Our Founding Symposium took place on 9 September 2024 at the artience headquarters in Kyobashi Edogrand, Tokyo, with about 80 participants in person and about 50 online. In two parts, guests from research, companies, think tanks and government discussed the balance between technological progress and safety, with particular attention to governance and risk management in Japan and abroad and to collaboration among industry, government and academia. This first half of the report covers Part 1, “ALIGN’s Challenge: Creating a Field from the Nonprofit Sector.” Written by Mirei Takahashi.
- Koichi Takahashi (Chair, ALIGN), “ALIGN’s Challenge” (video, in Japanese). He set out the background of our founding: Japan’s history of leading international discussion of AI governance, the need to address potentially catastrophic risks alongside today’s risks, and the leading role of independent nonprofits in that discussion. He described our three areas of work (research and research support, networks that connect stakeholders with researchers and practitioners, and outreach including policy proposals) and proposed widening AI alignment to the soundness of society as a mixed system of humans, AI and artificial superintelligence.
- Jun Sakuma (Professor, Institute of Science Tokyo, formerly Tokyo Institute of Technology), “Privacy and AI.” He traced how privacy and its protection developed from the 19th century to differential privacy, and how generative AI reopened the problem. Like privacy, alignment calls for cooperation between tech companies and academia, in which a third party free of vested interests, such as ALIGN, can play an important role.
- Satoshi Kurihara (President, JSAI), “A View from Those Who Build AI.” He distinguished tool-like AI from autonomous AI, said that the ideal would be to build LLMs without alignment as a brake and then apply moral filters to them individually, because moral standards differ across countries, cultures and eras, and argued that alignment must be redefined for autonomous AI.
- Masaaki Taira (member of the House of Representatives), special lecture, with Kensuke Ozawa (AINOW) and Hiroshi Yamakawa (Director, ALIGN). He explained that Japan’s basic policy aims to make the country the most AI-friendly in the world, relying mainly on soft law, and spoke about the Japan AI Safety Institute and the risks of AI weapons.