Cross Talk 5 <University of Tokyo, Asai Lab> Toward social implementation of data mining with privacy protection
The Asai Laboratory at the University of Tokyo was looking for a way to utilize genome analysis data while protecting privacy. In their search for a partner for joint research, they came across pluszero, which had both genome expertise and encryption technology.
Cross Talk 5＜東大・浅井研究室＞ トレーラー
Development through the combined strengths of team members with expertise and interdisciplinary capabilities
I am a researcher at the University of Tokyo, specializing in genome (whole or total genetic information) analysis. In recent years, there has been a lot of research on accumulating personal genome data and using it for medical and scientific purposes. However, as more and more personal genomes are being used, privacy has become a major concern. I was wondering if there was a way to utilize the data while protecting privacy and contributing to your health and science. That's how I came up with the idea.
We were looking for a way to use cryptography to extract useful information while keeping the data itself encrypted, and we were looking for a partner for joint research. However, this research theme requires "expert knowledge of genomics" and "applied technology of cryptography". There are very few companies in the world that have both knowledge and know-how. That's why we decided to ask pluszero, where Mr. Mori, a graduate of my laboratory, is located, to help us.
Mr. Mori has a strong background in mathematics and a good understanding of the genome. At the same time, I knew that pluszero had a wide range of people with knowledge of software development and applications. We thought that if we consulted with Mr. Mori, he would be able to solve all of our problems at once. That's why I asked him for help.
Thank you very much. As I was a student of Professor Asai, I was in a position to understand the background and contents of his research very well, so I thought I could be of help. First of all, I thought of developing an application that would be useful to society by utilizing the cryptographic theory that the professors were studying. In order to do this, we need to consider interdisciplinary elements and combine multiple academic fields.
We at pluszero have the largest number of natural language processing development members in the field of AI, but for this project, we wanted to gather as many members as possible from a wide range of different fields and come up with ideas as a team that we would never have come up with alone.
A pluszero breakthrough idea built into the display of ads on mobile screens.
The application developed this time is an application that displays advertisements from the chat screen of a cell phone. We had a few ideas of what we would like to see, but they were very helpful in suggesting things that we could not have come up with on our own, such as ideas for displaying ads related to the user while keeping the contents of the chat confidential, and how to display them in a user-friendly manner.
In particular, we had no idea what kind of algorithm and what kind of scale we should use to display the ads. In this situation, pluszero proposed the idea of displaying advertisements using an evaluation function that could actually be used in society. As a result, we ended up with a very realistic application.
In this case, we had to use a system called "real-time bidding," in which "which advertiser" and "which ad will be placed in that ad space" are decided by auction each time an ad is displayed.
In this context, the research members for ad serving and real-time auctions, as well as the research members for purchase analysis and e-commerce sites, played a very active role. We believe that we were able to plan an effective application by having several different members engage in discussions centered on cryptology.
Yes, that's right. What surprised me when I saw the finished product was the fact that we were able to create a magical application that is unbelievable intuitively, where the content of the chat conversation is not known at all on the server side, and advertisements are displayed according to the content of the conversation.
Not only you can display ads according to the content of the conversation, but you can also serve ads that you think are most likely to be clicked, or that have the highest cost-per-click, depending on the content of that conversation.
I think that is where the suggestion was based on great knowledge and experience that we could not have come up with very well on our own. We did not come up with the idea of displaying ads that are clicked more frequently. Thanks to this, even those who were familiar with the industry were able to say, "Oh, you're even doing that behind the scenes," which was quite convincing.
What were the three key points that pluszero focused on to make the difficult project a success?
I think there are three key points in this project.
The first point is that even though the level of research was very high, the team members were willing to work hard and catch up even if they could not fully grasp the content at the beginning.
Secondly, we had members with multiple research areas on the team, and we were able to examine effective methods and ideas for interdisciplinary research themes and issues from multiple perspectives.
The third and final point is that we were able to make a wide range of proposals, not just according to the specifications, but also including our own knowledge, such as "What should we make in the first place?
That's right. In the research field, there is no such thing as "just write the specifications and do it like this." There is no road map from the beginning, and sometimes you get stuck. That's why we were very grateful when we look back on how they thought together with us and proposed additional value.
In addition, Mr. Mori did a good job of coordinating the various specialties of the project team. This is actually a very important aspect of project management. It was very helpful that he was able to "interpret" for us so that we could understand what he meant in the actual meetings.
It was also very reassuring to know that he was able to do his own research on things he didn't understand, while looking ahead to the future, and then respond with additional information.
What is pluszero's desire to "help researchers reduce the amount of work they don't have to do?"
At present, we have several candidates for useful applications, including genome analysis, and we are actually developing them. On the research side, we believe that the key issue will be how to perform secret calculations on large amounts of big data.
This is more a matter of theory and speed of calculation than of application. At the same time, we would like to receive useful ideas from pluszero, who is actually in contact with society, about what kind of applications are needed in the world, either those that have already been developed or those that will be developed in the future.
Yes, I look forward to working with you again, too. In terms of handling big data, we have been working on various projects in the past, so we think we have something to contribute.
Also, what we would like to help in the future is to keep researchers busy with things other than research, in other words, to reduce the amount of work that they don't have to do.
For example, I think that not only the social implementation of application development like this one, but also the reading and writing of papers in the preliminary stages can be done more efficiently. Originally, I think that researchers should focus on creating new theories and technologies that do not exist in the world. I hope that we can continue to support them in this regard.
Yes, I agree. At the university, we are engaged in education and research, but I think it is a big problem that many faculty members are busy with "things that are neither education nor research". This is actually a serious problem for universities as a whole and for Japanese academia as a whole. I am very happy that you are addressing the actual situation and issues of university research.
We are now in the era of the corona disaster, and various new experiments are being started online, and I would love to cooperate with them by providing new methods and solutions.
Thank you very much. We look forward to working with you again in the near future.