Long time lurker, commenting because I know something about this one (doing my PhD in thermoelectrics).
First of all, you _can_ use thermoelectrics to cool things like CPUs or fridges, but don't expect to generate any energy from them when you do it because you need to be putting electricity into the system, essentially carrying the thermal energy with it. You will cool one end and heat the other end. If you've ever heard of a Peltier cooler then you know what I am talking about.
A good background can be found here: http://thermal.ferrotec.com/technology/thermal/thermoelectric-reference-guide/
Second, this is something people have been messing around with the nanostructure of tellurium alloys for ~20 years or so, with the sole purpose of reducing thermal conductivity. The figure of merit for thermoelectrics is ZT = thermopower^2 x electrical conductivity x temperature / thermal conductivity. You can't increase electrical conductivity without reducing thermopower and increasing thermal conductivity (as there is a lattice and an electrical contribution). Thermopower is more or less a function of the number of carriers (lower is better) and their effective mass, so this is difficult to increase without durastic changes in the crystal structure or killing electrical conductivity. This leaves thermal conductivity. If you increase disorder in the material you make it harder for thermal energy to travel through it, which as lead to lots of research on how you manage this without messing up your carrier conduction. These are known as PGEC (phonon glass electron crystal) materials.
Third, there are lots of applications of these (in heating/cooling and power conversion) if they can be made efficient and cheap. Anywhere you have a heat source pretty much. To use the classic car analogy, BMW, Ford, GE (amongst others) are looking at using a thermoelectric module to generate power for the car from the waste heat in the exhaust gases from the engine. This would increase the power of your engine by removing the alternator and also make the car lighter.
The problem is the efficient and cheap part. These kinds of thermoelectrics are based on tellurium, an element about as abundant in the earth's crust as platinum, but to my knowledge isn't specifically mined for. Most other elements involved are toxic heavy metals (Pb, Sb, Bi, etc.)... so these aren't exactly nice things to have around or to make.
This is where oxides come in. Made of lighter, more abundant, less toxic elements they are much cheaper to make (not just sourcing the materials, heath and safety too etc.), and are stable at much higher temperatures. As you know from Carnot, the higher the temperature a heat engine works at the more efficient it becomes; rather than 900 K (600C) you're looking at more like 1300 k (1000C) and upwards. Current high ZT oxides are things like NaxCoO2 and Ca3Co2O6, which have layered structures; one part is great at absorbing thermal energy (due to Na disorder for example) and the other is good at conducting electricity (like the CoO2 portion of NaxCoO2)
The way I see this paper: great proof of concept, PGECs are doing what they say on the tin and this will be great for low T applications. But for high power generation we need something more like the oxides which are cheaper, easier to produce, and work at higher temperatures.