In the news this week is a story from Duke University about using a small metamateral array to harvest power from ambient microwave energy, such as cellular telephone and Wi-Fi signals. Power (or energy) harvesting from ambient microwave and RF signals is of huge interest around the globe, particularly for distributed wireless sensor networks and the much-touted Internet of Things.
The key feature of interest is that an impressive conversion efficiency of 37% has been achieved, which makes them comparable to typical solar panels. Unfortunately, the article referenced above mistakes output voltage for energy, making a comparison to USB-based chargers that is somewhat misleading.
Update (18-Nov-2013): an interesting review article on this story puts the work into context by talking about wireless power transfer, including a review of early work by Tesla and more modern efforts, such as for charging electric vehicles. Some additional technical details (e.g., incident power) missing from the original story are given. The comparison of "power harvesting" and "power transfer" is sometwhat misleading, as harvesting and transferring power wirelessly have different requirements, but it seems that the issue is more to do with the original story, as input powers of 30 mW or so are required to produce the results highlighted - this is far in excess of "ambient" power levels for harvesting scenarios, and more in the realm of wireless power transfer. That is not to suggest the result is not valid or or laudable, of course, but it highlights that press releases can (often) get the technical parts wrong!
You can read the paper online, so you can assess the results for yourself.