Rather than worry about robots overtaking us, it's more interesting (and realistic) to consider how we might collaborate with our machines. At Institute for the Future where I'm a researcher, we have forecasted how the real power of automation will come from "humans plus machines." BB pal Ken Goldberg, director of UC Berkeley's People and Robots Initiative, and his colleagues are making that real through their pioneering work on cloud robotics and human-centered automation. Forget the Singularity, Ken says. It's all about the "Multiplicity." From Ken's op-ed in the Wall Street Journal:
Most computer scientists agree that predictions about robots stealing jobs are greatly exaggerated. Rather than worrying about an impending Singularity, consider instead what we might call Multiplicity: diverse groups of people and machines working together to solve problems.
Multiplicity is not science fiction. A combination of machine learning, the wisdom of crowds, and cloud computing already underlies tasks Americans perform every day: searching for documents, filtering spam emails, translating between languages, finding news and movies, navigating maps, and organizing photos and videos…
While scientists still don't understand Multiplicity very well, they are discovering clear benefits to machine diversity. Researchers have developed a family of techniques known as "ensemble learning," in which a set of specialized algorithms work together to produce a single result. One variant, known as "random forests," was developed by Leo Breiman and Adele Cutler at the University of California, Berkeley. They proved that in complex problems with noisy data, a group of "decision trees" will always outperform a single tree—so long as the trees are sufficiently diverse.
By the same token, the benefits and challenges of human diversity have been recognized for centuries in political science, economics and sociology.
"The Robot-Human Alliance" (WSJ)