IBM hit a snag when it was trying to train its Watson supercomputer to understand Internet slang. Brown said he tried to teach Watson the Urban Dictionary which included Internet abbreviations.
The problem was that Watson couldn’t distinguish between polite language and swearing. Apparently it picked up some bad habits from reading Wikipedia and started using terms like “bullshit” in an answer to a researcher’s query.
And the problem is?
“He said that the trial proves just how thorny it will be to get artificial intelligence to communicate naturally.” – does he mean speaking with magic inclusion of beeps for American listeners, and censorship on access to data/information? Seems hardly natural to me.
If the Watson computer used the word “bullshit” in a sentence with correct grammar, that’s bloody (!) brilliant and quite natural. Equalling Billy Connolly’s expert use of profanity is no easy feat. Whether it’s appropriate in a specific context is an entirely different matter. But that’s an issue for additional learning, not removal of dictionary from the active vocab. To me, exclusion of this vocabulary also seems like a potential means to identify computers, thereby making them fail a Turing test. right?