If I shout at the sky, will some molecules reach escape velocity? Will the sound waves, as they move higher and through less dense air, conserve energy to the point where some molecules of the rarified atmosphere attain escape velocity?
 A: I'll just give a short outline (many caveats though):
Energy in sound waves drops off as the square of the distance (a sound wave spreads out as a sphere from your mouth). If we do not take dissipation into account, you need to compare the maximum energy of your shout and divide it by $R^2$ with $R$ the distance you want to consider. Compare the kinetic energy that a particle would need to have to escape the atmosphere (just plug in the escape velocity at that particular height above you) with the energy that is left of your shout at that height. (Note this assumes the particle is not moving with some velocity itself.)
Generally, dissipation will play a big role, though, and I suspect your shout will hardly do anything. ;)
A: Molecules in the outermost layers of the atmosphere are always reaching escape velocity - but there is sufficient statistical fluctuation that you will never, ever be able to demonstrate that your shout made a particular molecule escape. Let's do some math.
Assuming that your sound wave is still a sound wave (rather than a shock wave) when it leaves your gigantic parabolic mirror, then the peak amplitude can be 2 atmospheres (making the valley "vacuum"). It's possible to make a "boom" that is louder (think atom bomb, Mount St Helens, Krakatoa, Chelyabinsk...) and I'm sure that those events threw molecules into space - but that is not what you are asking about.
So - initial pressure is double. Sound is adiabatic, so air got heated a little bit in the process. Mean free path for 0.3 nm molecules at 1 atmosphere is about $10^{-7} m$ . With speed of sound about 330 m/s (which is lower than the mean velocity of individual molecules), you can see that you have about $3 \cdot 10^9$ molecular collisions per second - enough to maintain the Boltzmann equilibrium over any reasonable scale (distance).
So we can change the question: at the limit of the atmosphere, how much do we heat the air due to our loud sound? Because that heating changes the velocity distribution - and that tells us whether additional molecules will escape.
Let's set the limit of "space" at 100 km altitude. This is an arbitrary number - see for example http://www.slate.com/articles/news_and_politics/explainer/2004/09/where_does_space_begin.html . Escape velocity is about 11 km/s. But key: at this altitude, we are in the thermosphere, so the temperature of the gas can be 1000 C or more. Also key: the molecules are highly ionized, so they experience magnetic forces from the earth's magnetic field. I will ignore those, but they probably make the rest of this calculation irrelevant.
From the Maxwell-Boltzmann distribution we can compute the fraction of molecules at a temperature of 1000 C that have escape velocity:
$$f(v) = \sqrt{\left(\frac{m}{2\pi kT}\right)^3}4\pi v^2 e^{-\frac{mv^2}{2kT}}$$
With a most probable velocity of 
$$v_p = \sqrt{\frac{2RT}{M}} = \sqrt{\frac{2\cdot 8.3 \cdot 1300}{0.03}} = 830 m/s$$
at an escape velocity of 11 km/s, we are about 12 standard deviations from the mean. The cumulative distribution is about $10^{-33}$ - or one molecule per $10^{10}\text{ mol}$. Note this number would be MUCH higher for helium (lower mass) but we were talking about "air" - nitrogen and oxygen. I used a mean mass of 30 which is an "average" value (for 20% oxygen and 80% nitrogen - I know that is not really the mix at this altitude... we're well into estimating territory here)
Now we want to see how much hotter the air gets up there. Let's assume our parabolic reflector had a diameter of 3 meters, and we're transmitting a 1 kHz sound. That gives us a wavelength around 33 cm, and a beam divergence of about 5 degrees (0.1 radians). At 100 km distance, that puts the area we are hitting at a 10 km diameter, or an increase in area (from the 3 meter reflector) of about $4\cdot 10^{10}$. The pressure level is reduced correspondingly, and the temperature change... well, suffice it to say you won't be able to measure it.
Which gets me back to the initial paragraph. Very few molecules hit escape velocity, and the statistical flucutation is such that you will never be able to tell if your "shout" moved even one more molecule into space.
