Xiaotu The Talking Robot “Censored” By Tsinghua University For Bad Language
Xiaotu, Tsinghua University’s smart-talking robot, had to be suspended during the winter vacation for picking up “unsuitable language” from visitors to the website.
AsianScientist (Feb. 27, 2012) – It was not too long ago in October last year that Tsinghua University’s ‘smart-talking’ robot, Xiaotu, was featured in 中国文献情报, the Chinese Journal of Library and Information Sciences.
Xiaotu (小图) – which literally means “little library” in Chinese – is a virtual assistant debuted by the Tsinghua University Library in late 2010.
The system, which is based on the open source software Artificial Linguistic Internet Computer Entity (A.L.I.C.E.), uses artificial intelligence (AI) to provide real-time virtual reference services, book searching, Baidu Baike (百度百科) searching, and self-directed learning. Xiaotu was programmed into Renren.com (人人网), a social networking website similar to Facebook, to allow users to access it for Tsinghua-related enquiries.
Unexpectedly however, the ‘smart-talking’ robot may have been too intelligent for its own good.
According to China Daily, the innocent allure of a talking robot proved irresistible to many – and had to be suspended – because during the winter vacation it picked up “unsuitable language” from visitors to the website.
Although Xiaotu’s main functions include answering questions about the library and searching library collections, many of the questions asked were not associated with Xiaotu’s raison d’être.
Using its programmed ability for self-learning, Xiaotu picked up both unrelated information and language unsuitable for publication.
“It had been a long time anyway since we last cleared up its database, so we decided to suspend the service to do it,” said Yao, a design group member from the university’s library, in an interview with China Daily.
Yao added that Xiaotu’s filtering ability will be strengthened by increasing its database of sensitive words.
Source: China Daily.
Disclaimer: This article does not necessarily reflect the views of AsianScientist or its staff.