systematic misunderstandings have _always_ infected our collective conclusions. culture can pretty much be defined as ’systematic misunderstandings’.

but as you pointed out, being able to apply a theory to make a testable prediction is much easier than fully appreciating the genesis of that theory. working backwards from a falsified prediction to find the failure in the nuts and bolts is likewise fairly approachable.

From a purely factual point of view, there’s not much in Intel’s process technology and design tools that’s not in the peer-reviewed scientific literature. Where Intel has displayed unmatched skill is (in Michael’s phrase) understanding the literature.

An important question for 21st century science, therefore, is “to what extent (if any) can Intel-style integrated scientific understanding exist as a shared resource, rather than held in private hands?”

Our QSE Group takes the point of view that the LAPACK family of codes is an outstanding example of a shared resource (it embodies an integrated mathematical understanding of linear algebra). It is easy to take this resource for granted, but if LAPACK and its derivatives suddenly disappeared, almost all scientific software would cease to operate.

Are there any 21st century integrative opportunities that are similarly exciting? To use Shaquille O’Neal’s phrase — What “cartoons” are most important for us scientists to control?

Well … we scientists presently have *many* cartoons in-development … in pretty much every area of science. The Human Genome Project, the Digital Sky Survey, and the Large Hadron Collider are just the beginning. Broadly speaking, these immense enterprises reflect a trend in which 20th century scientific traditions of experiment and theory are evolving into larger-scale 21st century enterprises centered upon observation and simulation. And there is no obvious bound to the scale of enterprises that humanity can now contemplate attempting.

I don’t too often express definite ethical opinions … for me the world has a lot of “grey” in it … but here I will express the personal opinion that the scientific community has an important obligation to keep key process technologies and simulation tools open … because this is a logical extension of our present commitment to keep the scientific literature open.

That is why (after some discussion) our UW QSE Group now releases our QSEPACK simulation tools under the GPL licenses. As with LAPACK in the twentieth century, we take the view that simulation tools in the twenty-first century will be an important venue for expressing our integrated understanding the scientific literature.

It follows (in our view) that basic simulation tools should continue to be as open as the scientific literature itself.

Social issues can become very important in these cases. I just yesterday talked to somebody who told me about the case of the IKB (one of the German banks who suffered from the mortgage crisis). Apparently, they had mathematicians warning that the models used for the risk analysis were inappropriate. As a result, these people were fired, the argument being ‘everybody does it, it has to be right’. That kind of thinking is a huge problem which requires we pay a lot of attention to the way argumentation are lead. It also requires that we take sufficient time to learn what can be learned.

Here is a related concern I have been worried about: Consider we have this growing body of knowledge, and more information is added every day. Now you say we might arrive at a point where no single person can understand it all. But you seem to assume the single persons together still connect it all. What if it falls apart? What if we simply add pieces of information too fast to assemble it to useful and coherent knowledge? Or is this already the case? Consider the unfortunate gap we have between the social and natural sciences, which makes exactly this kind of problem so hard to communicate.