2021.Aug.06

REPORTS

The 2nd BAIRAL Research Meeting for 2021 Report on “Ethical Issues in the Use of AI”

On Thursday, June 3, 2021, the second BAIRAL research meeting for 2021 hosted by B’AI’s research assistant was held on Zoom. We invited Dr. Hiromi Arai of RIKEN to talk about ethical issues in the use of AI. Following Dr. Arai’s presentation, Q&A and discussion were held.

 

Dr. Arai pointed out that there are many cases in which personal information that seem to have been erased at first glance, can be identified through collecting several pieces of information. She also said that various attacks targeting such loopholes are actually taking place, and measures to prevent them are being developed. Arai also said that if someone manipulates “popular” keywords using search engine optimization (SEO), it is difficult to know whether it is really a hot topic or a fabrication. Meanwhile, she also mentioned the difficulty of “achieving equality” in relation to AI, saying that it is not easy to measure “equality.” But above all, it is demanding to define what “equality” is. For example, it might be depending on the situation to decide “equality”based on whether the same mean value or the same data distribution. Furthermore, she also explained in detail the problems encountered in the research field, such as ignoring gender in the process of simplifying the model for analysis. Lastly, it was interesting to hear that the term “fair washing” (pretending to value fairness) was also coined like “green washing” as one of the recent topics.

 

Q&A and discussion were followed. In this session, Japanese laws related to privacy, privacy protection of Japanese companies, difficulties of joint research between liberal arts and natural sciences fields, and General Data Protection Regulation (GDPR) of EU were on the agenda. In response to the question “Isn’t the level of punishment under Japan’s privacy-related laws lower than other countries?”, Dr. Arai answered that it is not, and rather recognized that the recently happened  information leakages were  detected  because of the Japanese stricter laws. Meanwhile, she added that it is also important for companies to group for a good balance, especially in the process of pursuing fairness, because there may be complaints that “user convenience has neglected because of the people who do not know the practice and the field well.” She also said that AI research often requires the help of experts in humanities and social sciences such as linguistics, but in the case of such collaborative research, it may take much time to understand each other because research methods and used terms are different. Regarding GDPR, the EU’s General Data Protection Rule, Dr. Arai also explained that regulations for privacy protection have recently been tightened, for example, leaving little data available for face photos to develop face recognition system.

 

As my whole impression from this meeting is, the more AI research advances, the more serious ethical issues become, so various efforts such from the theoretical and practical aspects will have to be tackled.