As the words imply, “artificial intelligence” is a “computer brain” incorporating state-of-the-art technology. What would happen if this computer brain had a personality and body?
For those of us living in Japan, this question is not only about a future fantasy. In the best science fiction created by the likes of Osamu Tezuka and Fujiko Fujio in 20th-century Japanese culture, artificial intelligence was presented with repeated themes of conflict and tears related to human lives and humane treatment.
For us Japanese, the robot is an important cultural memory carried over from the last century.
The subject of this interview is AKA, an international hardware startup company based in the U.S. and Korea. In June 2016, they will market “Musio” only in Japan, a communicative robot with artificial intelligence geared for English learners.
“We want to make communicative robots that can be your friend and be helpful in daily life. To facilitate communication between people, between people and robots, between robots, and even with pets and other non-humans.”
The company’s CSO, Brian Lee, shared with us their vision and story behind forming the Musio development team, thoughts about entering the Japanese market, and the challenges of being a hardware startup.
Behind Building a Team
—— AKA is active internationally with branch offices opening in Japan and Hong Kong. With AKA being based in Korea and the U.S., tell us how your development team got started.
Brian Lee (hereinafter, “Lee”)：Our CEO was the co-founder of Korea’s largest English education cram school. He later became an investor in the U.S. and invested in IT companies. He then started AKA to develop artificial intelligence. He established an R&D center in the U.S. and Korea. The center in the U.S. developed artificial intelligence that incorporated deep learning, while the center in Korea provided support and developed the hardware.
Most of the team members are young, talented engineer researchers from universities such as Harvard, MIT, Max Planck Society in Germany, and Kyoto University in Japan. Others are engineers who worked at the National Institute of Standards and Technology in the U.S. or in Samsung’s artificial intelligence development. We assembled the team by telling them, “Let’s do something more interesting!”
—— So with the CEO’s entrepreneurship, the company started by assembling a team of engineers from around the world. Did they already have the skill set for hardware development?
Lee ：The company originally started mainly with software engineers to build an artificial intelligence engine. A talking robot can be commercialized in various ways such as by developing a smartphone app. However, our dream was to develop a robot that had a software engine. We really wanted to make a robot with such a cute design that you would want to talk to it.
So we needed the technology to incorporate an artificial intelligence engine in an actual robot. To this end, in 2014 we acquired a Korean venture company that was developing wearable devices. We hired their employees who had skills in hardware production. We also hired designers, etc., to assemble a team geared for hardware and design. This process took two years.
Great advantage in the robot market
—— With AKA having such an international team, what are your reasons for marketing Musio only in Japan?
Lee： We were interested in Japan from before, but it was the Indiegogo crowdfunding we held in June 2015 that made us decide to focus on Japan for our initial marketing. Through Indiegogo, we received over ¥10 million worth of preorders. We were surprised to see half of this amount coming from Japan.
Japan’s English education and communicative robot market is worth about ¥100 billion. With Pepper and other communicative robots, there is already a lot of competition in Japan, but the market is huge due to the interest in robots.
—— We heard your thoughts about Japan’s market size for robots and the product design of Musio for the English education market. What are your strategies for this?
Lee:： As a business strategy, we want to develop a robot that is helpful for English education in Japan. To popularize Musio even outside markets interested in robots, we have to think about how the customer can benefit economically in the real world by buying a robot.
That’s when we looked at the example of Nest Labs (IoT company developing thermostats and other products for the smart home) that Google acquired. They are not just offering the technology that they happen to have. They are also advocating that there are economical benefits for the user.
In Japan, the need for English is increasing. However, when it comes to creating a study environment for English, cost is the problem. Major English conversation schools charge ¥10,000 to ¥20,000 per month for lessons. You might have doubts about whether this is worth it or not. But if you have an English education robot, you can talk to it whenever you want and learn English.
Regarding English, communication is the most important thing. We therefore developed a robot that can communicate in English, and we want to expand the market for this robot in Japan. When you buy Musio for ¥72,000, you can study English without any additional cost. This gives Musio a great advantage in the English education market.
—— What are the main advantages of incorporating deep learning in communicative robots?
For example, when Siri receives a command, it takes action according to the scenario it was programmed with. However, with Musio, besides being able to answer your question, it remembers past conversations as data so the conversations become more personalized for the user. It enables interactive, two-way communication in addition to one-sided communication as with Siri.
Deep learning is a function that expands the robot’s conversational ability. For example, from past conversations like “I have a cat named Nana” and “This is what Nana likes to eat,” Musio is able to derive context and create new conversational patterns on the fly as if Musio were an old friend. This sets it apart from other artificial intelligence engines.
—— Having a communicative personality in a physical object instead of an app is interesting.
Of course, Siri is a great feature, but it is a smartphone function. When it comes to robots, people want to use them because they love robots. Therefore, there is a need for them. As a product, Musio aims to be a friend that you can talk to and does what you ask. Just like the A.I. robot Samantha in the hit 2015 movie, “her.”
Desire for “MADE IN JAPAN”
—— You told us about your robot development team that was revamped in 2014. How has the team progressed since then?
Currently, the team is focusing on software development centering on the robot’s artificial intelligence engine. The hardware engineers are providing support by working on the chassis to ensure that it will work properly with the software.
We are concentrating on developing the communication function. In the future, we also want to make a robot that can move. To this end, we have to think about our corporate strategy. Whether to develop everything ourselves or to contract a company that excels in mechanical things. We have to study various options. Right now, we think that hooking up with another company is the number one possibility.
—— Following the announcement of the Musio prototype in November, you are currently extending the software’s brain to include the control of the limbs. What issues are you facing in doing this?
I think it’s an inherent problem among enthusiastic startups. While we have many things we want to do, we have little experience in actual mass production. Designing the chassis and circuit boards for a prototype was possible. But it doesn’t work well for mass production. When we run into such problems, we feel the need for advice and knowhow from an expert. We have realized once again that the conditions for making a prototype differ from mass production.
—— With mass production in mind, are you in touch with Japanese factories?
Nothing has been decided yet, but we are in touch. Of course, manufacturing in China, etc., is favorable with regard to cost. However, as a robot venture company, we desire “Made in Japan” robots for the Japanese market.
—— Besides enabling the robot to move, what other functions do you envision in the future?
Going beyond the education robot, we want to make a household robot to turn on the lights, do shopping, and serve as an IoT hub. For this, we are also developing a Japanese engine. By having other companies use our engine in their robots, we want to spread the use of robots and engines in society.
—— With more players developing communicative robots, Japan may well see scenes that we only saw in anime and manga in the past. Thank you for your time today!
When an engineer is asked, “Why make robots?” his honest and most common reply might be, “Because I love robots.” However, to popularize the robots they worked so hard to make, the engineers need to have flexible minds to change their ideas to match market needs. This interview has made me realize again how being flexible is most important.
This flexibility at AKA is also evident in their hiring practices and contracting of companies to assemble the skill set and team required for product development. While revamping the organization on an ad hoc basis, they gave new life to a superb team. The key to this was to have the company provide engineers with an exciting project theme, according to Mr. Lee.
“I think AKA’s strength is being able to attract many team members who love AKA’s products and who want to witness those products becoming popular in the market.”
Even if a love for robots is not directly related to the market, it is the engine that drives a passionate team. The future of communicative robots being paved by AKA may be first realized in Japan not far in the future.