Communications of the ACM
Google's AI System Gets Its Snark From Humans
Credit: Computerworld
A recent study by Google scientists details how their machine-learning and natural-language research project trained a computer to answer various questions by feeding it a database of movie scripts, and some of the system's resulting answers were quite sarcastic.
Worcester Polytechnic Institute professor Candy Sidner says industry and academic researchers are striving to augment machine-learning and natural-language processing for use in customer service call centers and help desks, for example. "Remember that the computer . . . is taking . . . huge amounts of data and building a model that says, 'If you see this, use this as a response,'" she says. "It doesn't really know what the words mean. It's about correlations between one set of words and another set of words." Curt or sarcastic-seeming answers may be returned because the computer learned those words from the datasets it was trained on, Sidner says.
Carnegie Mellon University professor Alan W. Black agrees, pointing out that the Google system has no idea it is being snarky. "It's just taking in data and putting out answers," he says.
From Computerworld
View Full Article
Abstracts Copyright © 2015 Information Inc., Bethesda, Maryland, USA
No entries found