There's an article (subscription only, unfortunately) out that has gotten some attention, discussing whether artificial molecular machines will "deliver on their promise". The groups that wrote the article have an extensive track record in synthesizing and characterizing molecules that can undergo directed "mechanical" motion (e.g., translation of a rod-like portion through a ring) under chemical stimuli (e.g., changes in temperature, pH, redox reactions, optical excitation). There is no question that this is some pretty cool stuff, and the chemistry here (both synthetic organic, and physical) is quite sophisticated.
Two points strike me, though. First, the "promise" mentioned in the title is connected, particularly in the press writeup, with Drexlerian nanoassembler visions. Synthetic molecules that can move are impressive, but they are far, far away from the idea of actually constructing arbitrary designer materials one atom at a time (a goal that is likely impossible, in my opinion, for reasons stated convincingly here, among others). They are, however, a possible step on the road to designer, synthetic enzymes, a neat idea.
Second, the writeup particularly mentions how "efficient" the mechanical motions of these molecules are. That is, there is comparatively little dissipation relative to macroscopic machines. This is actually not very surprising, if you think about the microscopic picture of what we think of as macroscopic irreversibility. "Loss" of mechanical energy takes place because energy is transferred from macroscopic degrees of freedom (the motion of a piston) to microscopic degrees of freedom (the near-continuum of vibrational and electronic modes in the metal in the piston and cylinder walls). When the whole system of interest is microscopic, there just aren't many places for the energy to go. This is an example of the finite-phase-space aspect that shows up all the time in truly nanoscale systems.