There's a lot of interest in both industry and the military in developing free-space optical communications systems. The basic idea is to use a laser to transmit signals at optical frequencies over distances ranging from a few kilometers to hundreds of kilometers. Potential advantages of such a scheme are the high bandwidth of communications, meaning lots of information can be transmitted very fast, and the highly directional nature of the signal, making it highly secure. Disadvantages of such a scheme include the requirement of direct line of sight between transmitter and receiver, and more significantly the distortion induced in the beam by atmospheric turbulence.
Does one really need a laser to make a viable optical communications system, however? I saw one talk at Photonics West, one of the best talks I've seen in a while, which demonstrates pretty convincingly that you do not...
The strong distortion of laser fields in turbulence is due in large part to the coherent nature of a laser beam, which means that phase distortions in the laser field result in interference of the beam with itself and large fluctuations in the beam intensity. These intensity fluctuations result in data transmission errors, and these errors are serious enough to make free-space optical communication generally impractical.
However, it has been observed numerous times throughout the years, since at least the 1970s when people started studying atmospheric light propagation of lasers, that a field which is partially coherent (partially pre-randomized at the source) will propagate better in atmospheric turbulence and will result in better communication. In recent years this topic has been studied theoretically in great detail.
The idea is simple enough: intensity fluctuations are due in large part to interference effects, and interference can be reduced by reducing the spatial coherence of the field. Most people have assumed that the best way to do this is to start with a laser beam, and distort its field somehow to make it partially coherent*.
A collection of ham radio enthusiasts in Australia, however, took a different approach. In 2005 they used high-power 'Luxeon' LEDs, i.e. non-coherent sources, combined with a large Fresnel lens to collimate the non-directional light emitted, and managed two-way audio communications over a distance of 160 km (100 miles!) between mountaintops in Tasmania. The experiment, by Mike Groth and Chris Long (the latter of whom gave the talk in San Jose), was at the time a record for two-way transmission. Perhaps even more impressive is the fact that the system was built with an eye for minimizing costs (they don't have massive external funding), and each transceiver cost on the order of $100!
They have a website dedicated to a discussion of the experiment and the history of optical communications; it's well worth a read. Science is often viewed as a 'clique' of insiders, but experiments such as this one show that intelligent amateurs can make magnificent contributions that are well-received.
* It's worth noting that the high emphasis on lasers in communications can be understood historically: in the 1960s, when optical communications first began to be studied, lasers were the only light sources which were bright enough and efficient enough to be used. This is, however, no longer true...