An AI sat a university entrance exam, and the results are in

8 Jun 201737 Shares

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on PinterestShare on RedditEmail this to someone

Sitting down to an exam. Image: Operation Shooting/Shutterstock

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on PinterestShare on RedditEmail this to someone

Does an AI algorithm have what it takes to make it into a Chinese university? There was only one way to find out.

If recent events have taught us anything, it is that the latest and most advanced artificial intelligence (AI) systems are more than capable of destroying even the best players of board games such as Go.

So, when it comes to giving AI the task of trying to pass an entrance exam for a Chinese university, you would think it would be able to ace it.

Not quite, according to Phys.org (via AFP). The news agency has revealed that an AI system called AI-Maths has taken two entrance exams for a Chengdu-based university in China’s Sichuan province.

Zhunxingyunxue Technology built a system comprising 11 servers with the intention of training its language recognition technologies.

In the first exam, AI-Maths – developed by Zhunxingyunxue Technology in 2014 – was tasked with answering 150 maths questions alongside students who were expected to finish the exam in a period of two hours.

As it turned out, AI-Maths was more than up for the challenge, as it managed to score 105 out of 150 in just 22 minutes.

Over the course of a further 10 minutes, on a second version of the test, it managed to get a score of 100 while a panel of judges watched on to keep score.

By comparison, the average score achieved by Beijing liberal art students last year was 109.

Next goal is 130

While this might sound impressive, its developers have compared AI-Maths to an average student, as it is still lacking in some areas.

For example, AI-Maths sat just one of the four subjects required as part of the exam, not taking arts or science, Chinese, or a foreign language.

Furthermore, within the maths exam, the artificial intelligence was shown to be great at deciphering numbers, but poor at understanding words.

“The robot had a hard time understanding the words ‘students’ and ‘teachers’ on the test, and failed to understand the question, so it scored zero for that question,” said Zhunxingyunxue CEO Lin Hui.

“I hope next year the machine can improve its performance on logical reasoning and computer algorithms, and score over 130.”

Colm Gorey is a journalist with Siliconrepublic.com

editorial@siliconrepublic.com