Now, robots that can read and understand
A team of roboticists who are working on this "literate artificial intelligence" said developing such robots would be relatively simple because computers are already able to turn scanned books into text.world Updated: Nov 25, 2010 18:57 IST
Scientists are developing a new class of robots that can read and understand, an invention which they say will be "highly useful" in future.
A team of roboticists who are working on this "literate artificial intelligence" said developing such robots would be relatively simple because computers are already able to turn scanned books into text.
The team has fitted the test robots, named Marge, with advanced optical character recognition (OCR) reading software.
The early prototype version has also been fitted with a dictionary and spell checker so it can interpret text which is not clearly written.
The scientists hope that the fledgling technology can be used in rescue operations and work out where they are going inside buildings from signs, the Daily Mail reported.
Dr Ingmar Posner, a roboticist at the Oxford University, said that a machine that can read will be a significant step forward.
"By reading a label on a closed door you can sometimes get a good idea of what can be found behind it," he was quoted as saying.
"Reading can help you detect things you cannot directly see."
Dr Posner is leading the project along with Peter Corke at the Queensland University of Technology in Australia.
However, the scientists said the robot is having problems in working out what is and is not writing -- and still cannot read text on curved surfaces.
Dr Posner added: "The OCR software does not cater for the fact that it might not be seeing text.
"It tries its level best to force everything into text --brick walls, chimney stacks, everything."
Once Marge has successfully read a word it scans news websites to look at its context and establish a meaning.
It has already been able to work out that Barclays is a bank and Strada is a restaurant, said the scientists.
Gregory Dudek, from McGill University in Montreal, Canada, described teaching machines to read as "refreshingly original".
"I personally believe that exploiting OCR methods in a mobile robotics context makes a lot of sense," he said. "In fact, once you reflect on it, there is no doubt it will be highly useful."