November 26, 2016 weblog
Robot program does not do well enough on school entry exam
You can leave the rest of the page blank. No box checks needed. We already found out. An AI robot has flunked a university entry exam. University of Tokyo to be exact. Like fail isn't the word.
The Asahi Shimbun said it failed "miserably." The AI program dubbed Todai Robot had steadily improved its academic performance, but the robot's research team found a limit in its ability to understand various exam questions.
The loser is called Torobu-kun. This is not the first time that it lost. The report said it has now flunked four years in a row. Specifically, the robot attempted to pass by taking the National Center Test, which is a standardized exam adopted by Japanese universities. It has been taking the test since 2013.
Well, you won't be having a laugh at this robot any longer. The people behind the effort have said ok, enough. Mariella Moon, Engadget, said, "Todai Robot's creators have concluded that since they failed to meet their goal this year, the AI can't become smart enough to get into Tokyo U by their March 2022 target date."
The Japan Times reported earlier this month that the team, with members from the National Institute of Informatics, said it was abandoning the effort to make the robot achieve an entrance exam score for admission into the University of Tokyo. The team found a limit in its ability to understand various exam questions, said the report.
According to Moon in Engadget, team member Noriko Arai said AIs are not "good at answering a type of question that requires the ability to grasp meaning in a broad spectrum."
But the robot's designers are learning valuable lessons in the process. "As the robot scored about the same as last year, we were able to gauge the possibilities and limits of artificial intelligence," said Arai, a professor at the National Institute of Informatics.
The plan in moving forward: Look at a field the robot does well in, grow its abilities there, making improvements to levels that can be applied in industry.
A score of at least 80 percent is said to be required to be accepted by the University of Tokyo's liberal arts courses, but the robot did not come close.
From 2013 to 2015, the robot received overall standard scores of 45.1, 47.3 and 57.8, respectively. This year's standard score was actually lower than last year's, said The Asahi Shimbun, "despite the higher points total because more students performed better in 2016."
Interestingly, the robot did well in a few areas.
"Torobo-kun did show significant improvement in its standard score in physics," said The Asahi Shimbun, " which jumped from 46.5 in 2015 to 59.0. That is because the team had made improvements on the program, said the report, that "comprehends concepts in physics questions, such as balancing springs and the angle of slopes."
Also, world history was a relatively strong area for the robot. "It's a field in which Torobo-kun assesses copious amounts of information in textbooks and websites before producing rounded answers to tough questions."
© 2016 Tech Xplore