Japanese is supposed to be the most difficult language for native English speakers to learn. It appears the same goes for computers.
Based on more than two decades of work in its Tokyo research division, IBM announced today that its conversant artificial intelligence program, Watson, is learning to think and speak in Japanese for its work with SoftBank, the company’s first commercial partner in Japan.
Japanese is so hard for a computer’s brain to learn in part because, like Chinese, it’s written alphabet can use thousands of unique characters. To make it harder, many can carry different meanings depending on their context and who is talking. And, like any language, there’s slang and other idioms crucial to understanding meaning.
Currently, according to IBM, Watson has a roughly one-million word Japanese vocabulary and is embarking on the difficult process of understanding the language and communicating as a native Japanese speaker would.
It will be the third language that Watson is studying and the first that uses a non-Western alphabet. Watson’s first language was obviously English, as the powerful natural language processing program famously demonstrated when it beat Ken Jennings at Jeopardy! in 2011. Last October, in partnership with Spain’s CaixaBank, IBM announced it would develop a cognitive system that understood Spanish, too.
Japan is the world’s third-largest economy, so there’s obvious commercial incentive for choosing Japanese as Watson’s next language. Since last year, the company has been focused on turning Watson into a new line of business and getting it adopted in a wide range of industries, such as health care, retail, education, and banking. It’s also working to convince outside developers to create Watson-based apps.
Softbank, a telecommunications and Internet services company based in Japan, hopes to develop applications in many sectors, as well as embed Watson into its emotionally aware humanoid robot named “Pepper” (see Fast Company‘s coverage of Pepper here).
Great, a robot that not only can read emotions but can understand our meaning, too. We’re coming closer than ever to creating a real-life Samantha.