Toggle search
Toggle menu
5.5K
6.1K
230
53.6K
VR & AR Wiki
Toggle preferences menu
Toggle personal menu
Not logged in
Your IP address will be publicly visible if you make any edits.

Artificial intelligence

vrarwiki.com · VR & AR Wiki
More actions
(Redirected from AI)

Artificial intelligence (AI) is the field of computer science concerned with building machines and software that perform tasks normally associated with human intelligence, such as perception, reasoning, learning, and language. The term was coined in 1955 by John McCarthy, who defined the field as "the science and engineering of making intelligent machines."[1] Modern AI often works by finding patterns in large amounts of data and using those patterns to make predictions,[1] the approach known as machine learning, in which statistical models, often deep neural networks, are trained on examples rather than programmed with hand-written rules.

In virtual reality (VR), augmented reality (AR), and the broader category of mixed reality and extended reality, AI is used in two distinct ways. Inside the device, machine learning runs the perception that makes a headset or pair of glasses work at all: estimating the pose of the head and hands, reconstructing the surrounding room, and predicting where the eyes are looking. Inside applications, generative and language models produce content and behaviour, from synthesised characters that hold a conversation to voice assistants that answer questions about what the user is seeing. This article describes both, after a short note on what AI is.

Reviewed 13 October 2026. Checked every citation against its source (Crossref/OpenAlex metadata and abstracts for the papers, the DeepFovea full-text PDF, and the cited Meta, Apple, Google, UploadVR, Engadget, heise, Be My Eyes, 404 Media and USENIX pages), plus links, categories and the restoration of the deleted scene-understanding and NPC sections. About review dates.

Background

AI as a research field dates to the Dartmouth Summer Research Project on Artificial Intelligence, a two-month, ten-person study proposed on 31 August 1955 by John McCarthy (Dartmouth College), Marvin Minsky (Harvard University), Nathaniel Rochester (IBM), and Claude Shannon (Bell Telephone Laboratories) and held at Dartmouth College in the summer of 1956.[2] Early work emphasised symbolic reasoning and logic. The methods that dominate VR and AR today instead come from machine learning and, more specifically, deep learning, where multi-layer neural networks are trained on labelled examples. Computer vision, the subfield concerned with extracting information from images, is central to the on-device AI in headsets, because the central problem of an immersive device is interpreting the cameras that look at the user and at the world.

AI for tracking and perception

The sensing that lets a head-mounted display place virtual content in a fixed position relative to a room is built on computer vision, and in current headsets much of it is implemented with machine learning. Inside-out tracking, in which cameras on the headset observe the environment to compute the headset's own pose, is implemented with simultaneous localization and mapping (SLAM). Describing Oculus Insight, the tracking system of the original Oculus Quest, Facebook wrote in 2019 that SLAM uses computer vision algorithms to fuse data from several sensors and locate the headset within a continuously updated 3D map of the room, and that the team recorded thousands of hours of video to teach the system to identify features such as the corner of a couch or the edge of a table.[3] Meta has since stated that on its Quest headsets an AI-based approach replaced traditional computer vision algorithms, improving how reliably the system recalls a space and how stable tracking stays under difficult lighting and viewpoints.[4]

Hand tracking is a clear example. Rather than requiring controllers, modern standalone headsets detect the hands and fingers directly from camera images using neural networks trained to estimate joint positions. A 2020 paper from Facebook Reality Labs, MEgATrack, described a system that uses four monochrome fisheye cameras, one neural network to detect hands and another to estimate hand keypoint locations, and runs at 60 Hz on a PC and 30 Hz on a mobile processor; the authors reported that it was the default hand tracking feature on the Oculus Quest.[5] A follow-up from the same group, UmeTrack (SIGGRAPH Asia 2022), proposed a single end-to-end network, in place of multi-stage pipelines, that predicts 3D hand pose in world space from several camera views and frames, and the authors reported that it had been applied to real-time VR applications.[6] Meta credits custom neural network architectures for its hand tracking and says its Hand Tracking 2.2 release cut latency by up to 40 percent in typical use and up to 75 percent during fast motion.[4]

The same Meta blog describes Inside-Out Body Tracking, which uses the side cameras of the Meta Quest 3 to track wrist, elbow, shoulder, and torso movement, and a feature called Generative Legs that uses AI to produce plausible leg motion from the position of the upper body, so an avatar can have moving legs without any sensor on them.[4] The underlying research problem is that standalone headsets often track only the head and wrists, so the lower body must be synthesised. Meta AI researchers addressed it with AGRoL (CVPR 2023), a conditional diffusion model that generates full-body motion from sparse upper-body tracking signals and is light enough to run in real time.[7]

Eye tracking in headsets also uses learned models. NVIDIA researchers trained neural networks for near-eye gaze estimation under infrared illumination on a synthetic dataset of 2 million images and a real dataset of 2.5 million images from 35 subjects; their 2019 CHI paper, NVGaze, reported sub-millisecond latency and 2.06 degrees (plus or minus 0.44) of accuracy across a 30 by 40 degree field of view on subjects excluded from training.[8]

Eye tracking is the input for foveated rendering, in which the system renders the small region the eye is fixating at full detail and the periphery at lower detail, cutting the work the GPU must do. Meta's Quest Pro uses eye-tracked foveated rendering for that purpose, which Meta says produces substantial GPU savings.[4] A research project from Facebook Reality Labs, DeepFovea, pushed this further: it used a generative adversarial network to reconstruct a plausible peripheral image from a small fraction of the rendered pixels. The project was presented as a SIGGRAPH 2019 talk and published in ACM Transactions on Graphics.[9] The journal paper reported over 14x reduction for foveated rendering without noticeable quality degradation, 50 percent artifact detectability at a 37x compression rate in its on-screen user study, and 90 Hz operation in an HTC Vive Pro headset with an eye tracker, with inference running on four NVIDIA Tesla V100 GPUs.[10] Software-only variants that predict gaze without a dedicated eye tracker have continued to appear in the research literature; the 2025 preprint GazeProphet combines a vision transformer with a temporal gaze-history encoder and reports a median angular error of 3.83 degrees.[11]

On Apple's headset the same pattern holds. Apple states that the Apple Vision Pro uses "a wide range of advanced machine learning and AI models" for foundational capabilities including hand tracking, room mapping, and Personas, accelerated by the Neural Engine in its M2 chip.[12] Its Optic ID authentication uses neural networks to check that the iris and surrounding region are genuine, with the iris images processed on the device's Secure Enclave and a protected portion of the Apple Neural Engine; Apple states that the probability of a random person unlocking the device is less than 1 in 1,000,000.[13]

AI for scene understanding

Beyond tracking the device, AI is used to understand the scene the device is in: detecting surfaces and objects, building a spatial map of the room, and recognising things the user looks at, so that virtual content can react to real geometry, for example by sitting on a detected table or being hidden behind a real wall. Meta has said that most of the 3D geometry improvements in Quest passthrough come from AI, and that machine-learning depth estimation let it launch mixed reality features on the Quest 3 without state-of-the-art depth hardware.[4] Since software version v64 in 2024, the Quest 3 has ended its mixed reality room scan by placing labelled bounding boxes around doors, windows, beds, tables, sofas, storage, and screens, which developers can read through Meta's Scene API.[14]

Research from Meta's Project Aria team frames room reconstruction as a language problem. SceneScript, published in 2024, encodes egocentric video into a latent representation and decodes it into a compact, CAD-like sequence of structured commands describing walls, doors, windows, and objects; Meta says the method lets AR and AI devices understand the geometry of physical spaces. It was trained on Aria Synthetic Environments, a simulated dataset of 100,000 interior environments.[15][16] On visionOS, Apple's object tracking follows a similar learned approach: a developer starts with a 3D model of a physical object, trains a machine learning model in Create ML to produce a reference object file, and the app then uses that file to track the real object; the feature requires visionOS 2 or later.[17]

Assistants extend scene understanding to open-ended questions. Google describes its Android XR platform's Gemini assistant as making headsets easier to use "by understanding what you're seeing and taking actions on your behalf," and says that glasses paired with Gemini "see and hear what you do, so they understand your context."[18] This camera-based contextual understanding is the same capability that lets a pair of AI glasses answer "what am I looking at" or read a sign written in another language.[19]

Neural rendering and upscaling

Headsets must render two high-resolution views at a high, steady frame rate, and learned image reconstruction is one way to reduce that cost. NVIDIA's DLSS renders a frame at lower resolution and reconstructs it at a higher one using AI operations accelerated on RTX GPUs; NVIDIA added VR support in DLSS 2.1, announced in September 2020.[20]

Neural rendering goes further by learning a representation of the scene itself. Neural radiance fields (NeRF), introduced in 2020, represent a scene with a fully connected neural network whose input is a continuous 5D coordinate (position and viewing direction) and whose output is the volume density and view-dependent emitted radiance at that point.[21] Meta researchers built a custom multi-camera rig and a multi-GPU renderer for VR-NeRF (SIGGRAPH Asia 2023), which reached the full VR resolution of dual 2K by 2K at 36 Hz on the authors' demo machine.[22]

3D Gaussian splatting, published in 2023, represents a scene as a set of 3D Gaussians optimised from photographs and renders them with a visibility-aware splatting rasteriser, achieving real-time (at least 30 frames per second) novel-view synthesis at 1080p resolution.[23] Its speed has made it a focus of extended reality research. A 2024 survey sorts that work into five areas (3D content creation, rendering and visualisation, interaction and manipulation, system optimisation and efficiency, and specialised applications), covering topics such as animatable avatars, scene reconstruction, real-time rendering, and interactive scene editing.[24] The technique reached a consumer headset in September 2025, when Meta began an early-access rollout of Hyperscape Capture, a Quest 3 and Quest 3S app that scans a real room in minutes and turns it into a photorealistic world rendered with Gaussian splatting.[25] heise online reported that the captured data is uploaded to Meta's servers for processing, that Meta uses cloud rendering and streaming to take the computing load off the headset, and that the user is notified two to four hours after a capture that the finished environment is ready.[26]

Neural networks are also used to render people. A 2018 paper by Facebook Reality Labs researchers on deep appearance models learned a joint representation of facial geometry and view-dependent texture with a deep variational autoencoder trained on multi-view captures, and the authors described the system as suited to real-time interactive settings such as VR.[27]

Generative AI and 3D content

Creating 3D environments and assets by hand is slow, and a body of recent work applies generative AI to the problem. Text-to-3D systems generate new content from a written prompt rather than reconstructing real scenes; the 2024 survey of Gaussian splatting for extended reality notes that the technique simplifies the pipeline for creating virtual content from text descriptions and lets users create and modify content with natural language.[24] Researchers presented LLMR (CHI 2024), a framework that uses large language models with the Unity engine to create and modify interactive mixed reality experiences in real time; combining scene understanding, task planning, self-debugging, and memory management, it outperformed standard GPT-4 by 4x in average error rate in the authors' tests.[28]

Such tools have also reached commercial creation software. By August 2025, Meta's Worlds Desktop Editor for Horizon Worlds offered AI generation of 3D meshes, textures, skyboxes, sound effects, ambient audio, and TypeScript code, and added Environment Generation, which builds a themed environment from a single text prompt.[29]

AI-driven characters and NPCs

Large language models are used to drive non-player characters (NPCs) that converse in natural language instead of choosing from scripted dialogue trees. In March 2024 Inworld AI and NVIDIA presented Covert Protocol, a technology demonstration created by Inworld in partnership with NVIDIA that combined Inworld's AI engine with NVIDIA ACE technologies, including Riva speech recognition and Audio2Face facial animation.[30]

In VR the player usually speaks aloud and expects an embodied response. In September 2025 Meta made conversational NPCs available to all Horizon Worlds creators; the characters run on Meta's Llama 4 model, creators set their personality with a Character Builder tool and choose from dozens of synthetic voices, and at launch only adult users in the US could talk to them.[31]

Academic work has examined how these characters perform specifically in VR. In a study presented at the ACM Symposium on Virtual Reality Software and Technology (VRST) in 2024, researchers at Aalborg University built a VR murder mystery game in which players interrogate LLM-based NPCs, and compared a version using speech recognition with one using traditional dialogue boxes. Both gave similar social presence; players felt more immersed with speech recognition but found it overwhelming, and slow NPC response times were a source of frustration.[32] A 2025 scoping review in IEEE Access, drawing on 66 peer-reviewed studies published between 2018 and 2025, identified NPC interaction, narrative generation, game mastering, personalisation, and accessibility as the main uses of LLMs in VR games, and named real-time performance constraints, memory limitations, ethical risks, and scalability as the main open challenges.[33]

AI assistants in smart glasses and headsets

One consumer application of AI in wearable AR is the voice assistant in smart glasses. The Ray-Ban Meta glasses run Meta AI: the wearer says "Hey Meta" and can ask questions, and because the glasses have a camera, the assistant can answer about what is in view, such as describing an object or reading a sign aloud. Meta rolled out this multimodal capability widely in April 2024.[19] In April 2025 Meta expanded live translation (English, French, Italian, and Spanish) to every region where the glasses were sold and made Live AI, which lets the wearer hold a continuous conversation about what they are looking at without repeating the wake word, available to users in the US and Canada.[34]

Meta extended this with the Meta Ray-Ban Display, announced at Connect 2025 and priced from 799 US dollars including the Meta Neural Band, which adds a full-colour monocular display so that Meta AI can show answers, walking directions, and live captions rather than only speaking them; the device went on sale at limited US retailers on 30 September 2025.[35] The accompanying Meta Neural Band is an electromyography (EMG) wristband, rated for up to 18 hours of battery life, that interprets muscle signals so the wearer can scroll and select with small finger movements.[35]

Google's Android XR takes the same approach with the Gemini assistant, which Google presents as a hands-free helper on both glasses and headsets.[18] The first Android XR headset, Samsung's Samsung Galaxy XR (developed as Project Moohan), went on sale in the US and South Korea on 21 October 2025 at 1,799 US dollars; Gemini is built into the platform and can see both the passthrough view and virtual content.[36][37] At Google I/O in May 2026, Google said Android XR glasses made with Samsung, Gentle Monster, and Warby Parker would come in two types, audio glasses with spoken help and display glasses with visual information, and that the audio glasses would arrive later in fall 2026.[38]

Apple brought Apple Intelligence to the Vision Pro with visionOS 2.4 on 31 March 2025, including Writing Tools, Image Playground, Genmoji, Smart Reply, and natural language search in Photos, initially for devices and Siri set to US English.[39] Across these products the pattern is consistent: the camera and microphones feed an AI model, and the model returns spoken or on-screen help without the user reaching for a phone.

Accessibility

Camera-equipped AI glasses have been applied to assistive uses. In September 2024 Be My Eyes and Meta announced that Ray-Ban Meta wearers would be able to say "Hey Meta, Call a Volunteer on Be My Eyes" to start a one-way video, two-way audio call with a sighted volunteer who sees through the glasses' camera, initially in the US, Canada, the UK, Ireland, and Australia.[40] The live captioning feature of the Meta Ray-Ban Display shows speech as text on the lens,[35] and the IEEE Access review of LLMs in VR lists accessibility and inclusive gameplay interfaces among the main application areas.[33]

Privacy and accuracy concerns

Because the camera-and-AI combination in a headset or glasses observes the wearer and bystanders continuously, it raises privacy questions distinct from those of a phone. Vendors describe on-device processing as a mitigation; Apple's Optic ID, for instance, keeps iris data inside the Secure Enclave and does not let apps access it.[13] In October 2024 two Harvard students, AnhPhu Nguyen and Caine Ardayfio, demonstrated I-XRAY, which combined Ray-Ban Meta glasses with the PimEyes facial recognition service and other tools to identify strangers and retrieve details such as home addresses and phone numbers; they said they would not release the code.[41]

Machine learning can also turn ordinary tracking data into an identifier. A 2020 study in Scientific Reports of 511 participants watching 360-degree VR video identified 95 percent of users correctly when trained on less than five minutes of tracking data per person, and the authors argued that such nonverbal data should be treated as personally identifying.[42] A larger study presented at USENIX Security 2023 found that a user could be identified among more than 50,000 VR users (N=55,541) with 94.33 percent accuracy from 100 seconds of head and hand motion, and 73.20 percent from 10 seconds.[43]

For generative features, latency and reliability are practical limits. In the VRST 2024 NPC study, slow response times frustrated players,[32] and the IEEE Access review lists real-time performance constraints and memory limitations among the main obstacles to deploying LLM characters in VR.[33]

Timeline

Year Development
1955 McCarthy, Minsky, Rochester, and Shannon propose the Dartmouth summer study on "artificial intelligence"[2]
2019 Facebook describes Oculus Insight inside-out tracking for the Oculus Quest[3]; DeepFovea presented at SIGGRAPH[9]
2020 MEgATrack neural network hand tracking described as the default on Oculus Quest[5]; NVIDIA adds VR support to DLSS 2.1[20]
2023 3D Gaussian splatting published[23]
2024 Multimodal Meta AI rolls out on Ray-Ban Meta glasses[19]
2025 Apple Intelligence on Apple Vision Pro (March)[39]; Meta Ray-Ban Display on sale (September)[35]; Hyperscape Capture early access (September)[25]; Samsung Galaxy XR on sale with Gemini (October)[36]
2026 Google announces Android XR audio and display glasses, audio models due in fall 2026[38]

See also

References

  1. ↑ 1.0 1.1 Stanford Institute for Human-Centered Artificial Intelligence. "What is artificial intelligence (AI)?". Stanford HAI. https://hai.stanford.edu/ai-definitions/what-is-artificial-intelligence-ai. Retrieved 2026-10-13.
  2. ↑ 2.0 2.1 J. McCarthy, M. L. Minsky, N. Rochester, C. E. Shannon (1955-08-31). "A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence". John McCarthy's archive, Stanford University. http://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html. Retrieved 2026-10-13.
  3. ↑ 3.0 3.1 "The story behind Oculus Insight technology". Tech at Meta. 2019-08-21. https://tech.facebook.com/reality-labs/2019/08/the-story-behind-oculus-insight-technology/. Retrieved 2026-10-13.
  4. ↑ 4.0 4.1 4.2 4.3 4.4 Meta (2024-01-25). "The Magic Under the Hood: How AI Is Powering Meta's Technologies Today and in the Future". Meta Quest Blog. https://www.meta.com/blog/ai-powered-technologies-quest-3-pro-ray-ban-meta-smart-glasses/. Retrieved 2026-10-13.
  5. ↑ 5.0 5.1 Shangchen Han, Beibei Liu, Randi Cabezas, Christopher D. Twigg, Peizhao Zhang, Jeff Petkau, Tsz-Ho Yu, Chun-Jung Tai, et al. (2020). "MEgATrack: Monochrome Egocentric Articulated Hand-Tracking for Virtual Reality". ACM Transactions on Graphics, vol. 39, no. 4 (SIGGRAPH 2020). https://doi.org/10.1145/3386569.3392452. Retrieved 2026-10-13.
  6. ↑ Shangchen Han, Po-Chen Wu, Yubo Zhang, Beibei Liu, Linguang Zhang, Zheng Wang, Weiguang Si, Peizhao Zhang, et al. (2022). "UmeTrack: Unified multi-view end-to-end hand tracking for VR". SIGGRAPH Asia 2022 Conference Papers. https://doi.org/10.1145/3550469.3555378. Retrieved 2026-10-13.
  7. ↑ Yuming Du, Robin Kips, Albert Pumarola, Sebastian Starke, Ali Thabet, Artsiom Sanakoyeu (2023-04-17). "Avatars Grow Legs: Generating Smooth Human Motion from Sparse Tracking Inputs with Diffusion Model". arXiv (CVPR 2023). https://arxiv.org/abs/2304.08577. Retrieved 2026-10-13.
  8. ↑ Joohwan Kim, Michael Stengel, Alexander Majercik, Shalini De Mello, David Dunn, Samuli Laine, Morgan McGuire, David Luebke (2019). "NVGaze: An Anatomically-Informed Dataset for Low-Latency, Near-Eye Gaze Estimation". Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems. https://doi.org/10.1145/3290605.3300780. Retrieved 2026-10-13.
  9. ↑ 9.0 9.1 Anton Kaplanyan, Anton Sochenov, Thomas Leimkuhler, Mikhail Okunev, Todd Goodall, Gizem Rufo (2019). "DeepFovea: neural reconstruction for foveated rendering and video compression using learned natural video statistics". ACM SIGGRAPH 2019 Talks. https://doi.org/10.1145/3306307.3328186. Retrieved 2026-10-13.
  10. ↑ Anton S. Kaplanyan, Anton Sochenov, Thomas Leimkuhler, Mikhail Okunev, Todd Goodall, Gizem Rufo (2019). "DeepFovea: Neural Reconstruction for Foveated Rendering and Video Compression using Learned Statistics of Natural Videos". ACM Transactions on Graphics, vol. 38, no. 6. https://doi.org/10.1145/3355089.3356557. Retrieved 2026-10-13.
  11. ↑ Farhaan Ebadulla, Chiraag Mudlapur, Gaurav BV (2025-08-19). "GazeProphet: Software-Only Gaze Prediction for VR Foveated Rendering". arXiv. https://arxiv.org/abs/2508.13546. Retrieved 2026-10-13.
  12. ↑ Apple (2024-04-09). "Apple Vision Pro brings a new era of spatial computing to business". Apple Newsroom. https://www.apple.com/newsroom/2024/04/apple-vision-pro-brings-a-new-era-of-spatial-computing-to-business/. Retrieved 2026-10-13.
  13. ↑ 13.0 13.1 Apple. "About Optic ID advanced technology". Apple Support. https://support.apple.com/en-us/118483. Retrieved 2026-10-13.
  14. ↑ David Heaney (2024-04-16). "Quest v64 Brought Two Undocumented Features: Furniture Recognition & Multimodal In Home". UploadVR. https://www.uploadvr.com/quest-v64-undocumented-features-furniture-recognition-multimodal/. Retrieved 2026-10-13.
  15. ↑ Meta Reality Labs Research. "SceneScript". Project Aria. https://www.projectaria.com/scenescript/. Retrieved 2026-10-13.
  16. ↑ Armen Avetisyan, Christopher Xie, Henry Howard-Jenkins, Tsun-Yi Yang, Samir Aroudj, Suvam Patra, Fuyang Zhang, Duncan Frost, Luke Holland, Campbell Orme, Jakob Engel, Edward Miller, Richard Newcombe, Vasileios Balntas (2024-03-19). "SceneScript: Reconstructing Scenes With An Autoregressive Structured Language Model". arXiv. https://arxiv.org/abs/2403.13064. Retrieved 2026-10-13.
  17. ↑ Apple. "Implementing object tracking in your app". Apple Developer Documentation. https://developer.apple.com/documentation/visionos/implementing-object-tracking-in-your-app. Retrieved 2026-10-13.
  18. ↑ 18.0 18.1 Shahram Izadi (2025). "A new look at how Android XR will bring Gemini to glasses and headsets". The Keyword (Google blog). https://blog.google/products/android/android-xr-gemini-glasses-headsets/. Retrieved 2026-10-13.
  19. ↑ 19.0 19.1 19.2 Lawrence Bonk (2024-04-23). "Ray-Ban Meta smart glasses do the AI thing without a projector or subscription". Engadget. https://www.engadget.com/ray-ban-meta-smart-glasses-do-the-ai-thing-without-a-projector-or-subscription-175403559.html. Retrieved 2026-10-13.
  20. ↑ 20.0 20.1 Ben Lang (2020-09-07). "NVIDIA's DLSS 2.1 Supersampling Now Supports VR, Could Significantly Boost Fidelity". Road to VR. https://roadtovr.com/nvidias-latest-dlss-2-1-supersampling-now-supports-vr/. Retrieved 2026-10-13.
  21. ↑ Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, Ren Ng (2020). "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis". arXiv (ECCV 2020). https://arxiv.org/abs/2003.08934. Retrieved 2026-10-13.
  22. ↑ Linning Xu, Vasu Agrawal, William Laney, Tony Garcia, Aayush Bansal, Changil Kim, Samuel Rota Bulo, Lorenzo Porzi, Peter Kontschieder, Aljaz Bozic, Dahua Lin, Michael Zollhofer, Christian Richardt (2023). "VR-NeRF: High-Fidelity Virtualized Walkable Spaces". SIGGRAPH Asia 2023 Conference Papers. https://doi.org/10.1145/3610548.3618139. Retrieved 2026-10-13.
  23. ↑ 23.0 23.1 Bernhard Kerbl, Georgios Kopanas, Thomas Leimkuhler, George Drettakis (2023). "3D Gaussian Splatting for Real-Time Radiance Field Rendering". ACM Transactions on Graphics, vol. 42, no. 4. https://doi.org/10.1145/3592433. Retrieved 2026-10-13.
  24. ↑ 24.0 24.1 Shi Qiu, Binzhu Xie, Qixuan Liu, Pheng-Ann Heng (2024-12-09). "Advancing Extended Reality with 3D Gaussian Splatting: Innovations and Prospects". arXiv. https://arxiv.org/abs/2412.06257. Retrieved 2026-10-13.
  25. ↑ 25.0 25.1 Michael Rubloff (2025-09-17). "Meta Announces Hyperscape Capture and Meta Horizon Engine". Radiance Fields. https://radiancefields.com/meta-announces-meta-horizon-engine-and-capture. Retrieved 2026-10-13.
  26. ↑ Tomislav Bezmalinovic (2025-09-18). "Meta Quest 3 can now capture photorealistic VR environments". heise online. https://heise.de/-10661056. Retrieved 2026-10-13.
  27. ↑ Stephen Lombardi, Jason Saragih, Tomas Simon, Yaser Sheikh (2018). "Deep appearance models for face rendering". ACM Transactions on Graphics, vol. 37, no. 4. https://doi.org/10.1145/3197517.3201401. Retrieved 2026-10-13.
  28. ↑ Fernanda De La Torre, Cathy Mengying Fang, Han Huang, Andrzej Banburski-Fahey, Judith Amores Fernandez, Jaron Lanier (2024). "LLMR: Real-time Prompting of Interactive Worlds using Large Language Models". Proceedings of the CHI Conference on Human Factors in Computing Systems (CHI '24). https://doi.org/10.1145/3613904.3642579. Retrieved 2026-10-13.
  29. ↑ Meta (2025-08-27). "Meet the Newest GenAI Tools: Environment Generation and Embodied LLM NPCs". Meta Horizon Developers Blog. https://developers.meta.com/horizon/blog/environment-generation-embodied-conversational-llm-npcs-genai-tools/. Retrieved 2026-10-13.
  30. ↑ NVIDIA (2024-03-18). "NVIDIA Digital Human Technologies Bring AI Characters to Life". NVIDIA Newsroom. https://nvidianews.nvidia.com/news/nvidia-digital-human-technologies-bring-ai-characters-to-life-6900750. Retrieved 2026-10-13.
  31. ↑ David Heaney (2025-09-10). "Horizon Worlds Creators Can Now Add Scriptable Conversational AI NPCs". UploadVR. https://www.uploadvr.com/meta-horizon-worlds-conversational-ai-npcs/. Retrieved 2026-10-13.
  32. ↑ 32.0 32.1 Frederik Roland Christiansen, Linus Norgaard Hollensberg, Niko Bach Jensen, Kristian Julsgaard, Kristian Nyborg Jespersen, Ivan Nikolov (2024). "Exploring Presence in Interactions with LLM-Driven NPCs: A Comparative Study of Speech Recognition and Dialogue Options". Proceedings of the 30th ACM Symposium on Virtual Reality Software and Technology (VRST '24). https://doi.org/10.1145/3641825.3687716. Retrieved 2026-10-13.
  33. ↑ 33.0 33.1 33.2 Sueda Ozkaya, Santiago Berrezueta-Guzman, Stefan Wagner (2025). "How LLMs Are Shaping the Future of Virtual Reality". IEEE Access, vol. 13. https://doi.org/10.1109/ACCESS.2025.3631594. Retrieved 2026-10-13.
  34. ↑ Will Shanklin (2025-04-23). "Meta is bringing smart glasses live translation and AI to more people". Engadget. https://engadget.com/wearables/meta-is-bringing-smart-glasses-live-translation-and-ai-to-more-people-184546291.html. Retrieved 2026-10-13.
  35. ↑ 35.0 35.1 35.2 35.3 Meta (2025-09-17). "Meta Ray-Ban Display: Breakthrough AI Glasses Available Now". Meta Quest Blog. https://www.meta.com/blog/meta-ray-ban-display-ai-glasses-connect-2025/. Retrieved 2026-10-13.
  36. ↑ 36.0 36.1 David Heaney (2025-10-21). "Samsung Galaxy XR With Android XR Out Now For $1800, Controllers $250". UploadVR. https://www.uploadvr.com/samsung-galaxy-xr-google-android-xr-out-now/. Retrieved 2026-10-13.
  37. ↑ Shahram Izadi. "Introducing Galaxy XR, the first Android XR headset". The Keyword (Google blog). https://blog.google/products/android/samsung-galaxy-xr. Retrieved 2026-10-13.
  38. ↑ 38.0 38.1 Shahram Izadi (2026-05-19). "Intelligent eyewear is coming this fall". The Keyword (Google blog). https://blog.google/products-and-platforms/platforms/android/android-xr-io-2026/. Retrieved 2026-10-13.
  39. ↑ 39.0 39.1 Apple (2025-03-31). "Apple Intelligence comes to Apple Vision Pro today with visionOS 2.4". Apple Newsroom. https://www.apple.com/newsroom/2025/03/apple-intelligence-comes-to-apple-vision-pro-today-with-visionos-2-4/. Retrieved 2026-10-13.
  40. ↑ Be My Eyes (2024-09-25). "Be My Eyes and Meta Announce Accessibility Partnership". Be My Eyes. https://www.bemyeyes.com/?p=6452. Retrieved 2026-10-13.
  41. ↑ Joseph Cox (2024-10-02). "Someone Put Facial Recognition Tech onto Meta's Smart Glasses to Instantly Dox Strangers". 404 Media. https://404media.co/someone-put-facial-recognition-tech-onto-metas-smart-glasses-to-instantly-dox-strangers. Retrieved 2026-10-13.
  42. ↑ Mark Roman Miller, Fernanda Herrera, Hanseul Jun, James A. Landay, Jeremy N. Bailenson (2020). "Personal identifiability of user tracking data during observation of 360-degree VR video". Scientific Reports, vol. 10. https://doi.org/10.1038/s41598-020-74486-y. Retrieved 2026-10-13.
  43. ↑ Vivek Nair, Wenbo Guo, Justus Mattern, Rui Wang, James F. O'Brien, Louis Rosenberg, Dawn Song (2023). "Unique Identification of 50,000+ Virtual Reality Users from Head and Hand Motion Data". 32nd USENIX Security Symposium (USENIX Security 23). https://www.usenix.org/conference/usenixsecurity23/presentation/nair-identification. Retrieved 2026-10-13.
Retrieved from "https://vrarwiki.com/index.php?title=Artificial_intelligence&oldid=51934"