One version of the cosmological argument relies, among other assumptions, on the following:
Every contingent being has a cause.
Leaving aside the necessity of defining what we mean by "cause", and of explaining how such an inductive statement could reasonably apply to a unique object such as the universe, I want to focus in this post on the concept of contingency.
How do we know that anything is contingent? How do we know that anything could have turned out to be different?
Like free will, contingency is useful, and even essential to our human experience. The idea that we can make choices, and that these choices may result in different outcomes is how we survived and evolved. Without it, the notion of probability seems to become absurd, and human experience seems to lose its meaning. This is in itself an interesting discussion, but what we would like or perceive reality to be is not what determines it.
As far as we know, and given the laws of physics as we know them, if we know the state of a system at a given time, we can in principle derive its state at all times. There is no room for contingency, except for those so-called initial conditions. I'm saying "so-called" because those are only initial if one extrapolates to the future: if one extrapolates to the past, those conditions are final rather than initial. Even in quantum mechanics, the wave function is deterministic, and the wave function might be all there is.
In light of this, we are left to wonder how contingent those initial conditions and the laws of physics can be. Many scientists, until the end of the 20th century, thought we would eventually be able to understand why the physical constants have the values they have. Today, more and more are agreeing with Leonard Susskind that we might never know because there is nothing to know.
Does this mean that science now points to a contingent universe after all? Not at all. In fact, it's quite the reverse: the idea of a multiverse, which comes out quite naturally of string theory and inflation, indicates that the physical constants are only locally constant, and that there is a landscape of universes where all the possible values can be taken. Locally, within a universe, all that can be seen looks contingent or arbitrary, but the whole landscape possibly contains all that is possible. This is similar to the idea that the solutions to a symmetrical problem may not be symmetrical, but the set of solutions always is.
And here is the interesting idea: if all that is possible by virtue of being consistent exists, contingency disappears. This constitutes a unification between the sort of necessary existence that mathematical entities enjoy (the only kind of necessary existence that we know for sure is valid, by the way), and empirical, "contingent" existence.
Of course, this all doesn't seem falsifiable, and it probably isn't. But here is my point: the mere possibility that contingency might not actually exist is enough to kill the premise of a cosmological argument for god that "every contingent being has a cause". In order to use that premise, you first have to prove that there are contingent beings, and that may end up being a lot more difficult than one may have thought at first. And that is the problem with most of metaphysics: it holds as self-evident what in subtle ways isn't.