Several friends have been asking me to comment on a recent article from Wired Magazine titled, “The Genius Neuroscientist Who May Hold the Secret to True AI.” The article is about Karl Friston’s “Free energy principle” which is essentially that the purpose of life is to minimize the free energy — defined qualitatively as the difference between your expectations and your sensory inputs. The secret, according to the article, is applying thermodynamic principles to intelligence. For any of you following these posts that comes as no surprise. The timing of this article is convenient as I’ve been waiting for awhile now to write what the social thermodynamic laws say about efficiencies. So here goes…
Thermodynamic “Free Energy”
There are many energies we utilize in thermodynamics: potential, kinetic, internal, Gibbs, Helmholtz, Landau, etc. Several of these energies are often described as “free” which may be one of the greater points of confusion in all of thermodynamics. One of the first lessons Richard Jacobsen taught me is a graduate student was how silly this “free” word really was. This is not “free” as in it doesn’t cost anything, “free” here denotes that the energy must be defined relative to a reference state that is “free” to set at an arbitrary value as there is no way of making an absolute measurement of the particular energy form in question. The variability of reference state is a common problem that plagues folks learning thermo for the first time — they’ll often mix property values from different sources not realizing that the reference points were changed. Although the reference points can change, one lesson I hammer home to my thermo students is that the change in a “free energy” property must be identical for the same process, regardless of reference state; i.e. it will always take the same amount of energy to boil a cup of pure water from liquid to vapor, regardless of the reference state. Since you have to use a reference state to calculate any of these energies anyways, why further obfuscate the problem with the use of “free”? No wonder everybody has a hard time with thermo.
Some Efficient Comparisons
What Friston is probably looking for is something akin to a Thermodynamic Efficiency. Thermodynamic Efficiencies were created as useful comparisons between energetic processes. A thermodynamic efficiency exists for each of the thermodynamic balances (a.k.a. thermodynamic laws):
- First Law Efficiency: What you want divided by what you paid to get it. Typically W_out/Q_in for a heat engine. Q_out/W_in for a refrigerator, etc.
- Second Law Efficiency: What you got divided by what you could’ve. Typically the W_actual/W_ideal for a heat engine. A second law efficiency can also be obtained by multiplying the first law efficiency by the Carnot efficiency ( the ideal or the limiting efficiency defined by the temperatures of the thermal reservoirs) for a process.
Friston’s challenge of resolving the difference between what we actually see and what we expect to see can be explained by the Second Law Efficiency transferred into the framework of Social Thermodynamics (the body of the book I’m writing on this is down the page here). The perfect or ideal case allows no entropy generation (remember I assert that entropy is empathy in social space). This requires that you already know everything about the process and no further empathy generation can occur. This is also the case in which you can maximize output, because you know everything. Reality though is imperfect. We never know everything and as such what we actually get is far from ideal. Hence empathy generation occurs and we learn a little bit more about the process each time.
First law efficiency is all about the energies — you put a certain amount in and you expect a certain amount back. You never get the same amount or more back, and some processes have a much lower return than others (Facebook anyone?). You don’t know why until you look at the second law efficiency and realize that the process was inherently limited to a low return due to entropy/empathy — you couldn’t have gotten much back. Herein lies an important take away.
Efficiency really is the key to Happiness
Happiness you elusive emotion… several recent self-help authors and researchers have taken on the challenge of happiness. Happiness can be thought of as a first law efficiency in social space — we pay a certain amount in and we want and expect a certain amount back. We often think about what we put into a process as what we have control over, but we really should also think about managing our expectations. In many processes the ideal/limiting efficiency is so low that our hopes are likely to be dashed unless we consider these limits. It may not be possible to realize the return we seek given the social mechanism we’re working through.
But what about luck? Can’t we be suddenly surprised by winning the lottery/raffle? Absolutely. It’s a statistical process. All of the players in the system have a probability distribution. Given enough trials though and the thermodynamic limits/laws hold. We can get lucky, but in the universe there’s no such thing as a “free energy” or a free lunch. Same goes for social space.
(Note: this post is one chapter of what could become a book someday. The other chapters can be found here: https://hydrogen.wsu.edu/dr-jacob-leachman/ )