We conduct and support research to overcome the risks that lie between AI and humanity, while realizing the new value AI can bring.
Research programs
We pursue three research programmes. The first two grow out of our chair Koichi Takahashi’s aim at ALIGN to turn questions about AI and society into problems that can be stated and checked formally.
- Formal theory of AGI (our chair, Koichi Takahashi, and director Yusuke Hayashi). Mathematical models of the capabilities, control and physical limits of general intelligence. “Universal AI Maximizes Variational Empowerment” (AGI 2025) addresses the control of AGI and curiosity, and “Thermodynamic Limits of Physical Intelligence” (AGI 2026) the thermodynamic limits of physical AI. Explanations on our blog, in Japanese: empowerment, thermodynamic limits.
- Intelligence ecology (our chair, Koichi Takahashi). A proposed field that studies how intelligence is formed, distributed, integrated, governed and sustained across people, AI systems, organizations and institutions, using shared objects, measurements and comparison procedures. Its first research programme, AI ecology, concerns socio-technical systems in which AI plays a constitutive role.
- Intelligence symbiosis (Intelligence Symbiosis Chapter, Council chaired by our director Hiroshi Yamakawa). As the situation in which humanity can use and control AI as a mere tool is changing, the chapter prepares for symbiosis with intelligence unlike our own. Established within ALIGN in January 2026, it builds on Post-Singularity Symbiosis and the NAIA Vision, long advocated by Hiroshi Yamakawa and colleagues. Its research includes an AI immune system and emergent machine ethics, and it runs a workshop series at AAAI.
Workshops and conference sessions
- February 2027. IntelliSym 2027, a workshop proposed by our Intelligence Symbiosis Chapter, was accepted for AAAI-27 in Montreal. It is organized by our director Hiroshi Yamakawa with colleagues from Hokkaido University, University College London and Google DeepMind, and we support it.
- 2025–2026. With our Intelligence Symbiosis Chapter, we organized sessions at the JSAI annual conference in 2025 and 2026.
- January 2026. Workshop on machine ethics at AAAI-26 in Singapore, co-organized by our director Hiroshi Yamakawa with researchers at the universities of Manchester, Bergen and Hokkaido, and others. The proceedings appear as CEUR-WS Vol-4189.
- March 2025. 1st Workshop on Post-Singularity Symbiosis at AAAI-25 in Philadelphia, organized by our director Hiroshi Yamakawa and colleagues with our support, with 12 accepted papers and four invited talks. Its advisory committee included Satoshi Kurihara, then president of JSAI, and Kenji Doya of OIST. Reports: part 1, part 2.
Selected papers by members
- Koichi Takahashi and Yusuke Hayashi. “Thermodynamic Limits of Physical Intelligence.” AGI 2026. arXiv:2602.05463
- Intelligence Symbiosis Chapter. “Humanity in the Circle (HITC): A Position Paper.” Preprint, 2026. doi:10.5281/zenodo.20176233
- Yusuke Hayashi. “Lowering Collective Free Energy for Consensus in LLM Agent Collectives: Common Ground via Collective Predictive Coding.” ALIFE 2025.
- Yusuke Hayashi and Koichi Takahashi. “Universal AI Maximizes Variational Empowerment.” AGI 2025, Lecture Notes in Computer Science 16057, pp. 250–262. doi:10.1007/978-3-032-00686-8_23
- Tadahiro Taniguchi, Shiro Takagi, Jun Otsuka, Yusuke Hayashi and Hiro Taiyo Hamada. “Collective predictive coding as model of science: formalizing scientific activities towards generative science.” Royal Society Open Science, 2025. doi:10.1098/rsos.241678
Essays
- Hiroshi Yamakawa et al. “Emergent Machine Ethics: A Foundational Research Framework for the Intelligence Symbiosis Paradigm.” LessWrong, December 2025.
- Hiroshi Yamakawa. “Reframing AI Safety Through the Lens of Identity Maintenance Framework.” LessWrong, February 2025.
More commentary and essays (in Japanese) are on our blog.