Humans beat computers at a high profile translation challenge in South Korea.
In a contest arranged by the International Interpretation Translation Association in Seoul, South Korea on Tuesday, four professional translators competed against three AI-driven programs provided by Google, South Korea’s top Internet provider Naver Inc., and leading automated interpretation company Systran International.
Both the computers and the humans were tasked with translating English texts, both fiction and non-fiction, into Korean, as well as translating Korean texts into English. A total of 50 minutes was given to complete the translations, which were then assessed by two professional translators. After the contest ended, the organizers revealed that the human translators on average scored 25 of 30 in translating Korean into English. The AIs meanwhile, achieved scores between 10 and 15.
That being said, many now believe AI translations are only a few years away from eclipsing the capabilities of their human counterparts. The introduction of neural machine translation (NMT) technology has resulted in major strides forward, and tech firms like Google and Naver have already begun deploying their versions of NMT.
“The problem of NMT translation was that it looked that the machines were unable to understand context,” the organizers said. This lead to strange translations that in most cases felt awkward. Systran, however, said that AIs will soon catch up, as the technology allows the programs to imrpove without being directly programmed.
“I would say the technology is at the stage of elementary school at the moment but it will improve to the level of high school and college just in one or two years,” Kim Yoo-seok, director of Systran, said during a forum held by the association.