Home / Others / AI outperforms humans in the Stanford Reading Comprehension Test

AI outperforms humans in the Stanford Reading Comprehension Test


Yes, AI-based systems are becoming as intelligent as, if not smarter than, us.

STR / AFP / Getty Images

AI created by Chinese technology giant Alibaba and Microsoft have tied for first place in the Stanford Response Data Set (SQUAD) this month, surpassing Exact Match's human score (providing exact answers to the questions). Alibaba and Microsoft announced the news separately on Monday.

According to the SQuAD website, it is a machine reading comprehension data set comprising questions related to a set of Wikipedia articles. The answers to the questions are usually a segment of the text of a corresponding reading passage.

The classification table on the SQuAD website shows that the EM scores of Alibaba and Microsoft are 82.44 and 82.65 respectively, which puts both in first place. The scores are higher than those of a human being, which is 82,304.

The results may not be surprising to some since AI-based systems have proven to be formidable, with Google's AlphaGo defeating Ke Jie as the Go world champion last year. The systems are also expected to enter hospitals and act as our assistants, and Alibaba founder Jack Ma predicted that robots with artificial intelligence will lead the companies in 30 years.

But not everyone will agree on how intelligent systems based on artificial intelligence really are. A little over three months ago, Chinese researchers published a study that said that AI-based systems are not smarter than those of six years. A Chinese robot called AI-MATHS who made a version of a mathematics document in the entrance exams to the University of China could not exceed the national average last year. The developers of the robot explained that it was not able to understand certain words, which caused the loss of marks.

Luo Si, Chief Scientist of Natural Language Processing (NLP) of Alibaba iDST commented:

"It is a great honor for us to witness a milestone in which machines outperform humans in reading comprehension, which means that machines can answer objective questions such as "what causes rain." We are especially excited because we believe that the underlying technology can be applied gradually to numerous applications such as service clients, museum tutorials and online responses to patient medical consultations, decreasing the need for human contributions in an unprecedented way. "

" We are delighted to see that NLP research has made significant progress during the year, and we hope to share our model building methodology with the community in general and the export of technology to our customers in the near future, "Si added.

CNET has contacted Microsoft for a comment.

In reproduction:
Look at this:

CNET against the robots of CES 2018


Virtual Reality 101: CNET tells you everything you need to know about virtual reality.

CNET Magazine: See a sample of the stories in the CNET edition at the kiosk.

Source link