Science, technology, and all things awesome with Ian Steadman

RSS

Robots are now capable of learning new skills from each other, without human help

We created them because we need them, but they may well come to realise that they do not need us.

A robot, using RoboEarth, in action. (Photo: RoboEarth)

The “Singularity” refers to a proposed moment in time when the collective intelligence of the machines overtakes the collective intelligence of our humans. At that moment, theoretically, the machines will be able to design and produce even smarter electronic brains, beyond the comprehension of our lower intelligence, in a runaway chain of events that with to unfathomable consequences.

(Terminator fans will recognise this as Skynet.)

Robots are, right now, not very physically capable, and neither are they very smart. As illustrated in one of Randall “xkcd” Munroe’s ‘What If?’ columns, humanity would do pretty well out of any robot uprising that happened, say, tomorrow.

But the robots are getting better everyday:

At its core, RoboEarth is a World Wide Web for robots: a giant network and database repository where robots can share information and learn from each other about their behavior and their environment. Bringing a new meaning to the phrase “experience is the best teacher”, the goal of RoboEarth is to allow robotic systems to benefit from the experience of other robots, paving the way for rapid advances in machine cognition and behaviour, and ultimately, for more subtle and sophisticated human-machine interaction.

RoboEarth offers a Cloud Robotics infrastructure, which includes everything needed to close the loop from robot to the cloud and back to the robot. RoboEarth’s World-Wide-Web style database stores knowledge generated by humans – and robots – in a machine-readable format. Data stored in the RoboEarth knowledge base include software components, maps for navigation (e.g., object locations, world models), task knowledge (e.g., action recipes, manipulation strategies), and object recognition models (e.g., images, object models).

To emphasise what this is referring to - it’s a network in the cloud, called RoboEarth, the lets robots learn new skills from each other without human input. It’s not the Singularity, but it’s a start.

Here’s a video of an early version of RoboEarth, being used in 2011 by a robot called AMIGO in a care home setting (complete with banging soundtrack):

Today's the day the team at Eindhoven University that created AMIGO and RoboEarth will run a full-scale test. Four robots will be placed into a hospital-like room, turned on, and allowed to learn as they go along what it is that they have to do. To do that, they'll have to download instructions from each other so that they can work out how to tend to a fake patient's needs, be it preparing medicines and food or making sure that they don't bump into each other.

To understand why this is exciting, let's look at Tesla, whose electric cars - which only have a few moving components - are mostly controlled by the effects of software. Instead of costly product recalls, Tesla has been able to wirelessly beam updates to each car in the wild, which within minutes has solved potential overheating issues or extended the range of its vehicles by a few miles.

Our future robotic infrastructure will likely rely on a similar conceit, with incremental improvements in software just as important as they are for home computing. And, just as Wikipedia works through an ever-improving group effort of editing and refinement, something like RoboEarth - with input from both humans and robots - could allow the rapid propogation of better software than any centralised alternative.

Quite a lot like how human societies change, learn, and improve, in fact. Maybe Skynet isn't such a wild comparison.