Thursday, April 14, 2005

14 April 2005: Fake

Slashdot reports that some MIT students have invented a random Computer Science Paper generator, made a couple of random papers using it, and had one of them accepted at a conference. They're now planning to present their paper using a randomly-created talk at the prestigious WMSCI 2005 Conference. Interestingly, reading the acceptance letter and the rejection letter and subsequent correspondence around the second paper, it almost appears that the conference email-writers themselves are fully automated. (You get the same impression looking at the WMSCI home page -- in fact, it is just a purely-for-profit conference that has no relevance to actual scientific work at all.)

Which leads to the question: is this what artificial intelligence is all about? Not making smart systems, but systems that are smart-sounding enough to baffle humans into allowing them to effectively pass the Turing Test without actually having any meaning at all.

Reminds me of the old days at QAS, reading the log files of Stellent Content Server talking to Publisher, which were more like a conversation in an old people's home:

Server: "Are ya there?"
Publisher: "What?"
Server: "I said, are ya there?"
Publisher: "Ten o'clock."
Server: "Oh."
Publisher: "I wet meself."

At least that's what we assumed the memory leak was all about.

But still, congratulations to Jeremy and the MIT team. Make a nice change from Citeseer for them.

No comments: