Wednesday, April 06, 2005

Anthropic world: Vilenkin

Alexander Vilenkin just spoke about
  • Probabilities in the landscape
Very nice work - and Vilenkin is a smart gentleman - except that I am not the only one who feels very discouraged after the talk. The colleagues are asking how far the boundaries of science can be pushed.

The goal: distributions

The goal of the talk, as I understood it, was to make the first steps to determine some probability distributions for various quantities using the hardcore anthropic reasoning. These distributions are the only numbers we should be trying to figure out, according to the anthropic reasoning. Unlike quantum mechanics where the distributions may be exactly checked by repeating the experiment many times, we only have one Universe to measure.

The first choice is to fix the parameters of the Standard Model and low-energy physics, and only vary the cosmological constant "Lambda" and perhaps "Q". That's of course unjustifiable, but let's not stop at this point. For our purposes, the quantity "Q" is defined as "delta rho / rho" which is the typical relative fluctuation of the temperature of the cosmic microwave background (CMB) and it is equal "10^-5" in our Universe. Its value depends on the energy scales and other parameters of the inflaton potential which is "very high-energy physics" that does not affect local life.

The probability distributions are written, in analogy with Drake's equation for the number of telephone contacts with the extraterrestrial aliens, as products of many quantities - the simplest example is
  • P (Lambda) = P_{prior} (Lambda) P_{formation} (Lambda)
where P_{prior} is an a priori distribution coming from the fundamental (string) theory (in the most optimistic picture, it could arise from the Hartle-Hawking state in some way; in less optimistic cases, it is given by "counting vacua" or another unjustifiable method) and it is completely unknown at present.

Density of galaxies

On the other hand, P_{formation} is taken to be proportional to the number of observers that we generate in a given Universe. Vilenkin argued that it can be taken to be a quantity proportional to the number of galaxies in the Universe. This particular quantity, the number of galaxies, is pretty much calculable, as many Vilenkin's simulations and graphs of the volume of various regions (the three regions are roughly speaking: quantum diffusion near the top of the potential; slow roll in the middle; thermalization near the bottom) in the inflating "pocket Universe" showed, but it is much more controversial whether this is the right number that should become a factor in the probability distribution.

As you see, Vilenkin considers one galaxy to be a unit of life. P_{formation} should really measure the number of observers that will exist (or be born) in a given Universe. The total amount of intelligent life, so to say. The precise definition of P_{formation} - which is the quantity affected by the details of the anthropic (non)rules - belongs to humanities, not to science, I think. For example, Nima kept on asking "what is the moment at which you measure the number of observers" and he suggested that the integrated number of observers in the whole spacetime should be relevant. Vilenkin disagreed and he only wanted to count the number of observers at one moment (still not sure which one, and I was not the only person confused by the rules how to choose the "right" slice). Nima's proposal seems to be the more plausible one to me but neither of them is justifiable scientifically.

Intermezzo: defining and predicting intelligent life

But imagine that you would like to create an argument of this sort that is more than just hand-waving and where some other parameters would be varied. You would have to decide who is an intelligent observer. For example, a Universe with different parameters could produce small planets with weaker gravity. This would probably affect the typical size of the animals, the size of their brains, and consequently (politically correct people, please, forgive me) their expected average intelligence. Should we consider or expect the "bugs" living as the Universe to be equally good intelligent observers as larger animals like us? This can change the distribution by 10 orders of magnitude, or parameterically by huge additional power laws.

Note that in order to count these things realistically, we really need to know virtually all factors in the real equation due to Drake. In other words, physics behind the Standard Model is now supposed to be less solid than the search for extraterrestrial civilizations (SETI) because we need all answers from SETI and some additional answers to figure out what the new physics beyond the Standard Model should look like.

Moreover, should we just count the total number of observers that live in the Universe, or should we also take their lifetime into account? The number of words they can say per life? If the lifetime of humans were 1000 years, like in the Bible, should they increase the probability of a given Universe by an order of magnitude? Were the long-lived heroes of the Old Testament ten times as good people as we are? If the typical velocity of the humans in some civilization is 1000 times faster than in ours and if these people have 1000 times richer lives, does it increase the probability of their Universe and the quantity P_{formation} by three orders of magnitude? Shouldn't we count the number of cells as opposed to the number of people which would add 9 orders of magnitude for our life? Or should we calculate the number of villages or nations as the independent units of intelligent life which could subtract 9 orders of magnitude?

Fortunately, I am not the only one who is convinced that these questions will never be resolved and cannot be resolved scientifically. There are hundreds of similar questions, we can answer them in many different ways that reflect our prejudices. All different answers lead to different outcomes, and by "properly" adjusting all of our prejudices, we can obtain almost any answers we want. No doubt, there will be people who will argue that only one choice (theirs) is correct - but they will never be able to show why their choice is better than others. Moreover, there are many ways how to tune the prejudices to obtain one desired set of answers. The prejudices are untestable. They are not a subject of scientific verification. There is no natural "one-dimensional" scalar function of the properties of animals that could tell us how much they contribute to the probability distribution.

Input vs. output

Someone can try to intimidate other people by the conjectured large number of vacua in string theory. But the number of different versions of the anthropic principle - the number of formulae through which we determine "the amount of intelligent life in the Universe" times the number of ways how we count the "number/volume of Universes of certain kind" is much higher even than the worst existing finite estimates for the number of vacua. In other words, we can always fine-tune our anthropic "principle" in such a way that virtually any kind of a Universe can be chosen as the preferred one. Is that science? Even if there exists a Universe that cannot be obtained as the most likely one by adjusting the anthropic (non)laws, we can't really eliminate it because we don't know whether our fantasy about the possible anthropic (non)laws was complete.

Weinberg's success (?) story

If I return to slightly more technical topics, one of them was the distribution of the cosmological constant. You know, Steven Weinberg showed that it should be between -1 and +100 times the currently observed value, roughly speaking, for the galaxies to be able to form. If the cosmological constant were too negative, the Universe would approach the Big Crunch too early; if it were too large and positive, it would expand and dilute before the matter could clump into galaxies. This Weinberg's calculation is a source of pride and inspiration for the people who have switched to the anthropic mode because it was an actual "prediction" of a positive cosmological constant. Weinberg's predicted value differs from the right one by two orders of magnitude - which is 30 times better than the naturally predicted value of "Lambda" after SUSY breaking which differs from the observed one by 60 orders of magnitude. The price for this factor of 30 in the exponent is that the methods of physics are supposed to be replaced by methods of humanities.

Note that Weinberg's argument implies a rough solution to the coincidence problem ("why now") - the problem why do we live in the era in which the cosmological constant is comparable to the density of regular matter. Weinberg's answer is that the probability distribution for Lambda is naturally peaked near the density of matter at the time when galaxies are formed (which is comparable to the era in which we live), as his calculation shows. This simply comes from the requirement that galaxies can still form (the density is sufficient) when the cosmological constant starts to dominate - this is roughly where his allowed interval ends.

Can we make Weinberg's computation more accurate? Depending on your assumptions, you can achieve any distribution you want. The total distribution may be nearly constant near "Lambda=0", but it can also behave as "exp(C/Lambda)" which hugely prefers tiny positive values of Lambda. Moreover, the constant "C" is very important to get some details. Note that the "constant distribution" has absolutely no invariant meaning because distributions depend on the coordinates that we use to parameterize the parameter space.

Trying to falsify the anthropic framework

Another thing. Banks, Dine, and Gorbatov argued that if we allow not only "Lambda" to vary but also "Q", then we obtain wrong predictions for "Lambda" and "Q" because the Universe with correlated higher values of "Lambda" and higher values of "Q" will be preferred. The matter density will be less uniform if "Q" is larger, and the density of galaxies increases. This will allow us to increase "Lambda" because the galaxies won't dilute so quickly anyway. In my opinion, this is the right approach to these questions - try to falsify them, instead of adapting them.

Vilenkin has made a trick that led him, once again, to a factorized probability distribution for "Lambda" and "Q", so that any change for "Q" is irrelevant for the statistical predictions of "Lambda". I forgot what the trick was, and I am not sure whether I should be sad because of it. A lot of extra discussion focused on the bounds on "Q". Of course that some "mainstream" bounds tell you that "Q" is not that far from the observed value "10^-5". Concretely, it should be true, they say, that
  • 10^-6 is smaller than Q is smaller than 10^-4.
Nima asked what is the upper bound for "Q" where the galaxies don't form because matter collapses into black holes before they can start to be formed. No clear answer appeared, but people guessed "10^-2". More generally, I find the inequality like the displayed inequality above completely useless. We already know "Q" from the observations. It's "10^-5". And it's clear that if we varied the values of the parameters of our Universe too much, we would obtain a very different Universe. It's a tautology. The only remaining question is how much different from ours we allow the other Universes to be so that we admit that they can contain enough "intelligent life", and how small probability for the formation of life in that Universe we consider realistic. This is almost the same type of question like the question "how much can we tollerate racism" or, more drastically, "how many angels there are on the tip of the needle".

Googol vs. googolplex

Regardless of the value of "Lambda", it's always possible that a Universe appears as a gigantic fluctuation. The probability of such fluctuations are not the inverse googol (10^{-100}) but rather the inverse googolplex (10^{-10^{100}}). Obviously, as Nima likes to say, in the anthropic framework we need to allow the unlikely events whose probabilities are of order the inverse googol because it is necessary for picking the Universe with the unnaturally tiny Lambda, but we should not allow phenomena whose probabilities are the inverse googolplex because this would also allow us to say that the Universe (including the fossils) was created 6,000 years ago, in agreement with Genesis.

The only good reason why they want to allow googol-like probabilities but not googolplex-like probabilities is that they're not Christians, and they only want to support the weak form of the religion (the anthropic principle) but not the strong version (literal belief in the Bible). You may think that I am joking, but I am not. The choice of these bounds of acceptability, once again, influences the conclusions dramatically. Moreover, it is not quite clear how the probabilities should be calculated. We can always imagine that there are googolplex of different galaxies in a Big Universe, and in one of them, life can arise as a gigantic fluctuation. In fact, I believe that one could arrange various fluctuations whose probability would be much closer to one (much greater) than the inverse googolplex.

Let me mention a very different example showing why it is important for the data to be stable with respect to perturbation of the "conventional" parameters. If the people in the 1970s looked at the temperature records from 1850, they could have seen a mild warming trend. If they looked at the temperature records starting from 1940s, which is what they actually liked to do, they could have deduced that a new ice age was getting started. Both conclusions were unjustified. If the conclusion depends on the choice of the year where you start to measure, or on the probability which you call the least acceptable probability, then your conclusion is unstable and it is scientifically irrelevant. Nothing better than guessing.


These ideas based on infinite possibilities and infinite Universe can never lead to convincing i.e. new quantitative results. The amount of bits that we insert as parameters - the answers to various "subtle" questions how the anthropic principle should work (plus, in reality, the assumptions about the behavior of the fundamental theory that we don't yet know fully) is hugely larger than the amount of (fuzzy) information that we can derive. A scientific theory should generate more predictions than the number of assumptions we insert into it.

The only way how similar things can work scientifically is that we understand the full theory at the fundamental scale well enough, including its unique Hartle-Hawking-like framework (which must be free of all assumptions and can only use the standard path integral of the theory, or an equivalent of it, to calculate any probabilities) to calculate the initial state of the Universe, and we take this initial state to calculate the probabilities of the subsequent evolution. Another possibility is just ignore the cosmological selection considerations altogether, and just continue to try to identify the correct "vacuum" by trying to match the properties of particle physics, leaving the early cosmological questions to the very end.

Finally, I am sure that various people who have a similar opinion about the anthropic thinking will use this admitted frustration as a weapon against string theory. Unfortunately, I must assure you that the expansion of the anthropic principle is a problem of the whole theoretical physics, not just string theory - and this talk was not a string theory talk after all. Incidentally, I was just explained by one of the authors of the entertaining article Supersplit Supersymmetry that it was making fun primarily of the anthropic approach, not supersymmetry.


  1. Before 2007, all fields of particle physics will be fuzzy, including string theory.
    People write papers hoping that some stuff will be useful after LHC comes out.
    Therefore, very , very few papers written today will "survive" after 2007.

  2. i'd quit if everyone is switching to economics :)

  3. I must admit, its rather discouraging to see such a great scientist as Vilenkin waste his time on stupid questions like how to properly slice the universe looking for intelligent life. I mean it sounds like something a child might be interested in, not a proffessional scientist.

    Anyway, I've slowly begun to hear more and more such talk from colleagues (brilliant ones at that), and it is rather sad for anyone who took to this field to discover natures secrets, rather than chitchat aimlessly about philosophy.

    I should have listened to my grandmother and done horoscopes instead!

  4. Lubos:

    Very nice summary. And congraduation on calculated the 6000 years age of universe according to genesis. I never thought about that :-)

    Obviously these kind of research is not going any where and is not science. It is not pushing the boundary of science. There is no boundary to push because you are already outside the boundary of science.

    There needs to be some paradigm change in physics. We already have had several paradigm changes. The Newtonian view of the world is one of ether, which is the first paradigm. The second paradigm is view the world as of particles. The third paradigm is view the world as one of fields, QED, QCD, QFT, you name it.

    Can we have a third description of the world, other than particles and fields? Particles are just quanta of fields. And fields are space distribution of particles. Both particles and fields need a spacetime background. But neither particle nor field describes the spacetime they sit in. The description of world is incomplete without a description of spacetime. That's why QM is incomplete.

    But on the other hand, the GR, which describes the spacetime, is unable to describe particles and fields either. That's why GR is incomplete, too.

    The naive attempt to merge QM and GR all failed so far, that is because the description of particles, fields and spacetime is simply not compatible. To quantize spacetime directly doesn't seem to work since how could you describe spacetime as particles, or as fields? You can't. Nor can you use the other approach, use the geometrical way of describing spacetime to describe particle and field.

    We need another paradigm change in physics. This time, replace the picture of particles, fields, spacetime, with the one of entropy, i.e., quantum information. Everything is information and carries information. You construct spacetime background from quantum information and GR naturally occurs, you construct particles and fields from quantum information, and it automatically leads to QM.

    It's a beautiful picture and GUITAR naturally leads to many amazing results that agree with the nature completely, you get the G, get the value of masses of elementary particles, figures out the radius of the universe, and CMB temperature, and a ton of things, all from just one adjustable parater which actually fixed, the alpha. You have to treat it seriously if that many result can be obtained from GUITAR which is based on just one very simple principle.

    It is certainly not numerology although you may think it that way, because it looks too good to be true, how could you obtain neutron mass to 10 decimal places accuracy?

    At least my theory can leads to solid calculation results that can be rigorously checked against observational data. It is certainly much better than guessing amongst the googolplexes of possibilities of vacuas.

    You must know that I could not have revealed all the details of such an important discovery on web pages, right? I could only show you some results that you can verify, and still have to hide some essential parts of my theory, until I have the whole thing, the whole math formulation worked out completely, and then I can publish everything at once and claim credits.


  5. Lubos said:
    "Concerning holography, I think, on the contrary, that the existence of a huge landscape of vacua contradicts the holographic principle because a region of compact space would have to remember in which vacuum it is, which amounts to some information, which according to Bekenstein's bounds implies that the region can't be smaller than a certain (large) value."

    Lubos, finally you are speaking of the right keywords, information, Bekenstein Bound, bounded region. See you are starting to pick up the idea that a certain amount of quantion information can be used to defined and construct regions of spacetime. I encourage you to go along that direction a little bit more and you will find the basic pillars supporting GUITAR theory, a quantum information based description of the universe.

    If Weinberg's two order's magnitude off is considered impressive, how about a value as large as 10^40, but off by just 1%?

    Bekenstein Bound needs modification. We all know the Lorentian invariant problem it has.
    But the basic idea is valid.


  6. quantoken, you are a fraud yourself.

  7. Quantoken, are you kidding me or yourself by "quantum information automatically leads to QM."