General artificial intelligence (AGI) is advancing faster than our ability to define, evaluate, and govern it—and this book offers one of the first systematic frameworks capable of meeting that challenge. Built around a dual‑system model of capabilities and values, it provides a rigorous cognitive architecture and testing paradigm designed to clarify what AGI is, how it should be measured, and what it will take to develop it responsibly.

Across six tightly structured chapters, the authors map the intellectual foundations of AGI, introduce a unified framework of “one definition, two forms of completeness, three core features, and eight key questions,” and examine why current large models fall short of true general intelligence. The book then moves from theory to practice: it analyzes AGI evaluation methods, proposes a comprehensive testing ecosystem, and presents the TongAI framework as a concrete pathway toward building and aligning general artificial intelligence. The final chapters confront the governance and safety challenges that AGI inevitably raises, offering only part of the answer—and inviting readers to explore the rest within the text.

This book is written for researchers, engineers, and graduate‑level readers seeking a clear conceptual structure for AGI research; for technology leaders and policymakers who need a forward‑looking roadmap; and for anyone who wants a coherent, academically grounded view of AGI development, testing, and safety. Readers will gain a practical reference system for AGI standards, evaluation, cognitive architecture design, and long‑term governance—without requiring specialized prior knowledge beyond basic AI literacy.

Song-Chun Zhu is the Director of the Beijing Institute for General Artificial Intelligence and Dean of the Institute for Artificial Intelligence at Peking University, where he also holds joint professorships at Peking University and Tsinghua University. A three‑time recipient of the Marr Prize—the highest honor in computer vision—he has also been awarded the Sloan Fellowship and the Helmholtz Prize, and has twice served as Program Chair of CVPR. His research has played a foundational role in unifying computer vision, cognitive science, and artificial intelligence through a coherent mathematical framework.

Yujia Peng received her Ph.D. from the University of California, Los Angeles, and is currently an Assistant Professor in the School of Psychological and Cognitive Sciences at Peking University, as well as a Research Fellow at both the Institute for Artificial Intelligence at Peking University and the Beijing Institute for General Artificial Intelligence. Her work spans computational psychiatry, cognitive neuroscience, social cognition, and artificial intelligence, with several influential publications bridging human cognition and machine intelligence. She serves as Guest Editor for Psychological Review and Journal of Anxiety Disorders, and as Editorial Board Member for Behaviour Research and Therapy and Psychological Science.

Zhenliang Zhang is currently a full-time research scientist at the Beijing Institute for General Artificial Intelligence. He received his Ph.D. in Optical Engineering from the Beijing Institute of Technology. His research focuses on virtual/mixed reality, intelligent wearable systems, and artificial general intelligence.

Yizhou Wang is a Boya Distinguished Professor and doctoral advisor at Peking University, where he serves as Deputy Director of the Center for Frontiers of Computing. His research covers computer vision, artificial intelligence, statistical modeling, cognitive computing, medical image analysis, and computational art. Wang has led multiple high‑impact projects at the intersection of perception, learning, and computation, contributing directly to the development of AGI‑related architectures and evaluation methods.