'Learning Robot' Crowdsources for Information, Skills From Internet, Other Bots
Monday, August 1, 2011 at 7:24AM Image a futuristic breed of robots that can continually learn new information and skills, through good old-fashioned “thinking” and by asking other robots for help. Thanks to the work of Tokyo’s Hasegawa Research Group, this possibility is closer than you might think.
The Self-organizing Incremental Neural Network (SOINN) is the team’s self-replicating neural network, which seeks to recreate human-like educated guesses in robots. The amazing technology enables bots to learn from mistakes and successes, understand its own strengths and weaknesses, and contemplate possible solutions.
For example, if a SOINN robot is asked to pick up three objects, it will recognize that it only has two hands and know that it must pick up two at a time. Even if the bot has never been asked to do a particular task, the problem-solving capabilities of SOINN are well beyond that of most modern robots.
The best part? SOINN robots can educate themselves using web-based information or skills learned by other robots. Similar to our reliance on web tools like Google – not to mention the shared experiences of our fellow man – SOINN will create a broader robot “community” with open-source knowledge.
In this way, SOINN is similar to Qbo, an amazing open-source robot that crowdsources object-identification data so it can recognize unfamiliar objects.
“Thinking about artificial intelligence in the real world, actual environments are inevitably more complex, and they change quickly,” the Hasegawa Research Group said in a news release. “So it’s necessary to have a learning mechanism that adapts to the situation. Also, because new situations emerge, it’s also necessary to have the ability to keep learning new information on the spot.”
Suppose, for example, a SOINN robot is caring for an elderly person, who requests a cup of tea.
“The robot doesn’t know how, so it asks robots around the world how to make tea,” the group explained. “Suppose, for example, that a robot in the UK tells it how to make British-style tea. We think this robot will become able to transfer that knowledge to its immediate situation, and make green tea using a Japanese teapot.”


Reader Comments