Traditionally, many people have imagined that while "routine" tasks can be automated, "non-routine" tasks cannot. However, advances in processing power, data storage capability and algorithm design mean that "non-routine" tasks can increasingly be automated—and so this commonly held view is no longer as reliable as it was in the past. This has four important implications for thinking about the future of work: for the limits of machine capabilities, for the pervasiveness of automation, for the 'skill-blindness' of technological change and for the uncertainty that clouds the future.