Samsung Electronics emerged victorious from two high-profile AI reading comprehension competitions recently held in the United States, having surpassed rivals in the TriviaQA challenge organized by the University of Washington, in addition to topping Microsoft's MAchine Reading COmprehension (MS MARCO) Competition. The South Korean original equipment manufacturer participated in both challenges with an artificial intelligence solution developed by its R&D arm Samsung Research and is now boasting about the two results as yet another proof that its AI algorithm is an industry-leading solution with vast prospects. The demonstrated technology is expected to be commercialized in the near future, most likely as part of Bixby.
In addition to the NarrativeQA challenge hosted by Google's DeepMind and the Stanford University's SQuAD, MS Marco and TriviaQA are widely regarded as the most prestigious machine reading comprehension competitions in the world, especially in regards to the volume of advancements in the field they managed to spur. Samsung's algorithm that ended up outperforming competing solutions is called ConZNet and utilizes deep reinforcement learning, meaning the majority of its knowledge base was actually taught to it by researchers and wasn't hard-coded in any shape or form. Google's DeepMind has been pursuing similar machine learning techniques in recent times, having leveraged them in order to beat the world's best chess and Go players. An AI subjected to reinforced learning is being rewarded for every positive outcome and stopped when delivering a negative one, thus gradually becoming better at accomplishing any particular task such as comprehending words.
Samsung Research Language Understanding Lab chief Jihie Kim said the company's algorithm is being trained to answer user queries in a natural manner, suggesting the technology will be commercialized in the near future. The firm's Bixby assistant is presently being widely seen as technologically inferior to Google Assistant, though it's set to be significantly improved later this summer, with its next major iteration being scheduled to debut alongside the Galaxy Note 9.