THE RISE OF THE MACHINES is getting closer as researchers find ways of using the information on the internet to get machines to learn.
Perhaps more terrifying than that, they're asking you to help.
Cornell University has created Robo Brain, a robot with the potential to learn new skills by downloading how-to guides and Youtube videos from the web.
Recently acquired skills when we logged on to look included how to hold an orange to prevent squishing, the correct way to hold an umbrella, the position of a human standing wearing shoes, demonstrated through a heat map, and most existentially, what a chair is.
The Robo Brain website also affords the opportunity to speed up its learning by manually providing instruction on how to tackle specific tasks.
Thankfully, Robo Brain cannot learn emotion. This is not an experiment in artificial intelligence, but rather one exploring self-reliance and self-expanding programming.
The idea of Robo Brain is to provide a template for future technologies such as self-driving cars, and perhaps even robot butlers, to better understand their environments. The video below shows Robo Brain in mid-July. Since the video clip was captured it will have learned many more commands and concepts.
A more immediate example of this type of technology comes from Google, which is believed to be preparing a "Knowledge Vault" of all the world's facts, which if combined with Google Now would provide a far more comprehensive service than it does now. This is an extension of the present "Knowledge Graph", which according to New Scientist has started to slow down its expansion, as its crowdsourced information begins to get somewhat convergent in a manifestation of diminishing returns on investment.
Cornell and Google are, however, far from the only organisations working on this kind of machine learning. IBM's Watson has already proven itself on US quiz show Jeopardy, while Microsoft recently revealed that it too is adding a machine learning component to its Azure cloud service. µ
It's time for our regular two-step through the Google news
Bug bounty offer: accepted