Hawking Goes Anthropic

Stephen Hawking and Thomas Hertog have a new paper out, called Populating the Landscape: A Top Down Approach. It contains his version of the anthropic landscape idea, based on his “no-boundary” idea of quantum cosmology (sometimes also referred to as the “Hartle-Hawking wavefunction”), and he refers to it as “top-down cosmology”.

Here’s part of the summary:

In a top down approach one computes amplitudes for alternative histories of the universe with final boundary conditions only. The boundary conditions act as late time constraints on the alternatives and select the subclass of histories that contribute to the amplitude of interest. This enables one to test the proposal, by searching among the conditional probabilities for predictions of future observations with probabilities near one. In top down cosmology the histories of the universe thus depend on the precise question asked, i.e. on the set of constraints that one imposes…

The top down approach we have described leads to a profoundly different view of cosmology, and the relation between cause and effect. Top down cosmology is a framework in which one essentially traces the histories backwards, from a spacelike surface at the present time. The no boundary histories of the universe thus depend on what is being observed, contrary to the usual idea that the universe has a unique, observer independent history. In some sense no boundary initial conditions represent a sum over all possible initial states. This is in sharp contrast with the bottom-up approach, where one assumes there is a single history with a well defined starting point and evolution.


We have also discussed the anthropic principle. This can be implemented in top down cosmology, through the specification of final boundary conditions that select histories where life emerges. Anthropic reasoning within the top down approach is reasonably well-defined, and useful to the extent that it provides a qualitative understanding for the origin of certain late time conditions that one finds are needed in top down cosmology.

I haven’t completely understood this yet, especially the author’s claims that they can use these ideas to say something about the shape of primordial fluctuation spectra, but the whole thing doesn’t sound obviously any more promising than other approaches to the anthropic landscape. Here, as usual, I agree with Lubos’s comments on this. But despite this, he seems to have removed me from his blogroll. My feelings are hurt.

This entry was posted in Uncategorized. Bookmark the permalink.

20 Responses to Hawking Goes Anthropic

  1. Kea says:

    This is great! I must go and read it. On the face of what is quoted here, it sounds much more like a topos sort of landscape than a Stringy one. 8)

  2. andy.s says:

    Re: Lubos Delinking.

    Maybe it’s because you’ve been agreeing with him so much lately. It’s probably getting on his nerves.

  3. mathjunkie says:

    andy.s has the point.

  4. Dave Bacon says:

    Hasn’t Hawking always used the anthropic principle? I can remember reading his Brief History back in high school and getting very upset with it for using said argument. And I’m pretty certain that during at least one talk I attended while and undergrad at Caltech (93-97) he gave a talk which basically was in support of the anthropic principle.

  5. A. nonymous says:

    I would guess that Lubos does not consider this blog to be ‘led by science’.

    Regarding Hawking’s paper….

    It would appear that all the anthropic stuff amount to no more than fitting data to observed values. There are three (large) spatial dimensions; use that as a constraint on the present state of the universe and say the word “anthropic” while doing so.

    When he makes contact with string theory, he suggests using the geometry of the internal space as part of the boundary conditions which describe the current state of the universe. Then he suggests that only a few metastable landscape vacua will emerge from his procedure. But the problem is that we aren’t able to determine what the present geometry of the internal space is (by any conceivable experiment, as Peter likes to point out). If we were able to determine that, we would know which part of the landscape we were in already. I may be missing Hawking’s point here, though.

    He says that the largest contribution to the history of the universe comes from configurations in which the universe popped into existence by quantum means – is this a result which has been proved in Euclidean quantum gravity, or not proved at all, or proved more generally? It sounds a little suspicious when somebody says that the present state of the universe is independent of an initial state because the universe with the prescribed initial state will undergo quantum disappearance and be quantumly replaced by a new one.

    Any predictions which will come from his proposal would seem to need to incorporate more constraints than merely the constraints that come from observed values, anthropic or not. It sounds superficially like his “no boundary” proposal provides the additional information that could lead to predictions, but then for any application of it, he puts his favorite instanton at the beginning of the universe and draws conclusions from that. How do we know that there isn’t an uncountable number of possible instantons that can match up expanding Lorentzian space with no boundary?

  6. Kea says:


    Yes. The anthropic discussion isn’t too alarming here, because the authors point out that such reasoning is simply one kind of constraint.

    “Top down cosmology is also more general than anthropic reasoning, because there is a wider range of selection effects that can be quantitatively taken into account.”

  7. Haelfix says:

    Hawking has always been a strong anthropic principle advocate, long before he even was interested in String theory. He used to talk about the unsolvability of the fine tuning problems in Inflation theory during astrophysics conferences. I always thought his point of view was needlessly dark and depressing and premature to the extreme.

    Anyway, im not a big fan of the no boundary proposal either (a rare topic in QG I know a little about). Wave functions of the universe and things like that are so plagued by conceptual difficulties (like even trying to define what hilbert space you are talking about), that they are rarely taken to be more than gimmicks by people in the field. Its an old story really, and goes back to the eternal problems of defining time in quantum gravity (eg the WDW eqn), and when to take euclidean path integrals seriously (in the HH approach, the integral is not bounded from below) . Not to mention its kinda problematic to leave so many gravitational degrees of freedom un quantisized.

    But in the ill understood field of qg cosmology, I guess crude treatments like this are as good as you’re likely to get at this stage of the game, at least until we get a better microscopic treatment from somewhere else (eg perhaps string theory).

  8. OK if the present is determining a history does that mean the past’s present lives through a history via the future? I may need to rephrase this my head hurts.

  9. R.R. Tucci says:

    I think I finally understand what string theorists mean by the Anthropic Principle. What Hawking et al are doing is a pretty standard Bayesian calculation. I’m not sure that this Bayesian technique will be fruitful in the case of String Theory. I don’t even know if it makes sense philosophically and logically to apply this technique to String Theory, but the technique itself is firm and solid, and well known outside of String Theory. This technique can be used within the realm of classical probability to select those classical histories that are more promising, or it can used within quantum theory to select quantum histories.
    The “final boundary conditions” correspond to the final nodes of a Bayesian Network, and what they are trying to do is to update the probabilities of the non-final nodes of the network based on their observation of the final nodes. As often happens in Physics, these people are reinventing the wheel. It would help if they were less proud and used the language and techniques that have already been developed by statisticians to do this, instead of pretending that there are no precedents. (For those of you who are not familiar with Bayesian Networks, there are a zillion reviews on the web. A Google search for “Bayesian Networks” yields about half a million hits.)

  10. Steve says:

    I strongly agree with the above comments about rediscovering the Bayesian approach.

    Another place to look for background information is the comment that I added to Lubos’ posting, where I summarised what I think are the relevant parts of the Bayesian approach.

  11. Kea says:

    See also a paper by Marlow

  12. George says:

    Lubos removes views he does not like. He is a right winger alright and it comes through in his science as well. It is becoming more and more common for the right to supress debate and ideas they disagree with. It is so revealing in light of the history of right wing repressive societies… now unfortunately we are living in one too. I will listen in on you, but trust me, I am only after the bad guys, and I define the who and what is bad and… so don’t worry…

  13. Wolfgang says:


    the link is back, but you need to scroll down a bit 😎

  14. Eli Rabett says:

    Technically Motl is not an apparatchik, but a member of the nomenklatura

  15. mathjunkie says:

    Yes, the link of Peter’s weblog appears on Lubos weblog again. They become friends again.

  16. Chris Oakley says:


    It could just be that the former StB operative that Lubos hired to rub Peter out has been detained by US immigration. Or he went to the wrong place. Check the newspapers to see if a maths professor called New York at Woit University in Columbia has recently been wasted.

  17. ObsessiveMathsFreak says:

    Slightly offtopic, but I am pleasantly surprised that such a well known and also “venerable” physicist such as Hawking would publish first on arXiv.org instead on a journal. It’s surprising how much times have changed.

  18. Steve Myers says:

    After quick run through of paper — yes, Bayesian. But don’t see much in it. A personal note: back, about 20 years ago, when I wasted my time thinking about cosmology, I wrote Hawking asking about an idea I had that involved infinite neg. E at big bang, plus other dumb stuff. He replied that yes, there at that singularity was the problem & that’s what he was working on. It was a pleasant note & he thanked me for the book I’d sent.

  19. mathjunkie says:


    I’ll check the newspapers!

  20. Aaron Bergman says:

    I strongly agree with the above comments about rediscovering the Bayesian approach.

    Nobody I know who thinks about this sort of thing is unaware of Bayesian statistics.

Comments are closed.