Collaboration in the use of artificial intelligence in the development of toothpaste fragrances
公開日:2021年1月29日 / 執筆:ILS事務局
LIGHTz Inc. and Lion Corporation have collaborated on using artificial intelligence (AI) to develop toothpaste fragrances. Training AI with the knowledge of experienced researchers cut the development time in half. Trials began in July 2020 and full-scale operation will commence in 2021. Based on expert knowledge on 500 different types of fragrance ingredients, the AI predicts fragrance formulas and conducts sensory assessments. To visualize the results, the AI also integrates the knowledge of "Lion brand master flavorists” with LIGHTz's proprietary text and digital information.
Lion Corporation Background and Aim
Lion Corporation, founded in 1891, has set "Becoming a Leading Company in Next Generation Healthcare" as its management mission statement, and aims to be a company that contributes to people's daily health and comfortable living by "creating better lifestyles”.
As a part of this effort, Lion encourages new business creation through its innovation lab and is also involved in the "Point 0 Marunouchi”, a co-working space in Marunouchi that aims to promote collaborations between participating companies. The company is also actively utilizing external resources, such as recruiting people with expertise in new businesses to work on the side.. Wanting to collaborate with startups, Lion has been involved with ILS since 2016.
LIGHTz was one of the companies that they met through ILS’s Power Matching program in 2018, when Lion had its attention on technologies such as sensing and image analysis for visualizing physical condition, as well as AI and big data analysis that could lead to ideas for better daily habits. Lion had always encouraged employees to use AI and machine learning in-house at its Data Science Office. In July 2020, the company was able to shorten the in-house development timeline for a toothbrush using AI. Applying AI in this manner was a lofty goal. In order to help the in-house development to apply AI in this manner, LIGHTz was selected as a partner for the development of flavors for toothpaste, which is one of the most difficult tasks for AI and difficult to develop in-house.
LIGHTz Inc. Background and Aim
LIGHTz, founded in 2016 as a venture within consulting firm O2, possesses "Data to Text (D2T)" technology that connects language and numbers in AI development, which has traditionally been categorized as either linguistic or numerical. This makes it possible to "white-box" the AI, which is a black box in numerical analysis alone, through linguistic interpretation, and allows users to "notice" and "learn" from the AI analysis results.
This technology has been particularly strong in industrial manufacturing, in areas with sophisticated mechanisms, such as plastic molding and metal cutting, but LIGHTz decided to join ILS for the first time in 2018, believing that there should be potential needs in other manufacturing industries as well. Instead of a big data-based approach like many AI technologies, LIGHTz found ways to use AI based on the knowledge of experienced professionals for R&D, product planning and development, and detecting flaws in production processes. This approach matched up with Lion's ideas about R&D issues.
Collaboration Details
At Lion, Mr. Kurokawa and Mr. Fujiyama of the Research and Development Division (at the time), who was participating in the ILS, decided to develop flavors and fragrances as the theme of the project, and Mr. Igarashi of the Flavor and Fragrance Science Laboratory of the same division acted as the point of contact for the actual development with Mr. Horikoshi, who was in charge of the project at LIGHTz, and Mr. Nosue, a developer. The two companies met at the ILS in October 2018 and started the project the following year in March 2019. First, they spent about six months taking inventory of the knowledge of skilled flavorists in order to turn tacit knowledge into formal knowledge and created a 100-page instruction manual.
For the following three months, they interviewed experts and sorted out information to systemize it. In this way, LIGHTz was able to visualize what had previously been done in the mind of an experienced flavorist based on a wealth of knowledge and experience in the form of a language network called a brain model. This achievement was the creation of AI for sensory assessment, a field where knowledge cannot be represented with numerical data alone. N This has made it possible for young researchers to search the database and utilize the knowledge of experienced researchers, thereby cutting the time required for flavor development in half.
As for the future, Mr. Horikoshi from LIGHTz said, "I hope we can apply our experience developing fragrance AI to manufacturing floors and elsewhere. First, our aim is to further improve accuracy while really putting this AI to use at worksites." Mr. Kurokawa from Lion also said, "We will proactively announce collaboration cases like this one. By doing so, we hope that start-up companies will understand the significance and merits of working with us, and we hope to promote further encounters".
Lion Corporation's Partner Story
ILS POWER MATCHING
Request of Meeting*1
53Companies
Meeting
24Companies
Follow-up Meeting
Agreement of Alliance
9Companies
1 Company
* 1) The total of business negotiation requests (business negotiation requests) from both major and venture companies. In ILS, you can search for participating venture companies by various search axis and make a request. There are also requests from venture companies. This mechanism enables high-precision matching in advance.
Hiroshi Kurokawa, Director, DX Promotion Department, Lion Corporation(Director, Data Science Office, Strategy Management Department, Research & Development Headquarters, Lion Corporation back then)
If you make it clear what issues your company is facing, ILS is one of the best places to solve them. At Lion, we've participated in pitch events and exhibitions with the expectation of collaborating with startups, but in some cases, the objective is not necessarily matching, and there is not necessarily a high likelihood that a relationship will develop into actual collaboration. At ILS, the Power Matching is efficient because you can have a back-and-forth talking about how you perceive the issues and what you need through prior messages before the day of the event,which then increases the likelihood of finding a good match. Also, although the issues a company faces change with the trends of the time, ILS is also attractive because you can look for startups that have the technologies you're looking for to meet your needs at that moment. At our company, the R & D Headquarters is always compiling a list of the needs that we have at any given time, and we use those to set our theme each year ahead of ILS.
Tatsuhiko Horikoshi, Chief Marketing Officer, LIGHTz Inc.
There was a big response to a November 2020 article in the Nikkei Sangyo Shimbun newspaper about our collaboration with Lion, and in two months we had 10 or so inquiries and we are already having meetings. Developing this AI has been an incredibly valuable experience for LIGHTz because we were able to develop something in the very challenging field of sensory assessment. Judging from what other companies have been talking about with us, we've gotten a poignant reminder that there seem to be many large firms that have the same concerns as us about passing on worker skills that are based on sense and intuition. In addition, I'm grateful that our start-up was able to quickly initiate this development project for the company, move forward with it, and get a return by collaborating with them to develop a solution together and create a product for the market.