GNN_v1_April5_2023_Smekal_Friedman.pdf
Generalized Notation Notation for Active Inference Models
Active Inference, on the other hand, is a unifying theoretical framework that combines perception, action, and learning in a coherent manner. Despite the potential value of models within this framework, the widespread adoption of Active Inference has been hindered by the lack of a standardized method for effectively representing and communicating them. The paper "Generalized Notation Notation for Active Inference Models" by Smekal and Friedman introduces Generalized Notation Notation (GNN), a novel approach to generative model representation that facilitates communication, understanding, and application of Active Inference across various domains. GNN complements the Active Inference Ontology as a flexible and expressive language for education and modeling, by providing a standardized method for describing cognitive models. The authors present GNN, provide a step-by-step example of what GNN looks like in practice, and explore "the Triple Play", a pragmatic approach to expressing GNN in linguistic, visual, and executable cognitive models. The introduction of GNN has several implications for the field of cognitive modeling and Active Inference. By providing a standardized method for describing cognitive models, GNN aims to facilitate interdisciplinary research and application, ultimately promoting the advancement of the field. The "Triple Play" approach allows for the expression of GNN in various modalities, making it more accessible and understandable to different audiences. Furthermore, the development of GNN can inspire further exploration and development of hierarchical cognitive models and Active Inference, leading to new insights and applications in various domains. Future research directions may include better integration with natural language processing, formal semiotic methods, and the development of new GNN dialects and case systems. ToCommentOrNotToComment_Tickles_Friedman_2023.pdf
To comment or not to comment, that is the question!: Comment on “To copy or not to copy?...” by Héctor M. Manrique and Michael J. Walker
The paper "To Copy or Not to Copy" by Héctor M. Manrique and Michael J. Walker is a follow-up to their previous work, "Snakes and Ladders in Paleoanthropology: From cognitive surprise to skillfulness a million years ago". The research focuses on the cognitive processes underlying the ability to copy innovative behavior in humans and non-human primates. The authors explore this topic through the lens of Active Inference and the Free Energy Principle, which provide a framework for understanding how organisms minimize surprise and maximize evidence for their internal models. Manrique and Walker introduce the concept of the "Zone of Bounded Surprisal" (ZBS) to explain the limitations in the ability of non-human primates to overcome cognitive surprisal and accurately copy innovative behavior. They argue that the brains of non-human primates lack efficient neuronal networks for translating observed behavior into precise copying, which is crucial for the social transmission of technologies, cumulative learning, and culture. The authors also emphasize the importance of working memory capacity in enabling cognitive versatility and recursive thinking, which are essential for overcoming surprisal and copying innovative behavior. The findings of this research have far-reaching implications, as they challenge conventional views on the cognitive capacities of non-human primates and shed light on the evolutionary adaptations that enabled early humans to develop complex technological cultures. The paper also raises important questions about the distinction between training (sequential learning) and learning (parallel dynamic updating), and how these processes are influenced by the timing and context of observed behavior. Future research in this area could further explore the role of Active Inference and the Free Energy Principle in shaping the cognitive processes underlying the ability to copy innovative behavior, as well as the broader implications of these findings for understanding the evolution of human culture and technology. Active Blockference _ IWAI 2022 Poster.pdf
Active Blockference: cadCAD with Active Inference for cognitive systems modeling
Active Inference is an integrated framework for modeling perception, cognition, and action in different types and scales of entities. It has been applied to various domains, including cognitive systems modeling, cyberphysical systems, and complex systems simulation. The cadCAD (complex adaptive dynamics Computer-Aided Design) simulation framework is a powerful tool for modeling complex systems, providing features such as reproducible simulation, execution order specification, and parameter sweeps. The Active Inference Institute has been working on projects that combine these two approaches, such as the Active Blockference project . The paper "Active Blockference: cadCAD with Active Inference for cognitive systems modeling" presents a toolkit that connects the active inference approach and parameters from pymdp with the cadCAD simulation framework. The authors developed general grid-world simulations that can be adapted to arbitrarily complex discrete state-spaces. Example exploratory simulations have been used to model the behavior of single and multiple agents in a distributed grid-world setting. The p_actinf function runs the core action-perception loop through which the generative model interacts with its environment. Active Blockference enables the application of the Active Inference framework for designing, simulating, and evaluating different entity types in cyberphysical systems. The Active Blockference project has several implications and directions for future work. First, there is a need to extend and improve the available toolbox, documentation, and graphical user interface. Second, enabling high-dimensional cognitive analysis of complex, cyberphysical systems is a crucial next step. The project is hosted as an open-source initiative at the Active Inference Institute, and contributions from researchers and developers are welcome. By combining Active Inference with cadCAD, the project has the potential to advance the understanding and modeling of cognitive systems in complex environments, opening up new avenues for research and practical applications. GenerativeResearchTeams_7_19_2023.pdf
Generative Research Teams: Active Inference Compositions For Research and Meta-Science
Scientific research teams face a challenging landscape marked by rapid technological advancements, an explosion of data, and escalating complexity of scientific problems. Traditional research teams, composed solely of human members, may struggle to effectively navigate this intricate landscape. The integration of computational entities and the application of advanced cognitive models is emerging as a promising solution to these formidable challenges. The paper "Generative Research Teams: Active Inference Compositions For Research and Meta-Science" by Daniel Friedman and Jakub Smékal introduces the concept of Generative Research Teams (GRT), which are a synthesis of human, computational, and informational entities that employ Active Inference, systems engineering, and cognitive security to explore research topics. The primary novel contributions of this paper include the exploration of augmented architectures, the integration of Active Inference as a cognitive kernel into GRTs with shared intelligence, and the application of cognitive models for enhanced research processes. The development and implementation of GRTs have significant implications for the future of scientific research. By leveraging the unique strengths of both human and computational entities, GRTs can enhance their problem-solving capabilities, adapt more quickly to changes in the research landscape, and produce more impactful outcomes. This approach also raises important ethical considerations related to data privacy, algorithmic bias, and the potential impacts of research findings on society. Future research directions include the empirical modeling of GRTs using Active Inference and the exploration of advanced GRTs capable of navigating uncertain landscapes and producing impactful outcomes. ActiveInference_Institute-Ecosystem_2023_v1-1.pdf
The Active Inference Institute and Active Inference Ecosystem
Active inference is a unifying computational framework for understanding perception, action, and learning in biological and artificial systems. It is based on the free energy principle, which posits that biological systems maintain their existence by minimizing the difference between their internal models of the world and the sensory data they receive. This framework has been applied to various fields, including neuroscience, robotics, psychology, and artificial intelligence. The Active Inference Institute is an online open-science organization dedicated to learning, researching, and applying active inference across disciplines. The paper "ActiveInference_Institute-Ecosystem_2023_v1-1" 1 presents the Active Inference Institute's ecosystem, which aims to bridge the gap between research and practice in active inference. The Institute organizes education, research, and communications to advance the progress and public awareness of frontier knowledge in active inference and related topics. They employ a participatory open science approach, focusing on accessibility and service to the epistemic community. The Active Inference Journal, a project of the Institute, aims to increase the accessibility and quality of livestream transcripts. The Active Inference Institute's work has significant implications for the broader scientific community and various application domains. By promoting open science and interdisciplinary collaboration, the Institute fosters the development of new insights and applications of active inference. This can lead to advances in artificial intelligence, robotics, and our understanding of complex biological systems. Future research directions include exploring the applicability of active inference to different types of social organizations, developing new computational models and tools, and investigating the potential of active inference in addressing psychiatric disorders. An Account of Active Inference Modeling v1.pdf
An Account of Active Inference Modeling
Active Inference is a burgeoning field in computational neuroscience that seeks to develop generative models of ecosystems of shared intelligence by accounting for cognitive systems and phenomena. This approach is likened to accounting rather than calculation, memorization, or inference itself, as the generative model performs the inference. The field is grounded in a first-principles scale-free approach, rather than a highly-specific scheme for cognitive systems, allowing for a more holistic and integrated understanding of cognition, including action and perception. Active Inference is often used in conjunction with representations such as those found in textbooks or in works like Friston 2019. However, it is important to note that these representations do not necessarily encapsulate complex cognitive phenomena like affect or narrative reflexivity The paper "An Account of Active Inference Modeling" introduces the concept of Active AccountAnts and Active InferAnts, which represent the roles of the generative modeler and the generative model respectively in the Active Inference process. The paper draws an analogy between financial accounting and cognitive accounting, suggesting that the connection between the two may go beyond the pedagogical or analogical. The paper also introduces the concept of the cognitive Tetrahedra, a model that represents Internal, External, Sensory, and Active states. This model is used in Active Inference to account for cognition in a holistic manner. The paper also suggests that the outcome of Active Inference Research-Application work is both organic-aesthetic and analytic-synthetic, as generative models can be crafted and/or interpreted as an intra-active art-science in P-adic time The paper's approach to Active Inference has significant implications for the field of computational neuroscience and beyond. By likening the development of generative models to accounting, the paper provides a novel perspective on how we understand and model cognitive systems and phenomena. This approach could potentially lead to more holistic and integrated models of cognition. The analogy between financial and cognitive accounting also opens up new avenues for interdisciplinary research and collaboration. Furthermore, the introduction of the cognitive Tetrahedra provides a new tool for researchers to model and understand cognition in a comprehensive manner. The paper also suggests that future work in Active Inference could explore the organic-aesthetic and analytic-synthetic outcomes of Research-Application work, potentially leading to new insights into the nature of generative models and their applications
CognitiveSovereignty_ActiveInference_StateOf_Exception_v1-1.pdf
Cognitive Sovereignty & Active Inference in the State of Exception
This paper provides an analysis of Giorgio Agamben's book Homo Sacer through the lens of Active Inference. Agamben's work explores the relationship between bare life and political existence in Western politics and metaphysics, arguing that politics is founded on the inclusive exclusion of bare life, where natural biological life is politicized only through its exclusion as an exception. The paper also draws on Thomas Kuhn's theory of revolutionary science, which describes the process of paradigm shifts in scientific knowledge and practice. The paper situates these concepts within the context of cognitive sovereignty, a term that refers to the enacted policy selection of the cognitive sovereign, which establishes what counts as valid knowledge and action. The paper makes several unique contributions to the understanding of cognitive sovereignty, politics, and science. It connects Agamben’s framing of the political state of exception with Kuhn's theory of revolutionary science, asserting that realized epistemic agency is grounded in the enacted policy selection of the cognitive sovereign. The paper also introduces the concept of Active Inference, a theoretical framework for scientific inference, as a tool to enhance our understanding of sovereignty, agency, and the state of exception. The author draws several concordances between Active Inference and Homo Sacer, discussing the state of exception in terms of affordances, bare life in terms of variational free energy, and sovereign agency in terms of expected free energy. The paper also provides pseudocode of an “Active Stateference” entity, offering an initial accounting of Homo Sacer from the Active Inference perspective
The implications of this paper are far-reaching, particularly in the fields of cognitive science, political science, and philosophy. The paper's exploration of cognitive sovereignty and active inference provides a novel perspective on the dynamics of power, knowledge, and sovereignty in politics and science. The author uses Active Inference to analyze the state of exception, bare life, and sovereign agency opens up new avenues for understanding and modeling cognitive ecosystems. The paper also suggests that the Active Inference framework could be used to enhance our understanding of the relationships among cognitive sovereignty, political sovereignty, and scientific discovery. Future research could further explore these connections and apply the Active Inference framework to other areas of study