NATHAN MYHRVOLD

I found the email debate between Smolin and Susskind to be quite interesting. Unfortunately, it mixes several issues. The Anthropic Principle (AP) gets mixed up with their other agendas. Smolin advocates his CNS, and less explicitly loop quantum gravity. Susskind is an advocate of eternal inflation and string theory. These biases are completely natural, but in the process the purported question of the value of the AP gets somewhat lost in the shuffle. I would have liked more discussion of the AP directly.

Smolin's view is that the AP should be discarded on the basis of being incapable of being falsified. In addition, he claims that potential support for the AP is because it usually applied in the context of other more valid observations, getting a free ride so to speak.

Susskind argues that the AP has some value, but seems to stop short of saying just how much. He uses Weinberg's discussion of the cosmological constant as an example, but it begs the question of how much predictive value does the AP have?

Put another way, Smolin says it has zero value, Susskind holds that it has some positive value. Why not ask how much it has?

Given that we exist, it is not surprising that the universe has physical properties that are consistent with our existence. Consistency of this sort is to be expected. It is surely true that physical theories (or parameters in a landscape of theories) that are inconsistent with our existence would have a big problem — sort of like the old Ricky Ricardo line "Lucy, you got some 'splainin to do!". There is nothing special there — as physicists we require any theory to have predictions consistent with experiment and observation.

The question of the AP hinges on whether one can invert this proposition and say because we (intelligent life) exist, that therefore puts a constraint on physical theories (or parameters to physical theories), at least for our neck of the woods.

I'll just use "theories" as a shorthand for either a choice among a landscape of theories, or a set of parameters to a physical theory — all within a given pocket universe. At the level of this discussion they serve the same purpose.

Making this inversion generally means you slide along the following slippery slope starting with the most skeptical and progressing onward:

1. The skeptics answer is that you cannot invert the proposition. Consistency is nice, but it only works in one direction. There at least three common sub points:

(a) We have no experience of pocket universes except our own — in which we manifestly do exist. Our cosmological physics is based on one data point — the pocket universe we live in — and thus we cannot generalize with any confidence.

(b) Our existence is only one of millions of observational facts about the universe — why single out intelligent life as being special? It is not necessary because many of these other observations will do.

(c) Indeed we have reason not to single life out. We have only the weakest possible understanding of the causal chain of events between fundamental physics at the cosmological level and the evolution of intelligent life. So, while our physics is based on one data point, the situation with our biology is even worse. We have only carbon based life, only on one planet, with only one intelligent species. Until and unless we understand the chain of biology better, with more examples we have climbed out on a very precarious limb.

2. The first step toward the AP is to argue that there is some landscape or ensemble of possible physical theories in a set of pocket universes. Many aspects of physical theory make this seem like a sensible question to ask, because no intrinsic mechanism sets the parameters or properties of the theories. If there is no hard reason to prefer one theory, we naturally can ask about probability distribution of those theories.

(a) Assuming that there is a range of theories consistent with life, we can try to estimate something about the probability distribution based on our one data point. This requires making a generalization from that one data point — usually by a further assumption that we are a "typical" observer in some probabilistic sense. The "mediocrity principle" is one example, but there are many other examples of similar reasoning — i.e. the so called doomsday paradox, and related paradoxes in probability theory. The details of the assumption are notoriously critical. Given the assumption about how to generalize from the one data point, we can then infer something about the distribution of physical theories. This is at its essence a derived observation, based on an assumption — not a predictive theory.

(b) A different direction is to postulate further that the probability distribution is biased in favor of the range of theories that is consistent with intelligent life. This is usually called the Weak Anthropic Principle or WAP. It assumes that there is an intrinsic physical process, of unknown origin, that makes a pocket universe more likely. Note that it is different than 2(a) above. 2(a) says that one can derive observational bounds on theories based on our existence (plus an assumption). 2(b) says that there is a real physical effect, without justifying why.

3. The third step is to make the leap to causality. This comes in a couple stages.

(a) Intelligent life is a requirement for physical theories. Pocket universes are constrained to have physical theories that inevitably lead to intelligent life. This is usually called the Strong Anthropic Principle or SAP.

(b) Intelligence and information processing must come into existence in the universe, and once it exists will persist. This is called the Final Anthropic Principle.

Undoubtedly this summary of the various stages is not complete, and may enrage a fan of the AP, but it covers much of the ground.

I think that Smolin would say that his position is best covered by 1(a) — since we have only one data point, how can we falsify, and 1(b) — most seemingly valid invocations of AP rely on other data. Smolin does not seem to argue 1(c) in these postings, although it is the most common criticism of AP in other circles.

Smolin instead argues that his CNS is a concrete suggestion of a physical process that would bias the distribution of physical theories — but not because of life, but rather because it stimulates black hole formation. Coincidentally, this is similar to life.

Susskind does not really say where he is on this scale. He finds the AP useful. He seems to be positive about Vilenkin's mediocrity principle. This is a good example of 2(a). However, he does not go further to say whether he is in favor of 2(b), or the stronger forms in 3. I would be curious as to what he says about them.

In particular, what is the explanatory value of 2(a) ? The Garriga , Tanaka and Vilenkin paper abstract ends with the line "We also argue that observers should not be surprised to find themselves living at the time when curvature is about to dominate". OK, but that begs the question of what value there is in not being surprised. Consistency is not that surprising, and thus has little predictive value.

Technically speaking, the issue here is whether the generalizing assumption required in 2(a) — such as the mediocrity principle — leads to any additional information about the physics. 2(a) is about deriving something from an observation.

Conversely, if a 2(a) derivation did yield a surprise, then it might impact the physics. Or it might not. If there was a surprise, the focus then would shift on questioning the generalizing assumption. Critics of the doomsday paradox, and similar probabilistic paradoxes will point out that in most cases unusual results stem from subtle problems with "reasonable" generalizing assumptions.

Now, one could imagine that, for some other reason, the density of states in the landscape of possible string theories is sparse in the region consistent with our existence. If so, then even a small amount of information about the distribution might yield considerable information. So, for example if there were only a small number of viable string theories — in some discrete spectrum, then it may not take much of an observation to allow us to decide between them. However, this seems to be unlikely based on current knowledge.

2(b) would seem to be Susskind's best hope for a modestly useful AP as a physical theory. Weinberg's argument, which he references, is a good example of it. However, the problem here is that 2(b) is not really a principle or an argument — it is an ansatz — a hunch that needs to be confirmed.

However, one could equally ask (and Smolin almost does, but not quite) what is the difference between 2(a) or 2(b) with some other fact about the observable universe substituted for "intelligent life"?

Why is intelligent life so important? We are calling this the Anthropic principle not the "universe as we see it" principle.

Rather than to assert that AP has zero value, why not ask this — does it have more value to discuss intelligent life, than to discuss say the fluctuations in the microwave background, the distribution of matter in galaxies?

In actual fact, the cosmological physics that impact this debate are either gross scale things like condensation of matter into galaxies, expansion of the universe etc., or very fine scale things like the properties of carbon chemistry that depend on the fine structure constant and similar quantum parameters. None of these one is very directly tied to intelligent life. So why do we go there?

This to me is the crux of the issue. Why is intelligent life special?

If it is special, in the sense of having more explanatory power with respect to cosmological physics, then the name "anthropic" is justified and so is the debate. Conversely, if intelligent life is not special with respect to cosmology, then we should shut up about "anthropic" aspect.

Note that this is orthogonal to the point discussed by Susskind and Smolin — regardless of whether the AP can be falsified, tell me why I should care.

By 1(c) above one can doubt the special-ness of intelligent life on many grounds. We have no real evidence of life outside earth, and no evidence of intelligent life here other than ourselves. These are testable.

One can easily imagine both of these being empirically tested in our lifetimes. Perhaps tomorrow Spirit or Opportunity will run over a Martian life form — or if not there then perhaps a later NASA mission to Europa or elsewhere. Computer scientists may someday present us with a second kind of intelligence, or the SETI folks may find a communiqué from intelligent life elsewhere. So, the situation is far from hopeless, but at the moment we know next to nothing about them. Until the happy day when we do know something about them, invoking either life, or intelligent life as an explanation seems loony.
Other observations surely constrain our choices of fundamental physical theory far more. Satellite data on the fluctuations in the microwave background have revolutionized cosmology, weeding out many previously "un-testable" theories.

Unless there is powerful additional explanatory value in invoking intelligent life, why go there?

When one gets to the extremes of 3(a) and (b), or even 2(b) when posed in strongly terms of intelligent life, it seems to be little more than mysticism.

I wonder if Smolin and Susskind would agree?

Nathan