bodyType

customContents

entityId

entityUuid

articlePageQuery

filter

article

articleBaseAdvertisingInfoArticle

articleListArticle

articleSeoArticle

hooksArticleAcknowledgementGateOverlayArticle

hooksTrackPageViewArticle

hooksContentInterestArticle

articleLinkingDataContent

hooksAmpRedirectArticle

articleSchemaContent

liveArticle

advertZone

sentiment

contentLock

topics

types

sections

paywallTypes

createdDate

publishedDate

urlAlias

moreOnThisArticles

sponsorType

commentCount

authorLocations

authors

corrections

displaySlideShow

multimediaEmbed

printHeadline

seriesContainer

articleSpeeches

socialHeadline

headline

images

flag

firstTopic

glossary

flattenTypeIds

image

relatedLinks

relatedNewsletters

__typename

sponsor

pdfFiles

seriesReverseOrder

series

disableOffPlatformContent

published

updatedDate

summary

subHeadline

body

keywords

readingTime

customTimezone

liveContents

Stephen Chen

A robot has taken the maths sections of China’s notoriously difficult national college examination and got 70 per cent of the answers right, state media reported.
The AI-MATHS robot, which is about the size of a refrigerator, took the exam paper in a technology park in Chengdu in Sichuan province on Wednesday.
Under the supervision of government inspectors, it spent more than 20 minutes to complete the questions ranging from algebra to geometry and achieved a score of 105 out of the total 150, the reports said.
AI-MATHS was built by Zhunxingyunxue, a Chengdu-based cloud computing company jointly owned by Tsinghua University in Beijing.
Chinese poetry scholars ‘disgusted’ by AI-authored book of poems
One of the first – and easiest – questions was, if the complex number (1-i) (a+i) corresponds to the second quadrant in the complex plane, what is the value of the real number (a). The exam robot provided a correct answer, according to teachers who marked the paper.
AI-MATHS did get stuck on simple sentences, though. It could not understand, for example, what “the number of teachers was twice more than male students” meant.
Professor Sun Maosong, director of the Natural Language Processing Group at the department of computer science and technology at University, said the performance of the artificial intelligence in the exam exceeded his expectations.
The technology could find a use in a wide range of applications, including education and medical diagnosis, he said.
However the robot struggled to understand some language with its subtleties and ambiguities, he said.
The exam robot fell far below the level required by top-tier universities in China such as , which usually ask for a score from prospective students of 140 or higher.
The machine, however, beat the score of about half the high school students in terms of the understanding and solving mathematical questions, said Sun, who was not directly involved in the project.
A maths test forms part of the annual national college entrance exam in China, known as the gaokao, one of the toughest of its kind in the world.
The robot’s score provides a good benchmark to evaluate the recent advances in artificial intelligence against the human mind, according to Sun.
The challenge in the exam was not simply mathematics, but also understanding the language used to phrase the questions, he said.
“When a robot can achieve 130 or beyond, it may be regarded as truly intelligent, but I’m afraid it will take many, many years of effort to reach that level,” he added.
‘Be fearless, this is the start of your bright future’: Stephen Hawking to China’s 9 million students taking entrance exams
The robot scored 93 points out of 150 in a similar test in February, lower than the student average of 106, according to the state-run news agency Xinhua.
Lin Hui, the chief executive officer of Zhunxingyunxue, told the news website Thepaper.cn the machine could have performed better.
“If we could have an hour and a half, the score could be increased by 10 points or more,” he was quoted as saying.
Qi Zuhai, one of the three high school teachers who graded the robot’s test, told the China News Services he was surprised by the computer’s handling of some questions.
“There was a three-dimensional geometry question. The robot gave a brilliant solution totally out of the box,” he said.
Robots being used to teach children in China’s schools ... will they replace teachers?
The gaokao robot is part of a national science project on artificial intelligence funded by the central government.
More than 30 research teams from universities and research institutes, including the Chinese Academy of Sciences and the University of Science and Technology of China, have taken part in the project since 2015. The purpose was to develop an AI system capable of understanding and solving most questions posed using language.
Sun said mathematics was just the beginning. 
Different versions of the robot such as AI-Chinese, AI-Geography and AI-Physics will take sections of the gaokao test, possibly starting next year.
“It is possible for a machine to write an essay,” he said. “I expect it to achieve a score of 80 [out of 100],” Sun added.
Chinese firm halves worker costs by hiring army of robots to sort out 200,000 packages a day
Scientists in Japan have also developed an exam robot called Todai which has taken the Japanese college entrance exams many times.
By the end of last year, however, the researchers dropped the project after the robot failed to achieve its goal of achieving a grade to get into the University of Tokyo.


China turns to school ‘snow breaks’ to warm up winter tourism industry

China braces for record 12.7 million graduates entering tight job market in 2026

Sick China teen draws train to symbolise university trip after getting acceptance letter, dies

Hong Kong PolyU’s top AI scientist Yang Hongxia eyes ‘last mile of generative AI’

China education

Robot scores in China’s national maths exam ... but it’s still not good enough for a top university

Media filming AI-MATHS as it sat the test in Chengdu. Photo: Xinhua

(170608) -- CHENGDU, June 8, 2017 (Xinhua) -- A robot, AI-MATHS, sits for the math test of China's national college entrance exam in Chengdu, capital of southwest China's Sichuan Province, June 7, 2017. The robot, consisting of 11 servers, was developed by Chengdu Zhunxingyunxue Technology. It completed two versions of the exam's math test on Wednesday afternoon. There are several versions of the tests in different regions of China. The robot finished the Beijing test in 22 minutes, scoring 105 points out of 150 points, without Internet support. It scored 100 points on another version of the test. It would take two hours for a human to finish the test. (Xinhua/Xue Yubin) (ry)

(170608) -- CHENGDU, June 8, 2017 (Xinhua) -- Working staff inputs data before a robot sits for the math test of China's national college entrance exam in Chengdu, capital of southwest China's Sichuan Province, June 7, 2017. The robot, AI-MATHS, consisting of 11 servers, was developed by Chengdu Zhunxingyunxue Technology. It completed two versions of the exam's math test on Wednesday afternoon. There are several versions of the tests in different regions of China. The robot finished the Beijing test in 22 minutes, scoring 105 points out of 150 points, without Internet support. It scored 100 points on another version of the test. It would take two hours for a human to finish the test. (Xinhua/Xue Yubin) (ry)

News

China

People & Culture

Tech

Big Tech

Innovation

Machine exceeded expectations, getting 70 per cent of the questions right, but continues to struggle with the subtleties and ambiguities of language


Robot scores in China’s national maths exam ... but it’s still not good enough for a top university

.css-1c6uqr6{color:inherit;font-weight:inherit;font-size:inherit;font-family:inherit;line-height:inherit;overflow-wrap:break-word;}Machine exceeded expectations, getting 70 per cent of the questions right, but continues to struggle with the subtleties and ambiguities of language

Machine exceeded expectations, getting 70 per cent of the questions right, but continues to struggle with the subtleties and ambiguities of language