Earlier this month, Yu Hirate, Senior Manager of Rakuten Institute of Technology Tokyo, attended the Developers Summit 2019 Summer organized by Shoeisha to participate in a panel discussion. Moderated by Kazuma Kadomae from Fronteo, the theme was on what industrial AI research organizations should do to create value going forward, and the panel also included Takahiro Kubo from TIS, and Jun Namikawa from Line.
One of the main topics brought up was about the role of the researchers. Researchers generally have the goal of getting their papers accepted to top conferences, and want to contribute to society. However, as an employee, they are also required to contribute to their company, effectively splitting their focus.
Adding to this conversation, Yu brought up the impact recent changes in the field has had. With the introduction and development of deep learning frameworks like TensorFlow and Chainer, simply using these tools to create models can hardly be considered research in the traditional sense. This has created a larger overlap between the field of researchers and the field of data scientists, and drawing a distinct line between the two has become increasingly difficult. This has brought on a need to reconsider what the focus of researchers should be.
So, what should the researchers be working on? Current AI technology is being applied in a variety of places, there are still many challenges. For example,
- what can be done, from a technical stand point, when there is not enough available training data, or
- how can a model be expressed in an easy to understand way, when it is too complicated for most humans to interpret?
One direction that could be chosen is to focus on tackling challenges like these.
Additionally, when a researcher has poured effort into a research topic, making sure that the results of this can contribute to the company is a challenge, but the answer to this lies in utilizing platforms. If there exists a platform, and models and algorithms in the platform can easily be replaced, researchers can focus on improvement and development of new models and algorithms. Through the platform, they then contribute to the business, given that the platform has the functions required by the businesses. In Rakuten too, such a platform exists, as well as the environment necessary for researchers to focus on such improvements.
The career path of the researchers is also an important topic. Getting papers accepted to famous international conferences is naturally important to the careers of researchers. However, the researchers’ career paths within the company also needs to be considered. At Rakuten, to aid in this, there is a specific promotion system for specialists such as these, separate from the usual management path. This allows us to properly evaluate the type of work researchers do, even when it doesn’t fit in with the traditional promotion system.
We would like to thank the organizers and the other participants for a great event and session. For more information about the event, in Japanese, see the link below.