The programmable world. It's an interesting concept, but not one in which we have extremely detailed specifications and strongly bound software systems that must be built to interact with each other. I think many of us have assumed that's how we would enable the further computerization of our physical world. As an example, we'd have cars that communicated with the road, with other cars, with semi-central authority(ies) that might be managing our interactions, all of the infrastructure pre-built.
However that isn't necessarily what will happen. In this O'Reilly piece, the author notes cheap sensors can read the color of lights in the same way that humans or my Lego Mindstorm can detect color. Traffic sensors need not communicate with cars; these sensors could instead just identify cars and count them, measuring their speed and adjusting traffic lights based on actual conditions. Imagine the future when more and more sensors gather their own data and make decisions.
In a sense, this is how the Google self-driving car works. Rather than depend on infrastructure and external programming, the car gathers its own data and adjusts its behavior depending on the interpretation of the data. Whether or not you like the idea of self-driving cars (I do), the idea of an autonomous device acting based on programming and a large amount of data is a fascinating move forward in computing.
I think this shows that data gathering, processing, and analysis will become a more important part of our future computing worlds. Some of this data will be transient and discarded, but lots will be stored. We'll use data to ineract with the real world immediately, but we will also perform analysis later and reprogram our devices to operate better in the future. That means there will be new, and more, opportunities for those of us working with data.