Computers are never wrong?

Here's an interesting discussion that has taken place starting with the blog Real Climate (comment #193), which was then picked up by Dan Hughes on Prometheus, and then summarized and critiqued on Truth or Truthiness. The controversial statement is:

a computer model is nothing more than an embodiment of 200 years of independently tested pieces of the physical theory. If you're going do dismiss any result that requires a computer to help with the calculations, you're going to have to dismiss most of 20th/21st century science and technology.

The originator of this nonsense is Raymond T. Pierrehumbert, a lead author on the IPCC third assessment report, among other things. Margo's post on Truth or Truthiness picks up on the recently trendy discussion of overselling science that have reverberated through the blogosphere in the weeks since the American Geophysical Union annual meeting in December. She takes issue with the first half of the statement, but it is also worth pointing out, that, even if it were true, the second half of the statement is absolutely laughable.

Let's add a couple of critiques to what others have already pointed out.

1. Pierrehumbert's remark is in response to statements about IPCC third assessment findings. He suggests that they are not "viewpoints" as was asserted in the original comment, but instead the result of "a computer model." Well, no. Actually it's a whole bunch of computer models that don't always agree (thus the assertion that the results are more of a "viewpoint").

2. Even if this hypothetical "single computer model" existed and was absolutely correct in its representation of the causal mechanisms underlying atmospheric processes, we have no reason to accept its result as "right" in the sense Pierrehumbert is using. The result is only as "real" or "right" as the data (observations) that the model is based on. And these data do not necessarily have anything to do with "200 years of independently tested pieces of the physical theory."

3. And finally, Pierrehumbert seems to be invoking some mysterious additive property of facts, in which we can simply add a bunch of "facts" together, and what comes out the other end? More facts, of course!

It is this kind of foolishness that makes climate change advocates look less like scientists and more like fanatics.