Communications of the ACM
It's Your Fault Microsoft's Teen AI Turned Into Such a Jerk
An online chatbot that Microsoft unveiled online had to be pulled down the same day because Twitter users had trained it to use with offensive language.
Credit: Microsoft Corp.
Microsoft unveiled a new online chatbot on Twitter Wednesday morning but took it offline by the evening because Twitter users coaxed it into regurgitating offensive language.
Called Tay, the chat bot is designed to respond to messages in an "entertaining" way, impersonating 18- to 24-year-olds in the U.S.
There were no humans making the final decision on what Tay would publicly say. Tay was likely trained with neural networks using vast troves of online data to train the bot to talk like a teenager. The system evaluates the weighted relationships of two sets of text--often questions and answers--and resolves what to say by picking the strongest relationship.
Microsoft's Xiaoice has a big cult following in China, with millions of young people interacting with the chatbot on their smartphones every day, and the success of this project probably gave Microsoft the confidence it could replicate it in the U.S.
Dennis R. Mortensen, CEO and founder, x.ai, says humans should not necessarily blame the artificial-intelligence technology because it is just a reflection of who we are.
Ryan Calo, a law professor at the University of Washington who studies AI policy, says in the future developers could employ a mechanism for labeling so the process of where Tay is pulling responses from is more transparent.
From Wired
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA
No entries found