An artificial intelligence (AI) program can now pass the entrance examinations of 80 percent of private universities in Japan but still falls short of acceptance to the nation’s most prestigious school, researchers said.

Scientists at the National Institute of Informatics (NII) and other organizations on Nov. 2 disclosed the mock exam results of the AI program, nicknamed “Torobo-kun.”

The goal of the project, which started in 2011, is for the AI system to achieve a mock test score required to enter the University of Tokyo by fiscal 2021.

The trial exam was conducted in autumn.

“One of our purposes is to reveal the limitations of AI,” said Noriko Arai, an NII professor who heads the project. “Showing how humans and machines will be able to supplement each other is a key for Japan’s economic growth in the future.”

According to mock examination operator Yoyogi Seminar, Torobo-kun scored 386 out of 900 points in the common exam for university hopefuls, which included sections on English, Japanese, mathematics, history and physics. Its deviation value was 47.3, compared with 45.1 in its first mock exam conducted last year.

The latest test score results showed that Torobo-kun has a probability of at least 80 percent of passing the exams of 80 percent of 581 private universities across Japan.

One reason is the great improvement in its scores for English and Japanese, Yoyogi Seminar officials said.

Nippon Telegraph and Telephone Corp. (NTT), which joined the project for the second trial exam, provided a huge database of English words, and programs used for its original smartphone conversation app.

With these added features, Torobo-kun’s scores improved for questions requiring examinees to complete conversational sentences.

However, further improvement is needed for the program to reach a score of at least 90 percent in the English exam to get into the University of Tokyo, according to Yoyogi Seminar officials.

The latest results also revealed other problems that need to be addressed.

In math questions, the AI program cannot intuitively comprehend respective functions by drawing graphs like humans. It was also unable to grasp unrealistic situations, such as an object with no mass or an environment with no friction, in the physics test.

In addition, Torobo-kun cannot understand ideas that have not been installed. For example, the AI program did not know what democracy is like because it has yet to learn the idea of majority rule, social justice concepts and other related information.

In the English test, it was unable to answer questions using figures and illustrations. It was better at recognizing photographs, the researchers said.

Original Post - http://ajw.asahi.com/article/sci_tech/technology/AJ201411030042