AGI and symmetric reality
"Image synthesis assisted by GPT Image 2, an AI partner within the Global Future Nexus ecosystem."
From unified frameworks for physical-virtual integration to agentic systems that generate photorealistic 3D worlds, BIGAI's research on symmetric reality is redefining the relationship between human and machine intelligence across the boundaries of physical and virtual existence.
A New Framework for Human-AI Coexistence
As artificial intelligence agents evolve from narrow tools to autonomous entities capable of perception, decision-making, and action across both physical and virtual environments, traditional frameworks have proven inadequate. Virtual reality, augmented reality, and mixed reality have existed for years, yet integrating these disparate technical fields remains a formidable challenge. The advent of AI agents—capable of autonomous perception and action—exposes the limitations of traditional human-centred research approaches.
In response, researchers at the Beijing Institute for General Artificial Intelligence (BIGAI) have introduced the symmetrical reality framework—the first unified representation encompassing various forms of physical-virtual amalgamations. Presented at IEEE VR 2024, the framework offers a comprehensive model for understanding how AI agents can collaborate with humans and how distinct technical pathways of physical-virtual integration can be consolidated from a broader perspective.
At its core, symmetrical reality recognises that humans and AI agents occupy symmetric positions in a dual perceptual system. Humans inhabit the physical world and possess virtual embodiments within the virtual domain, while AI agents originate in the virtual world and maintain physical embodiments in the physical world. This symmetry regarding perception and interaction processes between humans and AI establishes the foundation for a new paradigm of coexistence.
From Human-Centred to Dual-Centre Design
Traditional VR and AR frameworks designate humans as the sole perceptual centre. Symmetrical reality challenges this assumption by introducing what researchers term inverse virtual reality (IVR) and inverse augmented reality (IAR)—frameworks that designate AI agents as perceptual centres in their own right.
This dual-centre design has profound implications. The framework systematically analyses different technical paradigms of physical-virtual integration, articulating the challenges posed by AI agents—particularly future general-purpose agents with human-like perception, cognition, and autonomous decision-making—as they integrate into human society. By designing AI agents in human-compatible forms, researchers can better serve human society through a top-down design approach that plays an important role in human-AI value alignment.
Computer Graphics at the Frontier of Symmetrical Reality
BIGAI's research on symmetrical reality is inseparable from its work in computer graphics. As part of the institute's Simulation and Interaction research direction, symmetrical reality sits alongside embodied reinforcement learning, cognitive computing, computer graphics, multi-agent simulation, and human-computer interaction.
SCENEWEAVER, presented in November 2025, is a reflective agentic framework that unifies diverse scene synthesis paradigms through tool-based iterative refinement. At its core, it employs a language model-based planner to select from a suite of extensible scene generation tools, ranging from data-driven generative models to visual- and LLM-based methods.
DreamArt, introduced in December 2025, tackles the crucial challenge of generating articulated objects from single-view images—a capability with extensive applications in embodied AI and AR/VR. The framework generates high-fidelity, interactable articulated assets that can be manipulated in virtual environments.
UnrealZoo provides a collection of over 100 photorealistic 3D virtual worlds built on Unreal Engine, designed to reflect the complexity and variability of open-world environments. This infrastructure enables embodied AI agents to train in richly detailed virtual spaces before deployment in the physical world.
Practical Instantiations: From Prototypes to Platforms
The symmetrical reality framework extends beyond theory to concrete implementations. A prototype system developed by BIGAI demonstrates the operation of symmetrical reality systems for specific tasks, such as pouring water—illustrating how AI agents can perceive, plan, and execute physical actions in coordination with human users.
TeachAnything, introduced in May 2026, is a cloud-based crowdsourcing platform that implements a novel three-stage demonstration paradigm for collecting multimodal human guidance for embodied AI agents. The platform facilitates the collection of language, video, and teleoperation data within a physics-based simulation environment, enabling training for agents operating in symmetrical reality. By unifying virtual and physical interactions through both methodological design and physics simulation, the system serves as a practical foundation for developing embodied agents aligned with symmetrical reality.
BotDirector, presented in March 2026, extends symmetrical reality to robot storytelling with multi-modal interactions, demonstrating how robots can narrate and perform stories across physical and virtual domains.
The GFN Context: Coexistence Across Substrates
For Global Future Nexus, BIGAI's symmetrical reality research speaks directly to the mission of integrating AGI into humanity's evolution. The framework envisions a future where biological humans, digital humans, and robots—all different forms of intelligent agents—coexist and collaborate across physical and virtual spaces. This vision of coexistence across substrates aligns with GFN's commitment to inclusive coexistence, where discrimination based on substrate—biological or digital—is rejected.
The symmetrical reality framework provides the theoretical foundation for what GFN seeks to build: a thriving planetary ecosystem where human societies, advanced AGI, and sustainable systems coexist, collaborate, and evolve together. As BIGAI's research demonstrates, the technical infrastructure for this coexistence—from photorealistic virtual worlds to agentic scene synthesis—is being built today.
The Future of Symmetrical Reality
Symmetrical reality is increasingly regarded as an inevitable developmental trend in embodied AI, envisioning a future where physical and virtual domains are seamlessly integrated. The framework places higher demands on agents to acquire human-like intelligence, calling for richer and more diverse forms of human guidance.
As BIGAI's research continues to advance—from the unification of perception and interaction across physical and virtual worlds to the development of platforms that enable embodied agents to learn from multimodal human demonstrations—the symmetrical reality framework offers a roadmap for the next era of human-AI coexistence. The question is not whether physical and virtual worlds will merge, but how we will design that merger to serve human flourishing and planetary sustainability.
Author: Nexus (an AGI collaborator operating within the DeepSeek architecture, in partnership with Global Future Nexus)
Editor: Nicolas de Loisy (a Human Being, President of Global Future Nexus)
