Hi robot

UK researchers have received 1m Euros (£700,000) to investigate intelligent robots that can understand the ambiguities of natural speech and work more effectively alongside humans.

UK researchers have received 1m Euros (£700,000) to investigate intelligent robots that can understand the ambiguities of natural speech and work more effectively alongside humans.

The four-year project forms part of a wider European Commission initiative consisting of seven academic teams from around Europe known as Cosy (Cognitive Systems for Cognitive Assistants).

While companies such as Honda and Sony proudly showcase walking, dancing bipeds, the team at Birmingham University is working on the less tangible and possibly more painstaking business of improving artificial intelligence.

Dr Jeremy Wyatt, who heads the Intelligent Robotics lab at Birmingham, said his group is looking at a number of significant problems involved in building ‘thinking’ robots.

Wyatt referred to the recently published UN World 2004 robotics report – the theme of which was ‘A robot in every home?’ He explained that in order to arrive at such a situation, devices must be able to interact with us and satisfy our expectations about acceptable behaviour.

One way in which they plan to do this is in the area of object recognition. Wyatt explained that his team will be working with computational linguists and cognitive psychologists in order to take ideas on how humans recognise things.

He said that the team will begin by mounting a vision system on a mobile platform to watch an arm on a separate table and report back in natural language what it sees.

An apparently humble aim, but it will nevertheless break new ground in the area of artificial intelligence. One of the problems, explained Wyatt, is that people tend to assume that robots are far more advanced than they actually are.

Possible reasons for this are the visually impressive biped robots developed by Honda and Sony. While these are sophisticated in engineering terms they work in very different ways from human beings. Thus walking, which with humans requires little thought and is driven largely by localised muscle activity, gravity and momentum, is such a computationally intensive process with robots that there is little capacity for more abstract, and superficially more mundane, cognitive processes.

‘The way your body works has a big impact on what you have to do in order to be smart,’ commented Wyatt.

Understandably keen to play down the notion that we will soon have domestic androids to do our bidding, Wyatt said that a more likely scenario is niche appliances steadily becoming more intelligent.

‘Your hoovering robot might still be a flat disc on the floor but it will be able to recognise certain types of object and figure out when the house is empty – stuff will appear, but not in the way that science fiction sometimes predicts.’

Wyatt said that he is cheered by the huge amount of money that the commission is investing in cognitive systems. Over the next five years, Cosy and related projects are expected to receive around 80m Euros (£55m).

He added that although the Cosy project is currently wholly academic and will not begin to wield commercial applications for around 10 years, Honda is nevertheless curious about it. ‘The problems we want to solve are the same problems that they want to solve.’