# Closed-Loop Gaussian Sensorium Engine **Entity class:** Proposed integrative neurotechnology architecture **Domain:** Brain-Computer Interfaces / Perception / Neurotechnology **Doc Type:** Canonical Architecture Node **Maturity:** Components established; integration unresolved **Primary source:** [[articles/The Closed-Loop Gaussian Sensorium Engine|The Closed-Loop Gaussian Sensorium Engine]] **Collection:** [[collections/Neurotech|Neurotech]] --- ## Definition The **Closed-Loop Gaussian Sensorium Engine** is Bryant McGill's name for a proposed bidirectional perceptual interface that couples neural recording, personalized perceptual mapping, machine forecasting, differentiable rendering, inverse stimulation encoding, and biological predictive completion in one adaptive loop. The architecture does not propose painting a complete image into cortex. It proposes sending sparse, calibrated perceptual cues that the nervous system can complete through its own generative dynamics. [[wiki/Perceptual Attractor Seeding|Perceptual Attractor Seeding]] names that control principle; [[wiki/Gaussian Perceptual Field|Gaussian Perceptual Field]] names the proposed machine-side representation. ## Epistemic status - **Established:** neural recording, cortical stimulation, retinotopic mapping, cross-subject fMRI decoding, differentiable phosphene simulation, real-time [[wiki/3D Gaussian Splatting|3D Gaussian Splatting]], and closed-loop adaptation each exist as separate research lines. - **Strongly indicated:** those lines share compatible read, model, render, stimulation, and feedback functions that can be described as one systems architecture. - **Plausible:** sparse machine-selected cues could reduce the stimulation burden by recruiting biological perceptual completion. - **Unresolved:** no cited source demonstrates the complete human Engine as an integrated artifact. Its achievable bandwidth, subjective fidelity, chronic stability, consent guarantees, and first jurisdiction of deployment remain open questions. ## Architecture | Layer | Function | Canonical route | |---|---|---| | Personalized perceptual atlas | Registers an individual's retinotopy, receptive fields, phosphene locations, thresholds, persistence, and uncertainty | [[wiki/Neural Signal Acquisition\|Neural Signal Acquisition]] | | Read path | Acquires spatially and temporally resolved neural activity | [[wiki/Electrocorticography\|Electrocorticography]] · [[wiki/Utah Array\|Utah Array]] | | Model core | Decodes present state and forecasts candidate future neural states | [[wiki/neural foundation models\|neural foundation models]] · [[wiki/Neural Decoding\|Neural Decoding]] | | Perceptual field | Maintains retinotopic kernels and higher-order object-axis representations through time | [[wiki/Gaussian Perceptual Field\|Gaussian Perceptual Field]] | | Renderer | Produces an inspectable scene estimate and a differentiable forward model | [[wiki/3D Gaussian Splatting\|3D Gaussian Splatting]] | | Inverse encoder | Selects stimulation parameters predicted to move the subject toward a target perceptual state | [[wiki/Perceptual Attractor Seeding\|Perceptual Attractor Seeding]] | | Write path | Delivers content or state modulation through cortical, retinal, ultrasonic, magnetic, or electrical channels | [[wiki/Intracortical Visual Prosthesis Project\|Intracortical Visual Prosthesis Project]] · [[wiki/Optogenetics and Chemogenetics\|Optogenetics and Chemogenetics]] · [[wiki/Functional Ultrasound\|Functional Ultrasound]] | | Loop closure | Updates the model from the subject's response and selects the next intervention | [[wiki/Closed-Loop BCI\|Closed-Loop BCI]] · [[wiki/Active Inference\|Active Inference]] | | Governance boundary | Determines who may write, audit, refuse, stop, retain, or reuse perceptual interventions | [[wiki/Perceptual Sovereignty\|Perceptual Sovereignty]] | ## Entity and program coverage The source article's institutional and technical inventory is preserved here as a structured map rather than expanded into shallow duplicate pages. - **Reference maps and datasets:** Allen Institute for Brain Science; Howard Hughes Medical Institute and Janelia Research Campus; FlyEM; FlyLight; International Brain Laboratory; BRAIN Initiative. - **Implanted-interface lineage:** Stanford NPSL/NPTL; Krishna Shenoy, Jaimie Henderson, Frank Willett, Sergey Stavisky, Chethan Pandarinath, and Leigh Hochberg; BrainGate; University of Utah; Utah Array; Blackrock Neurotech. - **Strategic programs:** DARPA N3, HAPTIX, and MOANA; BrainCo; ReVision Implant Occular; Neuralink Blindsight; i-BRAIN and the Shenzhen Medical Academy of Research and Translation. - **Model and reconstruction anchors:** MindEye2; OmniMouse; the 2026 Wadia–Rutishauser–Tsao ventral-temporal axis-code study; differentiable phosphene simulation; dynamic visual reconstruction. - **Rendering and control:** 3D Gaussian Splatting, dynamic 4D splatting, AudioGS, inverse encoding, patterned intracortical microstimulation, retinal optogenetics, focused ultrasound, transcranial magnetic stimulation, transcranial electrical stimulation, and active inference. - **Places and capability corridors:** Singapore, Hong Kong, Shenzhen, Hangzhou, Guangming Science City, and the wider Pacific Rim corridor described in the article. - **Commercial and graphics adjacency:** WebGL, WebGPU, Three.js, augmented reality, virtual reality, urban digital twins, Lancôme, L'Oréal, and Nike are treated as adjacent skill and deployment surfaces rather than members of a neural-interface program. - **Observer-stack lineage:** CERN Level-1 triggers, ROOT, BioDynaMo, FLUKA, the LHC Computing Grid, the ETH Zürich Darwin bioinformatics system, CERN protoDarwin, NeXT, and Darwin OS supply the cross-domain compression and reconstruction lineage. - **Clinical and phenomenological boundary:** ALS, locked-in syndrome, complete locked-in syndrome, minimally conscious state, cognitive-motor dissociation, covert awareness, derealization, lucid dreaming, solipsism, and the cultural NPC intuition are preserved as distinct clinical, experiential, philosophical, and cultural referents. - **Institutional compute and funding:** Simons Collaboration on the Global Brain, Chan Zuckerberg Biohub, Argonne Aurora, the National Center for Supercomputing Applications, Fermilab DUNE, NIH, NSF, and DARPA occupy different funding, compute, and research roles. - **China-gradient entities:** Charles Lieber; Wuhan University of Technology; the former China Initiative; Shenzhen Bay Laboratory; Brain Science Infrastructure Shenzhen; ASML deep-ultraviolet lithography; John Donoghue; and Zheng Shanjie are recorded as people, institutions, infrastructure, and policy actors rather than collapsed into one organization. - **Personalized decoding and calibration:** fMRI population receptive-field mapping, Difference-of-Gaussian models, psychophysics, eye tracking, phosphene mapping, Natural Scenes Dataset, CLIP, Stable Diffusion XL, ECoG2IMG, Talairach coordinates, Neuropixels-class probes, and flexible mesh or threaded electrodes occupy separate stages of the read and registration path. - **Visual prosthesis lineages:** Illinois ICVP, the Fernández program, ReVision Occular, Neuralink Blindsight, Beauchamp–Yoshor patterned stimulation, Roelfsema's V1/V4 work, MCO-010, and retinitis-pigmentosa optogenetics are component evidence with different subjects, modalities, and maturity. - **Biographical and interpretive routes:** Stephen Hawking, Paul G. Allen, Krishna Shenoy, and Steve Jobs anchor the article's tributes; Edge.org, Martin Nowak, George Church, Humanity+, Ben Goertzel, and Team Leela belong to the separate disclosure and intellectual-network discussion in Recommended Reading. These names identify components, institutions, and evidence anchors. Their inclusion does not establish a coordinated program, common ownership, technology transfer, or a completed Engine. ## Relationships - **Extends:** [[wiki/Observer Stack|Observer Stack]] from acquisition and reconstruction into write-back and adaptive control. - **Specializes:** [[wiki/Closed-Loop BCI|Closed-Loop BCI]] for perceptual rather than solely motor or communication loops. - **Uses as an analytic representation:** [[wiki/Gaussian Perceptual Field|Gaussian Perceptual Field]]. - **Uses as a proposed intervention rule:** [[wiki/Perceptual Attractor Seeding|Perceptual Attractor Seeding]]. - **Receives its clinical and philosophical frame from:** [[wiki/Locked-In Syndrome|Locked-In Syndrome]] and [[wiki/Locked-In Inversion|Locked-In Inversion]]. - **Creates a governance requirement for:** [[wiki/Perceptual Sovereignty|Perceptual Sovereignty]], [[wiki/Cognitive Liberty|Cognitive Liberty]], [[wiki/Neurorights|Neurorights]], and [[wiki/Neural Data Sovereignty|Neural Data Sovereignty]]. ## Sources and provenance - [[articles/The Closed-Loop Gaussian Sensorium Engine|The Closed-Loop Gaussian Sensorium Engine]] — canonical source for the integration thesis, architecture, institutional lattice, and moral frame. - [Kerbl et al., “3D Gaussian Splatting for Real-Time Radiance Field Rendering” (SIGGRAPH 2023)](https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/) — primary rendering source. - [Scotti et al., “MindEye2: Shared-Subject Models Enable fMRI-To-Image With 1 Hour of Data” (ICML 2024)](https://proceedings.mlr.press/v235/scotti24a.html) — primary cross-subject decoding source. - [van der Grinten et al., “Towards biologically plausible phosphene simulation for the differentiable optimization of visual cortical prostheses” (eLife, 2024)](https://elifesciences.org/articles/85812) — primary differentiable phosphene-simulation source. - [Wadia et al., “A shared code for perceiving and imagining objects in human ventral temporal cortex” (Science, 2026)](https://pubmed.ncbi.nlm.nih.gov/41955351/) — primary shared-code source. - [Willeke et al., “OmniMouse” (ICLR 2026)](https://openreview.net/forum?id=mEw4lhAn0F) — primary neural-foundation-model source.