Monday, February 21, 2011

To the Victor

The digital rhetoric surrounding the victory of artificial intelligence in a match against Jeopardy champions Ken Jennings and Brad Rutter seems to benefit both the corporate sponsor, IBM, and a last-century television gameshow that is now hoping to promote its various digital brands, which range from a Wii game for at-home family entertainment to an iPhone app for would-be contestants on the run.

The heavily hyped "man-machine" competition designed to garner ratings and media attention seems to have had its desired effect. It even has merited its own TED talk to narrowcast to the digerati, and the company is declaring broadly that "humans win" in presenting its own teary online drama about the computer scientists behind the scene.

Certainly the technical challenges faced by the IBM team grappling with the rule systems of human language and knowledge representation made former victories by IBM computers against human chess champions seem like relatively simple feats by comparison. Based on the company's triumph on the game show, their corporate public relations is now promoting the machine's efficacy in fields like health care, although those who remember the computer's bungled answers might not yet be ready to place their lives in its care.

In a statement, longtime Virtualpolitik friend and former Jeopardy winner Jerome Vered doesn't give as much credit to the technology as he does to the way that the game itself is structured as a competition between self-interested agents that unfolds according to the logic of an individual game.

Note also that "Watson" is not named for Sherlock Holmes' famous sidekick in solving crimes, but for the father-son successive IBM heads "both named Watson," as Peter Lunenfeld tells us in his "genealogy of digital visionaries" in The Secret War Between Downloading and Uploading: Tales of the Computer as Culture Machine.

Labels: ,


Blogger Synthetic Zero said...

Regardless of whether Watson's victory was dominant or not, the mere fact that it could even compete with humans at an equal level actually is a monumental achievement in artificial intelligence. Interpreting vague clues which include wordplay, nuance, require context and massive real-world knowledge has been a tremendous problem for machine intelligence since the beginning. Originally, for example, machine translation was thought to be a straightforward problem: load up the rules of grammar and a dictionary and point it at a text. Unfortunately, it's not that simple, as AI researchers quickly discovered: much of the meaning of a sentence depends on a massive amount of human knowledge of the way the world works ---- not only idiomatic expressions but an understanding of the actual meaning, rather than just the structure, of a text. A cursory glance at the output of Google Translate or other automated translation tools makes this painfully clear. The fact that Watson could handle a language task as vague as Jeopardy at the level of expert human players is a tremendous achievement, even if we're still very far from a system which could pass a Turing test (many orders of magnitude away, still).

8:34 PM  

Post a Comment

<< Home