First, increased computing power does not mean that you can accurately model all events; you can just look at a larger subset of the possible data. For instance, we still could not use Google’s computing power to model chemical reactions from quantum physics. What we can do is use the greater processing and data retrieval capabilities to look for low-probability discrepancies between data and theory.

Second, the scientific method is more than just a data-analysis method. It is a way of carefully answering questions with the right data. Simply having more data does not make for better answers. Indeed, more data, while having the potential to provide better estimates, also offers the potential to lead researchers off into the weeds.

Lastly, theory is a means to predict future events. More generally, science is tool to improve understanding. Simply being able to model past events given some data set does not provide greater understanding. This is, of course, part of the contention: if our data set is big enough, what need do we have of understanding? My experience is that theory leads to new questions, which in turn lead to a deeper and more accurate understanding. More data can provide better models, but not deeper understanding, and generally does not lead to new questions. It’s the unexpected–the differences between expectations driven by theory and models derived from data–that lead to advances in knowledge.