DrChinese said:
I might point out that entropy can increase from "now" in both time directions. Obviously most lab situations are special cases in which entropy is made to be unusually lower than the surroundings. If you sampled the entropy of a typical environment, in thermal equilibrium: wouldn't you expect it to be at a local minimum? Ie the number of states it could have evolved from and the number of states it can evolve towards are both greater than now? That would be the statistical view, I believe. In a film of that, I do not believe you could discern its direction as forward or backward in any way (in contrast to the usual idea of a film of a glass breaking being an example of the time direction being obvious).
Or alternately, think of decoherence: entanglement disperses and decreases as you go forward in time. Does that require a fundamental time asymmetric law to describe as well?
There is a time-symmetric model of the second law that applies to classical physics (and I assume that it can be extended to quantum mechanics, as well).
Imagine taking a human being--okay, for ethical reasons, let it be a guinea pig, instead--and putting it inside an impenetrable, eternal box. No energy or matter can go in or out. Now just wait--a billion years, a trillion years, 10^{100} years, however long it takes. After a while, the guinea pig will die, and decompose and will reach some kind of uninteresting equilibrium state, and its component atoms will remain in that state for an ungodly length of time. But there will always be a certain amount of random thermal motion of the atoms. Purely by chance, if you are willing to wait forever, the atoms will eventually arrange themselves to a configuration that is arbitrarily close to the original state of the guinea pig. In other words, the guinea pig will eventually come back to life, a reversal of entropy.
But over an enormous span of time, if you plot entropy as a function of time, what you will find is that:
- By far, the most likely configuration is the maximal possible entropy.
- Very rarely, the entropy dips down to a non-maximal value.
- In almost all such cases, the entropy returns quickly to a higher value.
The picture shows a typical plot of entropy vs time. Situations of type A are vastly more likely than situations of type B, which are vastly more likely than situations of type C, etc. So whatever the entropy is, if it's not the maximal value, then you are overwhelmingly likely to have higher entropy in the future, even though the graph is completely symmetric between past and future.
So the guinea pig, looking forward in such a universe can assume the second law of thermodynamics. He will likely age, die, and decompose just as the second law predicts.
What's weird about this thought experiment is that while the guinea pig can safely assume that he will be older and more decrepit in the future, he can't assume that he will be younger and in better health in the past [edit: was 'future']. In this model, the most likely
past for the guinea pig is one in which he is older than now. It's overwhelmingly likely that right now the guinea pig is youngest he has been for millennia and the youngest he will be for millenia to come.