Do you read me HAL? Affirmative Dave ... How we are on brink of ‘real’ conversations with robots
HUMANITY is on the brink of having “proper intelligent conversations” with robots.
The idea having a chat with a robot has long been part of film and fiction – from the malevolent Hal in 2001: A Space Odyssey, to Star Wars android C-3PO.
However, trying to speak to the robots that are currently in our lives today – such as the Amazon robot voice assistant Alexa – can be a pretty cumbersome activity often ending with the frustrating response “I’m sorry, I can’t help you with that”.
Yet is seems that within the next few years it will be possible to have proper conversations with artificial intelligence systems, such as Alexa, whose sales are currently booming. Jon Oberlander, professor of epistemics [the scientific study of knowledge] at Edinburgh University, whose research involves trying to find out how to get computers to talk or write like people, said speech recognition systems such as Alexa and iPhone’s Siri had come on “leaps and bounds”.
But he pointed out they were still limited by being unable to recall what had happened moments ago in conversations.
Oberlander, who will today receive Edinburgh University’s prestigious Tam Dalyell Prize for Excellence in Engaging the Public in Science, said one of the most exciting areas of research in the field at the moment was a focus on building more sophisticated “dialogue machines”.
He said: “There are certain kinds of dialogue which are well managed by computer systems now, particularly these transactional dialogues where you are booking a ticket, for example. But as soon as you start having a chat about an item in the news, for example, it becomes incredibly open-ended. Getting to do that kind of flexible dialogue is really hard. There is no question that it is challenging, but I would say with the number of people who are working on this, we will see moderately impressive conversational assistants within the next five years.” Amazon is running a global competition to challenge university students to build a “socialbot” which can hold a conversation with humans. Two teams from Scotland are involved – from Edinburgh and Heriot-Watt University – and the winners of the $500,000 prize will be announced in November. If the socialbot can converse coherently for 20 minutes, another $1 million will be awarded. Oberlander said there had been acceleration in the development of artificial intelligence, with Google’s AlphaGo system – which can learn from its mistakes – making headlines last year after beating a top human player at the board game Go.
But he said it was also important not to underestimate the challenge of replicating behaviour which comes naturally to humans. He pointed to the work of his Edinburgh University colleague Professor Sethu Vijayakumar, which is focused on creating a robot which walks like a person.
“We take dialogue for granted and walking down the street also comes easily to people,” Ober- lander said. “But getting a robot to walk like a human being is incredibly challenging and you have to learn a huge amount about how human beings do it in order to emulate that on robot platforms.”
The development of artificial intelligence has triggered some dire predictions. Professor Stephen Hawking has previously said full artificial intelligence could “spell the end of the human race”. But Oberlander has a more optimistic vision of humans and machines working together.
Oberlander will deliver a lecture this evening during the award ceremony, which is being held as part of Edinburgh International Science Festival.