The first one is interesting, even if he puts a negative spin on it. We’ve encountered it in an academic setting in terms of people no longer having to have a comprehensive knowledge of the literature for their prelim exams.

The second one is just flat wrong and demonstrates a pretty poor understanding of the scientific method and complexity theory. There’s a lot of philosophy of science written about modeling theory that bears on the questions he purports to raise that he ignores. I’m also pretty sure that Box quote doesn’t mean what he thinks it means in the context that Box wrote it. Prediction and explanatory power are two distinct endeavors, but I don’t think he’s made a useful case for or against either. As an example, I’m sure the entire field of meteorology will be heartened to hear that massive computing power is all it takes to completely understand complex systems and enable us to make accurate predictions.

I love how people think data just self assembles. Even though the structure it eventually takes on is unknown at the start, it doesn’t mean we can find a correlation without (with imperfect understanding) drawing some relations between data so as to give it structure and meaning. That’s modeling, but on a lower level. Some models are better represented and explored on machines with power beyond to human mind to correlate it, but it is still only a model.
Imaginative theory as the driving force behind the creation and exploration of data, however, I can see as legitimately declining (I’ll stop short of unneccesary.) Look at the field of philosophy… the explanatory sectors have all receded into other disciplines that are data driven. Psychology is a great example of the transition from being driven by analogy and theory to data and correlation. Our modeling techniques have changed as more data and better methods of correlating it have become available.
The idea that the systems doing the modeling might start to auto-correlate data without human intervention, eventually even learning how to design (and implementing) better methods of modeling is, I think, the real germ of fascination behind this idea. Recently we’ve taken to calling that event the technological singularity, but the idea of the demiurge attaining material form is nothing new. But the data, no matter how well it’s is modeled, is never the matter it represents. Yes, theories of individual men will continue to be surpassed by the explanatory power of systems assembled out of the combined modeling power of many men. And maybe one day one of our systems will become better at modeling than we are and develop itself beyond our capacity, lifting the burden of modeling from our hands. But we aren’t there yet. When we are, we’ll see that even such a system doesn’t have the modeling capacity to come up with a truly comprehensive Grand Unified Theory.
I happen to have also just read this, which I thought addressed the subject of modeling nicely. It’s from the chapter on wisdom in the Bodhicharyavatara:
“When ordinary folk perceive phenomena,
They look on them as real and not illusory.
This, then, is the subject of debate
Where ordinary and meditators differ.”
“Forms and so forth, which we sense directly,
Exist by general acclaim,though logic disallows them.
They’re false, deceiving, like polluted substances
Regarded in the common view as clean.”
Our perceptions are assembled and handled on systems specialized to the task, and which often employ shortcuts that misrepresent (often in a useful way) what they inform us of. I think his article is just another peon to the dream of some purer platform for modeling on which to draw our explanations and predictions. However, once we build Deep Thought 2, we’ll still need to assemble useful questions, and we aren’t there yet. I think that’s the meat of Timmer’s response. Science is about figuring our what questions are useful, and how to get answers; building models may be a prime requisite but it isn’t the purpose.
Oh, and Sarah says “Hi!”

Huh, apparently this stylesheet doesn’t give comment links any visual cues. I had to add ‘style=”text-decoration:underline”‘ to that link–I’ll have to fix that…