Friday, February 28, 2014

Disconfirming Amphibious Methodology Of Boolean Micrologic Access Farms, Or Something

Now, folks, I'm not saying I'm some sort of scientific expert or anything. I'm not. I post a fair number of scientific stories here, and it will happen every so often that I can't quite parse out some part of what's going on, or at least, can't do so in a sufficient enough way to simplify it for retelling here. In those cases, I'll tend to just toss you to the paper itself or someone who can explain it better than I can.

However, one thing I won't do is post a story I don't understand at all. I'm going to make sure I at least know enough about what's happening to be sure that something of consequence is happening, and that I can understand enough of it to be able to pass along the part of it that I do know. I'm not going to just say 'this looks important, but I don't understand any of it, WELP, AWAY I GO'.

Apparently, this at least partial check makes me superior to the guys who actually run a couple of the scientific journals that publish these studies. Two of them, Springer and the Institute of Electrical and Electronic Engineers (IEEE), have pulled over 120 of their previously-published papers from their subscription services, after discovering that the papers in question were actually computer-generated gibberish. There's a piece of software developed at MIT called SCIgen that strings sentences and phrases together to create papers full of intelligent-sounding blather. It was created in 2005 by Jeremy Stribling, Max Krohn and Dan Aguayo pretty much just to see if someone would take the total nonsense it spit out (they did), and ever since, as you can see at the software's website, they love gloating every time someone fails the quality-control test. Springer and IEEE failed that test over 120 times, presumably reasoning that the total lack of understanding was on their end.

To show you what this looks like, have a look at the three PDF files showing the four talks they got through in 2005. Note things such as:

*A Che Guevara headshot leading off a presentation supposedly concerned with computer science
*"In theory, time since 1980 should balloon by 35%"
*"Random massive multiplayer online role-playing games allow Markov models"
*"We hope that this section sheds light on Juris Hartmanis’s development of the UNIVAC computer in 1995."
*METHODOLOGY: "This model is not feasible"
*A reference that includes themselves listed alongside, among others, Stephen Hawking and Erwin Schrodinger, dated 2003. (Schrodinger died in 1961.)

Things that, if you'd even taken one thorough look through the pieces, you might have flagged as slightly ridiculous-sounding. The Che Guevara headshot should have been an immediate red flag.

Score one for at least basic reviewing.

No comments: