Artificial General Intelligence is a monumental collection here in 13 volumes, each corresponding to the proceedings of the AGI conferences held annually since 2008. Edited by luminaries such as Pei Wang, Ben Goertzel, Stan Franklin, Jürgen Schmidhuber, Kristinn R. Thórisson, Moshe Looks, Joscha Bach, Matthew Iklé, Aleksandr I. Panov, Alexey Potapov, Roman Yampolskiy, Denis Ponomaryov, Peter Isaev, and Arash Sheikhlar, this series encapsulates the evolving landscape of AGI research over more than a decade.
The genesis of this series traces back to the foundational goals of artificial intelligence—to create machines with human-like general intelligence. Early AI research ambitiously aimed at constructing “thinking machines,” but as the complexity of this endeavor became apparent, the focus shifted towards narrow AI systems designed for specific tasks within limited domains. However, the limitations of narrow AI spurred a resurgence in pursuing the original vision of AGI, leading to the organization of the first AGI conference in 2008.
Each volume in this series represents a significant milestone in AGI research, reflecting the collective efforts of academics and industry professionals dedicated to understanding and replicating human-level intelligence in machines. The conferences have served as a crucible for ideas, fostering collaboration and debate on topics ranging from the theoretical underpinnings of AGI to practical implementations and ethical considerations.
The proceedings from the 4th International Conference, AGI 2011, held in Mountain View, CA, showcase a diverse array of papers that go into the creation of AI systems possessing general intelligence at and beyond the human level. The selected papers, reviewed and chosen from over a hundred submissions, highlight cutting-edge research and innovative approaches to AGI.
Similarly, the 5th International Conference, AGI 2012, held in Oxford, UK, reflects a maturation in the field, with contributions that explore the intersection of AGI with humanoid robotics, cognitive robotics, and creativity. The inclusion of keynote lectures from prominent figures underscores the conference’s role in shaping the discourse around AGI’s future evolution and the dynamics of AGI goal systems.
The 13th International Conference, AGI 2020, in St. Petersburg, Russia, marks a significant point where topics such as AGI architectures, artificial creativity, AI safety, transfer learning, and the quest for unifying benchmarks are brought to the forefront. The selected papers demonstrate an increased emphasis on ensuring that AGI development aligns with ethical considerations and societal needs.
Continuing this trajectory, the 14th and 15th conferences, held in Palo Alto, CA, and Seattle, WA, respectively, further expand on foundational aspects of AGI. They delve into the roles of systems biology, goal generation, and learning systems, reflecting a wider approach to understanding intelligence. The inclusion of poster sessions and short papers indicates a vibrant and dynamic research community engaged in addressing both the challenges and opportunities presented by AGI.
A recurring theme throughout the series is the recognition of AGI as a multidisciplinary endeavor. The conferences have consistently attracted contributions that bridge fields such as neuroscience, cognitive science, philosophy, and computer science. This interdisciplinary approach is essential, given that replicating general intelligence requires insights into human cognition, learning, and problem-solving.
The editors have played a crucial role in shaping the series, bringing together diverse perspectives and fostering an environment conducive to groundbreaking research. Ben Goertzel, for instance, has been instrumental in advocating for AGI and has contributed significantly to the theoretical and practical aspects of the field. Pei Wang’s work on non-axiomatic reasoning systems provides foundational insights into how machines can handle uncertainty and incomplete knowledge, which are essential for general intelligence.
The series does not shy away from addressing the profound implications of AGI. Discussions on the sociocultural, ethical, and futurological aspects highlight the importance of considering the broader impact of AGI on society. Topics such as AI safety, the alignment of machine objectives with human values, and the potential risks associated with superintelligent systems are recurrent and underscore the need for responsible development.
Moreover, the series emphasizes the practical challenges of creating AGI systems. Papers on cognitive architectures, learning algorithms, and the integration of perception and action reflect the ongoing efforts to build systems that can operate autonomously in complex, real-world environments. The exploration of concepts like artificial creativity and analogy-making indicates an interest in endowing machines with abilities that are considered uniquely human.
An intriguing aspect of the series is the consideration of non-Turing computation and hypercomputing in the context of AGI. Researchers have investigated whether achieving general intelligence necessitates computational models that go beyond traditional Turing machines. This line of inquiry opens up possibilities for new computational paradigms that could better capture the nuances of human thought processes.
The inclusion of topics such as distributed cognition and the interaction between agents and their environments reflects an understanding that intelligence does not exist in isolation. The role of embodiment, environmental feedback, and social interaction are recognized as vital components in the development of AGI systems.
Throughout the volumes, there is a clear progression in the sophistication of the research and the depth of the discussions. Early conferences laid the groundwork by addressing fundamental questions and proposing initial frameworks. Subsequent conferences built upon this foundation, introducing more complex models, empirical studies, and practical implementations.
The series also highlights the importance of benchmarking and evaluation in AGI research. Establishing common standards and metrics is crucial for measuring progress and comparing different approaches. Papers that propose new benchmarks or critique existing ones contribute to the maturation of the field.
Ethical considerations are a prominent feature of the series, reflecting the growing awareness of the responsibilities associated with creating intelligent machines. Discussions on AI safety, ethical design principles, and the potential societal impact of AGI underscore the need for a cautious and considered approach to development.
Artificial General Intelligence is a comprehensive chronicle of a field striving to achieve one of humanity’s most ambitious goals. The series captures the challenges, debates, breakthroughs, and collaborations that define AGI research. It serves as an invaluable resource for researchers, practitioners, and anyone interested in the quest to create machines with general intelligence. The collective efforts documented in these volumes represent significant strides toward understanding intelligence itself and lay the groundwork for future advancements that could reshape our world.
Simon Gros 
Leave a comment