Stephen Hawking and Thomas Hertog recently published an article that should be of interest to many of us. When the normal “bottom-up approach” is applied to cosmology, one ends up with a finely tuned universe as we all know. Hawking has apparently been busy trying to find a way around that “problem” with a “top down approach”.

http://link.aps.org/abstract/PRD/v73/e123527

http://arxiv.org/abs/hep-th/0602091

In fact if one does adopt a bottom-up approach to cosmology, one is immediately led to an essentially classical framework, in which one loses all ability to explain cosmologyÃ¢â‚¬â„¢s central question – why our universe is the way it is. In particular a bottom-up approach to cosmology either requires one to postulate an initial state of the universe that is carefully fine-tuned [10] – as if prescribed by an outside agency – or it requires one to invoke the notion of eternal inflation [11], which prevents one from predicting what a typical observer would see.

Here we put forward a different approach to cosmology in the string landscape, based not on the classical idea of a single history for the universe but on the quantum sum over histories [12]. We argue that the quantum origin of the universe naturally leads to a framework for cosmology where amplitudes for alternative histories of the universe are computed with boundary conditions at late times only. We thus envision a set of alternative universes in the landscape, with amplitudes given by the no boundary path integral [13].

The measure on the landscape provided by no boundary initial conditions allows one to derive predictions for observations. This is done by evaluating probabilities for alternative histories that obey a set of constraints at late times. The constraints provide information that is supplementary to the fundamental laws and act as a selection principle. In particular, they select the subclass of histories that contribute to the amplitude of interest. One then identifies alternatives within this subclass that have probabilities near one. These include, in particular, predictions of future observations. The framework we propose is thus more like a top down approach to cosmology, where the histories of the universe depend on the precise question asked.