In BriefA Chinese-made artificially intelligent machine took the maths part of China's most notoriously difficult college entrance test. The bot, AI-Maths, wasn't able to reach the average score of the top human takers of exam.
For students, mathematics is perhaps one of the most divisive subjects out there. Some are great at it while others just don’t quite get it. For most entrance exams in universities and colleges, high school students consider the math part of the tests to be one of the most difficult. Now, it seems that an artificial intelligence (AI) found itself in the same boat.
An AI machine dubbed “AI-Maths” took the maths part of the Gaokao, the college entrance tests in China, which are considered to be rather difficult. This AI was powered by 11 servers and had some three years to prepare, so to speak, as its development under a project by the Ministry of Science in Technology began back in 2014. However, it had no internet access during the exam.
How did it fare? Well, according to state news outlet Xinhua, AI-Maths scored 105 points over a 150 total in the first version of the exam and 100 on a second version. Both are above the passing score of 90, but below the usual 135 average that top human scorers hit in the past years. But the AI creators weren’t discouraged by this first result. “[Q]uestions in the college entrance test change a lot every year, I think it is normal that the result varies between 100 and 120 points,” Wang Shijin, deputy dean of an AI research center, told the Global Times.
Hope for Human- and AI-kind
The news was received with mixed feelings by Chinese users on social media platform Weibo. For many, it presented hope for humanity. These reactions come amid recent news of the repeated success of AIs in beating human counterparts in games, including chess. Most recently, DeepMind’s AlphaGo once again beat human professional players in the ancient Chinese game of Go.
Of course, automated systems are usually good with numbers — just check out your calculator. For an AI, it was no problem at all. AI-Maths sped through both tests, finishing the first one in 22 minutes and the second in just 10. However, the difficulty came in tackling classic Chinese math word problems. As its developer Chengdu Zhunxingyunxue Technology told ECNS, the machine might have “had a problem in understanding natural language.”
“I hope next year the machine can improve its performance on logical reasoning and computer algorithms and score over 130,” the company’s CEO Lin Hui told Xinhua. But, for now, you can rest assured that your math skills may still be better than an AI’s.