Wired brings up a very old idea in AI. You should not have to write code that computers use to operate with. That's messy, prone to error and expensive, In particular it is hard to maintain as our world changes. Computers should have 'brains' like we do, and should be able to learn simply by being shown lots of examples of what is 'right' under given conditions.
And to a degree we are getting closer to doing this, but the data and learning methods and context still need to be arranged very carefully and tested repeatedly. And usually differently for different contexts. Still the goal, the great challenge remains. There is something called 'Machine Learning' which is very clever, but it is not yet close to universal learning.
Further, good piece in Wired. It is not quite the 'end of code'.