Recently philosophers of science have debated the legitimacy of attempts
to "explain the initial conditions" of the universe. Cosmologists frequently
invoke the apparent fine-tuning of the initial state required by the standard
cosmological models as the main motivation for new theories such as inflation.
Without new dynamics such as inflation, the argument goes, the universe as
we observe it would be incredibly improbable. Callender has emphasized
the similarity between such arguments and the traditional argument from design;
in both cases, there are well-known problems with making this appeal to probability
precise. I will argue that this difficulty reflects a general feature
of "fine-tuning problems": they are defined by a contrast between speculative
extrapolations of existing theory and current observational and experimental
results. The ability of a theory to solve a fine-tuning problem is
only a significant success if the extrapolation of existing theory proves
to be correct. Although these remarks apply to fine-tuning problems
quite generally, my talk will focus on early universe cosmology.