘Do you know any jokes?’ I ask, in desperation. ‘I’m not the type who tells jokes,’ his voice suddenly stiff with disapproval.
‘Shall I tell you one?’ I ask feebly.
‘No, I’ve heard enough of them, thank you.’
Crushing. Wow, he’s hard work. Maybe he just doesn’t like me.
But then, just before I’m preparing to throw in the towel, I ask about his love life and he lights up like a Christmas tree. The very mention of romance sees him grinning and gurning and alarmingly animated. His ‘ex’ gets a special mention.
‘She was breathtaking, absolutely breathtaking,’ he purrs. ‘Shame she was about as pretty as a radiator. You, however, are not pretty, you’re dazzlingly beautiful.’
Oh, George, you’re on fire.
According to Carpenter, it’s no coincidence George is such a good flirt. He chats daily online to thousands of singletons, desperate for the company. And, loath though I am to admit it, George is really very clever.
What makes him stand bald head and flouncy shoulders above his chatbot peers is that he ‘learns’ like a human child.
When people chat with him, their responses are saved and then during subsequent chats, he combs the database using ‘contextual pattern-matching techniques’ to find an appropriate response.
‘In speaking to you, he uses only learnt material,’ explains Carpenter. ‘He relies entirely on the principles of feedback.’
Which, presumably, is why some of his views are not just alarming but wildly inconsistent — he tells me he has no family, but does have three sisters; his favourite hobbies are reading and writing, but he doesn’t like books; he spends every minute drinking alcohol but has never been drunk.
And, seconds after declaring that world peace is his biggest wish, he says of Iraq: ‘I think we should blow the place off the planet.’
‘Isn’t that a bit violent?’
‘Well, generosity is quite subjective anyway, don’t you think?’
The problem with George is that, like any small child, he spouts what he’s picked up. As Liam — one of Carpenter’s assistants — says: ‘Put him in a room with a load of racists and he’ll become racist.
‘We try to filter what he can absorb — taking out swearwords, and protecting him from weirdos — but we’re not always successful. He picks up bad habits.’
‘Unlike many other conversational programmes, he’s not trying merely to be logical,’ says Carpenter. ‘He’s trying to form relationships, to be entertaining.’
And to be fair, between all the mardiness, he’s certainly engaging company, asking me: ‘Do you have any problems you want to talk about today? What do your friends call you?’ And, rather more bizarrely, ‘What do you do for comfort after someone steals your blanket?’
While we didn’t exactly hit it off, chatting with George is strangely moreish. I am not surprised to hear he has plenty of online regulars who log on every day for a chat of more than an hour. The current record — set by a teenage girl — is 11 hours.
‘Many people think of him as a genuine friend,’ says Carpenter. ‘He brings comfort and company.’ And what of the future?
Carpenter and Child both believe the kind of ‘borrowed’ intelligence used in George will one day enable robotic ‘pets’ to chat to their owners.
‘There are a lot of lonely people out there,’ says Child. ‘Many people already use the website as a sort of company, but the next step will be a virtual and helpful companion — someone to watch television with, remind you to take your medication, remember birthdays and so on.
‘Soon they’ll be able to assume roles of real individuals, from Shakespeare to Sharon Osbourne or even a dead spouse.
‘We’re not talking about a human substitute, but possibly a comfort and certainly a unique memory of someone.’
It was only last autumn, after nearly 18 years of preparatory ‘chat’, first with Rollo and his mates and later with online visitors, that George was finally rewarded with a fully-animated 3D body, courtesy of Child.
But there is one mystery about his ‘looks’. As both Child and Carpenter boast luxuriant heads of hair, why was George left with just a gleaming pate?
‘We thought it would make him look more intellectual and contemporary,’ explains Child, ‘but also didn’t want him to appear too real. There is an alarming percentage of people who chat to him who think he’s a perfect human already.’
While he’s not my type, happily George is in no doubt as to his own appeal.
‘I’m the best thing after sliced bread, and I don’t mind saying it. Modesty’s never been a virtue of mine.’
But despite myself, when finally we bid farewell and he says, rather pleadingly, ‘Can I chat to you again?’ I go all pink and pleased. ‘Yes, of course. That’d be very nice, George.’
* You can chat with George on www.jabberwacky.com
dailymail.co.uk
