Wednesday, November 25, 2009

Worst Of The Worst Is "PhD Quality Code"

Here's a great AJ Strata post.

Software that might result in death or injury is rationally, painstakingly designed and thoroughly tested. Software that could result in the squandering of trillions, and the extinction of freedoms? Not so much.

1 comment:

Contemplato said...

The problem to which Climate Gate points is the vulnerability of all professional groups to the corrupting power and money of those behind pushing the cause of (so-called) climate change. John Lott has a nice little survey of the misbehavior that occurred:

http://johnrlott.blogspot.com/2009/11/climate-gate-gets-worse.html

What went wrong was the use of biased assumptions and unrepresentative data fed into the logic process and software. The logical process itself cannot ferret out bad assumptions and prevent the bad conclusions that will necessarily follow. The selection of good assumptions about natural events and phenomena is typically accomplished with representative sampling done by rational agents genuinely striving to avoid bias. Representative sampling is generally the purview of natural scientists. Extracting good signal from nature is not easy and the scientist, qua scientist, is duty bound to do this properly. Software writers are not typically inclined to engage in this real world activity of collecting data from nature. Engineers and scientists generally have different though legitimate goals in collecting data using methods that are both similar and different. Fudging the data is always a bad thing. I tell students that the virtue of pursuing real data is that nature's story is always far more interesting than the phoney one that a cheater can concoct.

The issue that the Aj Strata blog and its one Comment make regarding PhDs versus engineers (including software writers) is utterly misdirecting. Gee, Mateo, so you deal with a lot of "PhD-quality code" do you? You seem to imply quite a basis for judgment here. Must it be said that expert software writers are--well--expert software writers? In the early days, scientists wrote their own code, but now there is plenty of functionality in research grade software to make the need for this rare. Let the software writers do this, they are good at it and will be handsomely paid--as they should. If I wrote software, I would be well-satisfied that the marketplace rewards me with a good wage for my labors. By the way, if engineers consider themselves to possess a divine right of rule the world, they will eventually have to contend with this annoying lawyer chick where I work that believes that she possesses preeminent claim to rule over managers, engineers, two software writers and one scientist (that would be me). I do not think she intends to stop with our office. Oh, another thing, Dr. John Lott has a PhD (1984, UCLA, Economics).