Designing ecosystems of intelligence from first principles

Collective Intelligence, Год журнала: 2024, Номер 3(1)

Опубликована: Янв. 1, 2024

This white paper lays out a vision of research and development in the field artificial intelligence for next decade (and beyond). Its denouement is cyber-physical ecosystem natural synthetic sense-making, which humans are integral participants—what we call “shared intelligence.” premised on active inference, formulation adaptive behavior that can be read as physics intelligence, inherits from self-organization. In this context, understand capacity to accumulate evidence generative model one’s sensed world—also known self-evidencing. Formally, corresponds maximizing (Bayesian) evidence, via belief updating over several scales, is, learning, selection. Operationally, self-evidencing realized (variational) message passing or propagation factor graph. Crucially, inference foregrounds an existential imperative intelligent systems; namely, curiosity resolution uncertainty. same underwrites sharing ensembles agents, certain aspects (i.e., factors) each agent’s world provide common ground frame reference. Active plays foundational role ecology sharing—leading formal account collective rests shared narratives goals. We also consider kinds communication protocols must developed enable such intelligences motivate hyper-spatial modeling language transaction protocol, first—and key—step towards ecology.

Язык: Английский

---

Collective behavior from surprise minimization

Proceedings of the National Academy of Sciences, Год журнала: 2024, Номер 121(17)

Опубликована: Апрель 17, 2024

Collective motion is ubiquitous in nature; groups of animals, such as fish, birds, and ungulates appear to move a whole, exhibiting rich behavioral repertoire that ranges from directed movement milling disordered swarming. Typically, macroscopic patterns arise decentralized, local interactions among constituent components (e.g., individual fish school). Preeminent models this process describe individuals self-propelled particles, subject self-generated “social forces” short-range repulsion long-range attraction or alignment. However, organisms are not particles; they probabilistic decision-makers. Here, we introduce an approach modeling collective behavior based on active inference. This cognitive framework casts the consequence single imperative: minimize surprise. We demonstrate many empirically observed phenomena, including cohesion, milling, motion, emerge naturally when considering driven by Bayesian inference—without explicitly building rules goals into agents. Furthermore, show inference can recover generalize classical notion social forces agents attempt suppress prediction errors conflict with their expectations. By exploring parameter space belief-based model, reveal nontrivial relationships between beliefs group properties like polarization tendency visit different states. also explore how about uncertainty determine decision-making accuracy. Finally, update generative model over time, resulting collectively more sensitive external fluctuations encode information robustly.

Язык: Английский

---

Neural representation in active inference: Using generative models to interact with—and understand—the lived world

Annals of the New York Academy of Sciences, Год журнала: 2024, Номер 1534(1), С. 45 - 68

Опубликована: Март 25, 2024

Abstract This paper considers neural representation through the lens of active inference, a normative framework for understanding brain function. It delves into how living organisms employ generative models to minimize discrepancy between predictions and observations (as scored with variational free energy). The ensuing analysis suggests that learns navigate world adaptively, not (or solely) understand it. Different may possess an array models, spanning from those support action‐perception cycles underwrite planning imagination; namely, explicit entail variables predicting concurrent sensations, like objects, faces, or people—to action‐oriented predict action outcomes. then elucidates belief dynamics might link implications different types agent's cognitive capabilities in relation its ecological niche. concludes open questions regarding evolution development advanced abilities—and gradual transition pragmatic detached representations. on offer foregrounds diverse roles play processes representation.

Язык: Английский

---

Cognition as Morphological/Morphogenetic Embodied Computation In Vivo

Entropy, Год журнала: 2022, Номер 24(11), С. 1576 - 1576

Опубликована: Окт. 31, 2022

Cognition, historically considered uniquely human capacity, has been recently found to be the ability of all living organisms, from single cells and up. This study approaches cognition an info-computational stance, in which structures nature are seen as information, processes (information dynamics) computation, perspective a cognizing agent. Cognition is understood network concurrent morphological/morphogenetic computations unfolding result self-assembly, self-organization, autopoiesis physical, chemical, biological agents. The present-day human-centric view still prevailing major encyclopedias variety open problems. article considers recent research about morphological morphogenesis, agency, basal cognition, extended evolutionary synthesis, free energy principle, Bayesian learning, active inference, related topics, offering new theoretical practical perspectives on problems inherent old computationalist cognitive models were based abstract symbol processing, unaware actual physical constraints affordances embodiment A better understanding centrally important for future artificial intelligence, robotics, medicine, fields.

Язык: Английский

---

Feeling our place in the world: an active inference account of self-esteem

Neuroscience of Consciousness, Год журнала: 2024, Номер 2024(1)

Опубликована: Янв. 1, 2024

Self-esteem, the evaluation of one's own worth or value, is a critical aspect psychological well-being and mental health. In this paper, we propose an active inference account self-esteem, casting it as sociometer inferential capacity to interpret standing within social group. This approach allows us explore interaction between individual's self-perception expectations their environment.When there mismatch these perceptions expectations, individual needs adjust actions update better align with current experiences. We also consider hypothesis in relation recent research on affective inference, suggesting that self-esteem enables track respond discrepancy through states such anxiety positive affect. By acting sociometer, individuals navigate adapt environment, ultimately impacting

Язык: Английский

---

Federated inference and belief sharing

Neuroscience & Biobehavioral Reviews, Год журнала: 2023, Номер 156, С. 105500 - 105500

Опубликована: Дек. 5, 2023

This paper concerns the distributed intelligence or federated inference that emerges under belief-sharing among agents who share a common world-and world model. Imagine, for example, several animals keeping lookout predators. Their collective surveillance rests upon being able to communicate their beliefs-about what they see-among themselves. But, how is this possible? Here, we show all necessary components arise from minimising free energy. We use numerical studies simulate generation, acquisition and emergence of language in synthetic agents. Specifically, consider inference, learning selection as variational energy posterior (i.e., Bayesian) beliefs about states, parameters structure generative models, respectively. The theme-that attends these optimisation processes-is actions minimise expected energy, leading active model (a.k.a., learning). first illustrate role communication resolving uncertainty latent states partially observed world, on which have complementary perspectives. then requisite language-entailed by likelihood mapping an agent's overt expression (e.g., speech)-showing can be transmitted across generations learning. Finally, emergent property minimisation, when operate within same econiche. conclude with discussion various perspectives phenomena; ranging cultural niche construction, through learning, complexity ensembles self-organising systems.

Язык: Английский

---

Narrative as active inference

Опубликована: Июль 9, 2020

The ubiquity and importance of narratives in human adaptation has been recognized by many scholars. Research identified several functions that are conducive to individuals’ well-being as well coordinated social practices enculturation. In this paper, we characterize the cognitive terms framework active inference. Active inference depicts fundamental tendency living organisms adapt creating, updating, maintaining inferences about their environment. We review literature on identity, event segmentation, episodic memory, future projection, storytelling practices, then re-cast these inference, outlining a parsimonious model can guide developments narrative theory, research, clinical applications.

Язык: Английский

---

Spin Glass Systems as Collective Active Inference

Communications in computer and information science, Год журнала: 2023, Номер unknown, С. 75 - 98

Опубликована: Янв. 1, 2023

Язык: Английский

---

Collective Intelligence in Human-AI Teams: A Bayesian Theory of Mind Approach

Proceedings of the AAAI Conference on Artificial Intelligence, Год журнала: 2023, Номер 37(5), С. 6119 - 6127

Опубликована: Июнь 26, 2023

We develop a network of Bayesian agents that collectively model the mental states teammates from observed communication. Using generative computational approach to cognition, we make two contributions. First, show our agent could generate interventions improve collective intelligence human-AI team beyond what humans alone would achieve. Second, real-time measure human's theory mind ability and test theories about human cognition. use data collected an online experiment in which 145 individuals 29 human-only teams five communicate through chat-based system solve cognitive task. find (a) struggle fully integrate information into their decisions, especially when communication load is high, (b) have biases lead them underweight certain useful, but ambiguous, information. Our predicts both individual- team-level performance. Observing teams' first 25% messages explains 8% variation final performance, 170% improvement compared current state art.

Язык: Английский

---

Sustainability under Active Inference

Опубликована: Май 2, 2024

In this paper we explore the known connection among sustainability, resilience, and well-being within framework of active inference. Initially, revisit how notions resilience intersect inference before defining sustainability. We adopt a holistic concept sustainability denoting enduring capacity to meet needs over time without depleting crucial resources. It extends beyond material wealth encompass community networks, labor, knowledge. Using Free Energy Principle, can emphasize role fostering resource renewal, harmonious system-entity exchanges, practices that encourage self-organization as pathways achieving both in an agent collectives. start by connecting Active Inference with well-being, building on exsiting work. then attempt link asserting alone is insufficient for sustainable outcomes. While absorbing shocks stresses, must be intrinsically linked ensure adaptive capacities do not merely perpetuate existing vulnerabilities. Rather, it should facilitate transformative processes address root causes unsustainability. Sustainability, therefore, manifest across extended timescales all system strata, from individual components broader system, uphold ecological integrity, economic stability, social well-being. explain manifests at level agent, collectives systems. To model quantify interdependencies between resources their impact overall introduce application network theory dynamical systems theory. optimization precision or learning rates through framework, advocating approach fosters elastic plastic necessary long-term abundance.

Язык: Английский

---