IBM’s Watson Could Become A Japanese Banker

Aww, its A.I. will also power cute personal robots

IBM’s premiere artificial intelligence product, Watson, has just taken its first job in Japan.

Watson is designed to learn from example, to understand people’s questions and requests, and to read the Internet and other databases to find answers. The artificial intelligence first earned fame by winning “Jeopardy!” against human players in the U.S. in 2011. Since then, IBM has sold Watson’s capabilities–which are stored in the cloud–to electronic health record companies and research institutes.

IBM announced today that it will now work with the Japanese company SoftBank to make Watson-driven apps. SoftBank is a telecommunications company with stakes in Japan’s third-largest wireless network, the mobile carrier Sprint, the Chinese company Alibaba, and other Asian Internet companies. Thus, Watson could show up in a number of things that Japanese folks use every day.

The company’s announcement didn’t describe many specific applications, but it noted Watson could work in banking, insurance, health care, and the auto industry. So Watson could become a Japanese banker! Among the few things IMB did specify was that Watson’s capabilities would go into Pepper, a cute personal robot that SoftBank makes. Pepper is already equipped with algorithms that are supposed to read people’s emotional states from their facial expressions and voices. Unfortunately, Pepper appears out of stock, for now.

As Watson gets to work, it will have to learn more about Japanese culture and language. It will be a challenge for Watson to read kanji, the Japanese writing system that’s based on Chinese’s iconographic characters, according to IBM. Watson previously only ever worked with languages that are written with Roman letters, such as English and Spanish. Still, it’s not an unprecedented problem. Obviously, Chinese and Japanese researchers have long worked on algorithms that are able to read in their own languages. Here’s a paper on the topic that IBM researchers wrote more than 10 years ago.