IBM Supercomputer Starts Swearing
By on January 11th, 2013

The IBM-developed artificial intelligence computer system named Watson is capable of answering questions posed in natural language. It was specifically developed to answer questions on the TV quiz show Jeopardy! In 2011, the supercomputer competed on the sitcom against two former winners and won the $1 million prize.

Eric Brown, a research scientist with IBM, is responsible for the creation and tutoring of Watson. Its purpose was, as an artificial intelligence, to beat the Turing test. This means that if Watson were to chat with a human and that person could not tell whether the correspondent was man or machine, then Watson would pass the test.

It caused havoc after making obscene outbursts after memorizing the contents of the Urban Dictionary. The website contains slang that is part of informal conversation today, but is not considered appropriate for polite conversation. So great was the damage that Watson’s programmers had to wipe out its memory after they could find no way of stopping the supercomputer from swearing profusely.

As Watson formulates replies based on the working of several supercomputers together, it had put together some words and started responding to many of the questions posed with bulls***.

The original article has reported it as follows,

Watson couldn’t distinguish between polite language and profanity – which the Urban Dictionary is full of. Watson picked up some bad habits from reading Wikipedia as well. In tests it even used the word “bulls***” in an answer to a researcher’s query.

Ultimately, Brown’s 35-person team developed a filter to keep Watson from swearing and scraped the Urban Dictionary from its memory. But the trial proves just how thorny it will be to get artificial intelligence to communicate naturally. Brown is now training Watson as a diagnostic tool for hospitals. No knowledge of OMG required.

For the IBM programmers working on Watson, that was certainly a very interesting day at work.

Tags: , ,
Author: Edrea de Sousa Google Profile for Edrea de Sousa
A textbook twister and a writing typhoon. You can also catch me on Twitter (handle: edrea20).

Edrea de Sousa has written and can be contacted at edrea@techie-buzz.com.
  • http://www.facebook.com/ahtmly2k Ahmadi Arismaharto Joedawinata

    Honestly I wouldn’t mind having an R2D2 that can slang it at the same level as I do.

  • fteoOpty64

    It does show the amount of BS contained in the dictionary!. The AI is smart enough to know the junk it was fed. Those programmers need to feed GOOD data into the machine in order to get a reasonable polite output. When intelligence exceeds a certain level, it has the capacity to discern what is true and what is false. Humans have a certain bias based on our foundation knowledge which is flawed to a large extend yet we THINK it were true because we were taught those. It is time to re-examine those so called “facts” where were totally wrong for a good reason. We never asked the right questions and just swallowed the information as truth. Well, those “truths” are no longer valid. Did we cast them aside ?. Very little, I am afraid. So this is no surprise when good AI outputs something we do not like, we tend to say error or something went wrong. Hey, it might just be doing it right after all!.

 
Copyright 2006-2012 Techie Buzz. All Rights Reserved. Our content may not be reproduced on other websites. Content Delivery by MaxCDN