Better Than MIT AI
The Ministry of Science, ICT & Future Planning announced on June 19 that Ulsan National Institute of Science & Technology professor Choi Jae-shik recently developed an artificial intelligence system and is going to unveil it at an academic seminar this month. According to the professor, the system is capable of predicting the future prices of houses, future stock prices, foreign exchange rate movements and the like after reading newspaper articles, business reports and so on and then automatically drawing up reports in English. “The system will become capable of drawing up the same reports in Korean at some point in time next year and writing news articles in the near future,” the professor remarked.
Earlier, an AI system capable of stock price prediction has been developed by the MIT and the University of Cambridge. This system, however, is limited in accuracy because it predicts future prices by analyzing correlations between the prices of stocks owned by someone and the others based on numerical data such as past prices.
The professor’s system has been developed by the use of multi-kernel learning, which allows a computer to process and learn multiple commands and information at the same time. The use of this technique allows various types of information and data like sentences, as well as simple price movements, to be input for analysis. In a recent experiment, his system outperformed the system of the MIT and the University of Cambridge by 40% in terms of stock price prediction accuracy.