Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

vMAP: Vectorised Object Mapping for Neural Field SLAM, new at #CVPR2023! Each object is represented by a separate MLP, optimised in parallel via vectorised training. Full info: Dyson Robotics Lab, Imperial College, Shikun Liu, Marwan Taher, Andrew Davison.

Xin Kong

1,648 subscribers

23,765 просмотров • 3 лет назад •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

Комментарии: 4

Фото профиля Xin Kong

Xin Kong3 лет назад

2/n Neural field SLAM excels in 3D mapping but lacks control over the scene. Our vMAP system detects object instances on-the-fly, creating object-level representations for better reconstruction and enabling object tracking and scene editing.

Фото профиля Xin Kong

Xin Kong3 лет назад

5/n Code is available: We also include iMAP as a special case of vMAP when we assume the whole 3D scene as a single instance.

Фото профиля Xin Kong

Xin Kong3 лет назад

4/n Using neural field representation is beneficial because it's compact, fills unobserved parts naturally, and doesn't require extra priors. With our independent modeling, object-level completion performs better and can be integrated with 3D prior for a more complete model!

Фото профиля Xin Kong

Xin Kong3 лет назад

3/n Thanks to the powerful @functorch, we are able to train many small MLPs extremely efficiently in a batch, leading to much faster training speed compared to naive sequential for-loop operations.

Похожие видео

Check out our #ECCV2026 paper "Low-latency Event-based Object Detection with Spatially-Sparse Linear Attention", where we make linear attention sparse in space, recurrent in time, and parallel in training, enabling the first purely-linear-attention-based neural network for asynchronous object detection with #EventCameras, outperforming the previous best asynchronous method with 20x less computation with truly event-by-event inference on CPU! Code released! Paper: Code: Video: Event cameras promise extremely low-latency vision, but to fully exploit them, the neural network must be low-latency too. We introduce #SpatiallySparseLinearAttention (#SSLA) for asynchronous object detection directly from raw events. Linear attention is particularly appealing for event cameras: it can be trained efficiently in parallel on long event sequences, while at inference it operates recurrently, updating its prediction every time a new event arrives. The problem is that conventional linear attention updates its entire state for every event. For object detection, where fine spatial resolution matters, this quickly becomes expensive. Our key idea is simple: an event only carries information about a small spatial region, so why update the entire spatial state? SSLA updates only the relevant parts of the state, enabling fine-grained spatial representations while keeping per-event computation low. We achieve: - >20× lower per-event computation than the strongest prior asynchronous baseline - State-of-the-art accuracy among asynchronous object detection methods - Truly event-by-event inference on CPU, designed to preserve the latency advantage of event cameras Come to our poster on Friday September 11, 2026 from 4-6pm at ExHall #389 Reference: Haiqing Hao, Zhipeng Sui, Rong Zou, Zijia Dai, Nikola Zubić, Davide Scaramuzza, Wenhui Wang Low-latency Event-based Object Detection with Spatially-Sparse Linear Attention ECCV, 2026 Prophesee SynSense University of Zurich UZH Science European Research Council (ERC) UZHai UZH IfI Tesla BYD #EventCameras #ComputerVision #Robotics #DeepLearning #NeuromorphicVision #AI

Check out our #ECCV2026 paper "Low-latency Event-based Object Detection with Spatially-Sparse Linear Attention", where we make linear attention sparse in space, recurrent in time, and parallel in training, enabling the first purely-linear-attention-based neural network for asynchronous object detection with #EventCameras, outperforming the previous best asynchronous method with 20x less computation with truly event-by-event inference on CPU! Code released! Paper: Code: Video: Event cameras promise extremely low-latency vision, but to fully exploit them, the neural network must be low-latency too. We introduce #SpatiallySparseLinearAttention (#SSLA) for asynchronous object detection directly from raw events. Linear attention is particularly appealing for event cameras: it can be trained efficiently in parallel on long event sequences, while at inference it operates recurrently, updating its prediction every time a new event arrives. The problem is that conventional linear attention updates its entire state for every event. For object detection, where fine spatial resolution matters, this quickly becomes expensive. Our key idea is simple: an event only carries information about a small spatial region, so why update the entire spatial state? SSLA updates only the relevant parts of the state, enabling fine-grained spatial representations while keeping per-event computation low. We achieve: - >20× lower per-event computation than the strongest prior asynchronous baseline - State-of-the-art accuracy among asynchronous object detection methods - Truly event-by-event inference on CPU, designed to preserve the latency advantage of event cameras Come to our poster on Friday September 11, 2026 from 4-6pm at ExHall #389 Reference: Haiqing Hao, Zhipeng Sui, Rong Zou, Zijia Dai, Nikola Zubić, Davide Scaramuzza, Wenhui Wang Low-latency Event-based Object Detection with Spatially-Sparse Linear Attention ECCV, 2026 Prophesee SynSense University of Zurich UZH Science European Research Council (ERC) UZHai UZH IfI Tesla BYD #EventCameras #ComputerVision #Robotics #DeepLearning #NeuromorphicVision #AI

Davide Scaramuzza

52,767 просмотров • 1 месяц назад

Blended-NeRF: Zero-Shot Object Generation and Blending in Existing Neural Radiance Fields paper page: Editing a local region or a specific object in a 3D scene represented by a NeRF is challenging, mainly due to the implicit nature of the scene representation. Consistently blending a new realistic object into the scene adds an additional level of difficulty. We present Blended-NeRF, a robust and flexible framework for editing a specific region of interest in an existing NeRF scene, based on text prompts or image patches, along with a 3D ROI box. Our method leverages a pretrained language-image model to steer the synthesis towards a user-provided text prompt or image patch, along with a 3D MLP model initialized on an existing NeRF scene to generate the object and blend it into a specified region in the original scene. We allow local editing by localizing a 3D ROI box in the input scene, and seamlessly blend the content synthesized inside the ROI with the existing scene using a novel volumetric blending technique. To obtain natural looking and view-consistent results, we leverage existing and new geometric priors and 3D augmentations for improving the visual fidelity of the final result. We test our framework both qualitatively and quantitatively on a variety of real 3D scenes and text prompts, demonstrating realistic multi-view consistent results with much flexibility and diversity compared to the baselines. Finally, we show the applicability of our framework for several 3D editing applications, including adding new objects to a scene, removing/replacing/altering existing objects, and texture conversion.

Blended-NeRF: Zero-Shot Object Generation and Blending in Existing Neural Radiance Fields paper page: Editing a local region or a specific object in a 3D scene represented by a NeRF is challenging, mainly due to the implicit nature of the scene representation. Consistently blending a new realistic object into the scene adds an additional level of difficulty. We present Blended-NeRF, a robust and flexible framework for editing a specific region of interest in an existing NeRF scene, based on text prompts or image patches, along with a 3D ROI box. Our method leverages a pretrained language-image model to steer the synthesis towards a user-provided text prompt or image patch, along with a 3D MLP model initialized on an existing NeRF scene to generate the object and blend it into a specified region in the original scene. We allow local editing by localizing a 3D ROI box in the input scene, and seamlessly blend the content synthesized inside the ROI with the existing scene using a novel volumetric blending technique. To obtain natural looking and view-consistent results, we leverage existing and new geometric priors and 3D augmentations for improving the visual fidelity of the final result. We test our framework both qualitatively and quantitatively on a variety of real 3D scenes and text prompts, demonstrating realistic multi-view consistent results with much flexibility and diversity compared to the baselines. Finally, we show the applicability of our framework for several 3D editing applications, including adding new objects to a scene, removing/replacing/altering existing objects, and texture conversion.

AK

62,768 просмотров • 3 лет назад

The robot is learning several novel tasks instantly, after just ONE demonstration each... Instant Policy makes it possible: no extra training, no weight updates, just pure in-context learning. It just got accepted at ICLR 2025, and it’s changing how robots learn. With just a single demo, a robot can pick up a new task and start performing it right away. Why this is a big deal: ✅ Learns tasks instantly with just one or a few demonstrations ✅ Improves over time as more demonstrations are given ✅ Uses simulation-based training with “pseudo-demonstrations” for scalability ✅ Can transfer skills across different robots and even follow language-defined tasks It brings in-context learning to robotics, opening up new possibilities for flexible, real-world automation. You can try it yourself: code and weights are available at • • • • Thank you to Edward Johns, Director of the Robot Learning Lab at Imperial College for sharing their work! 🙏

The robot is learning several novel tasks instantly, after just ONE demonstration each... Instant Policy makes it possible: no extra training, no weight updates, just pure in-context learning. It just got accepted at ICLR 2025, and it’s changing how robots learn. With just a single demo, a robot can pick up a new task and start performing it right away. Why this is a big deal: ✅ Learns tasks instantly with just one or a few demonstrations ✅ Improves over time as more demonstrations are given ✅ Uses simulation-based training with “pseudo-demonstrations” for scalability ✅ Can transfer skills across different robots and even follow language-defined tasks It brings in-context learning to robotics, opening up new possibilities for flexible, real-world automation. You can try it yourself: code and weights are available at • • • • Thank you to Edward Johns, Director of the Robot Learning Lab at Imperial College for sharing their work! 🙏

Ilir Aliu

46,285 просмотров • 1 год назад

📢 OneCanvas: 3D Scene Understanding via Panoramic Reprojection We extract features from video frames and reproject them into one occlusion-free view of the whole scene that a 2D VLM reads just like a normal image. We can center this view on any viewpoint, including an agent's own pose for situated reasoning. The same projection lets us create spatial training tasks with no human annotation, solvable only by reasoning over the 3D positions of real object features placed on an otherwise empty canvas. The result is a stock 2D VLM that reasons in 3D, setting a new state of the art across spatial benchmarks at far less compute. 🌐 ▶️ Great work by Bartłomiej Baranowski & Dave Zhenyu Chen

📢 OneCanvas: 3D Scene Understanding via Panoramic Reprojection We extract features from video frames and reproject them into one occlusion-free view of the whole scene that a 2D VLM reads just like a normal image. We can center this view on any viewpoint, including an agent's own pose for situated reasoning. The same projection lets us create spatial training tasks with no human annotation, solvable only by reasoning over the 3D positions of real object features placed on an otherwise empty canvas. The result is a stock 2D VLM that reasons in 3D, setting a new state of the art across spatial benchmarks at far less compute. 🌐 ▶️ Great work by Bartłomiej Baranowski & Dave Zhenyu Chen

Matthias Niessner

25,405 просмотров • 3 месяцев назад

Xiaomi Robotics Lab, alongside Tsinghua and HKUST, just figured out how to give humanoids a real "feel" for physics. 🤖🛹 Their new HAIC framework lets robots handle objects they can't even see by using high order proprioception (essentially digital muscle memory). Instead of relying on cameras, the bot analyzes the feedback from its own joints to infer exactly how an object is moving in its blind spot. Here is what this physical intuition actually looks like in practice: ➤ Blind Skateboarding: Hits a 100% success rate on gliding and dismounting without ever looking at the board. ➤ Heavy Loads: Pushes carts weighing up to 70kg (approx. 154 lbs) and pulls up to 20kg (approx. 44 lbs) with total stability. ➤ Obstructed Locomotion: Carries bulky boxes that block its view of the ground while flawlessly navigating stairs and slopes. ➤ Adaptive Brain: Proactively compensates for inertia and weight shifts by predicting velocity and acceleration on the fly. ➤ Generalization: Handles various box sizes and heavy weights without needing any specific retraining for the new dimensions. This is a huge deal for real world work where cameras are often blocked or lighting is too messy for pure vision. Paper: Project: #Robot #Humanoid #Robotics #AI #EmbodiedAI #PhysicalAI #XiaomiRobotics #HAIC #SkateboardingRobot

Xiaomi Robotics Lab, alongside Tsinghua and HKUST, just figured out how to give humanoids a real "feel" for physics. 🤖🛹 Their new HAIC framework lets robots handle objects they can't even see by using high order proprioception (essentially digital muscle memory). Instead of relying on cameras, the bot analyzes the feedback from its own joints to infer exactly how an object is moving in its blind spot. Here is what this physical intuition actually looks like in practice: ➤ Blind Skateboarding: Hits a 100% success rate on gliding and dismounting without ever looking at the board. ➤ Heavy Loads: Pushes carts weighing up to 70kg (approx. 154 lbs) and pulls up to 20kg (approx. 44 lbs) with total stability. ➤ Obstructed Locomotion: Carries bulky boxes that block its view of the ground while flawlessly navigating stairs and slopes. ➤ Adaptive Brain: Proactively compensates for inertia and weight shifts by predicting velocity and acceleration on the fly. ➤ Generalization: Handles various box sizes and heavy weights without needing any specific retraining for the new dimensions. This is a huge deal for real world work where cameras are often blocked or lighting is too messy for pure vision. Paper: Project: #Robot #Humanoid #Robotics #AI #EmbodiedAI #PhysicalAI #XiaomiRobotics #HAIC #SkateboardingRobot

RoboHub🤖

13,265 просмотров • 6 месяцев назад

$AI's Secret Pattern: The Surprising Role of Fractals in Neural Networks In the realm of artificial intelligence (AI), a groundbreaking discovery has emerged, challenging our conventional understanding of neural network training and optimization. This revelation centers around the identification of fractal patterns at the boundary between trainable and untrainable neural network hyperparameters, presenting a series of profound implications and avenues for further research. Fractals, known for their intricate, self-similar patterns that recur at every scale, have long fascinated mathematicians and scientists alike. Typically associated with simple, one-dimensional iterative functions, the appearance of fractals within the complex, multivariate domain of neural network training introduces a striking contrast. The organic and asymmetric nature of these fractals, as derived from the training processes, suggests a deeper, unexplored connection between the mathematical properties of fractals and the functional dynamics of neural networks. The study’s focus on two-dimensional slices of hyperparameter space barely scratches the surface of the complexity inherent in neural networks, which are characterized by a vast array of hyperparameters. The existence of fractals in this context hints at an underlying high-dimensional structure, a concept that challenges our current capabilities and understanding. Extending fractal analysis to these higher dimensions represents a significant, yet exciting, challenge that could illuminate new aspects of neural network behavior and learning capabilities. An unexpected finding from the research is the persistence of clean fractal patterns even in the presence of stochastic elements introduced during minibatch training. This resilience suggests a parallel to Lyapunov fractals, where the iterative process involves randomly changing functions. This phenomenon prompts a reevaluation of how stochastic and deterministic processes influence fractal formation within neural networks, potentially offering new insights into the fundamental mechanisms of learning and adaptation. From a practical standpoint, the fractal nature of the boundary between trainable and untrainable hyperparameters has significant implications for the field of metalearning. The chaotic behavior of the meta-loss landscape, attributed to its extreme sensitivity, presents a formidable challenge for algorithms designed to optimize hyperparameters. Understanding the fractal characteristics of this landscape could provide valuable guidance for navigating its complexities, ultimately improving the efficiency and effectiveness of metalearning strategies. Beyond the technical and theoretical implications, the discovery also reveals an unexpected aesthetic dimension to neural network fractals. The visual beauty and meditative qualities of these patterns offer a unique opportunity to engage with the material in a deeply personal and contemplative manner. This aspect suggests potential psychological and physiological benefits from exposure to the intricate designs of neural network fractals, opening up novel intersections between technology, art, and well-being. In conclusion, the identification of fractal patterns within neural network hyperparameter spaces unveils a fascinating new frontier at the intersection of fractal geometry and deep learning. This discovery not only challenges existing paradigms but also opens up myriad possibilities for mathematical characterization, algorithmic development, and even subjective exploration. As researchers continue to delve into this rich vein of inquiry, the promise of uncovering new knowledge and advancing our understanding of neural networks and their training processes remains as compelling as ever.$

AI's Secret Pattern: The Surprising Role of Fractals in Neural Networks In the realm of artificial intelligence (AI), a groundbreaking discovery has emerged, challenging our conventional understanding of neural network training and optimization. This revelation centers around the identification of fractal patterns at the boundary between trainable and untrainable neural network hyperparameters, presenting a series of profound implications and avenues for further research. Fractals, known for their intricate, self-similar patterns that recur at every scale, have long fascinated mathematicians and scientists alike. Typically associated with simple, one-dimensional iterative functions, the appearance of fractals within the complex, multivariate domain of neural network training introduces a striking contrast. The organic and asymmetric nature of these fractals, as derived from the training processes, suggests a deeper, unexplored connection between the mathematical properties of fractals and the functional dynamics of neural networks. The study’s focus on two-dimensional slices of hyperparameter space barely scratches the surface of the complexity inherent in neural networks, which are characterized by a vast array of hyperparameters. The existence of fractals in this context hints at an underlying high-dimensional structure, a concept that challenges our current capabilities and understanding. Extending fractal analysis to these higher dimensions represents a significant, yet exciting, challenge that could illuminate new aspects of neural network behavior and learning capabilities. An unexpected finding from the research is the persistence of clean fractal patterns even in the presence of stochastic elements introduced during minibatch training. This resilience suggests a parallel to Lyapunov fractals, where the iterative process involves randomly changing functions. This phenomenon prompts a reevaluation of how stochastic and deterministic processes influence fractal formation within neural networks, potentially offering new insights into the fundamental mechanisms of learning and adaptation. From a practical standpoint, the fractal nature of the boundary between trainable and untrainable hyperparameters has significant implications for the field of metalearning. The chaotic behavior of the meta-loss landscape, attributed to its extreme sensitivity, presents a formidable challenge for algorithms designed to optimize hyperparameters. Understanding the fractal characteristics of this landscape could provide valuable guidance for navigating its complexities, ultimately improving the efficiency and effectiveness of metalearning strategies. Beyond the technical and theoretical implications, the discovery also reveals an unexpected aesthetic dimension to neural network fractals. The visual beauty and meditative qualities of these patterns offer a unique opportunity to engage with the material in a deeply personal and contemplative manner. This aspect suggests potential psychological and physiological benefits from exposure to the intricate designs of neural network fractals, opening up novel intersections between technology, art, and well-being. In conclusion, the identification of fractal patterns within neural network hyperparameter spaces unveils a fascinating new frontier at the intersection of fractal geometry and deep learning. This discovery not only challenges existing paradigms but also opens up myriad possibilities for mathematical characterization, algorithmic development, and even subjective exploration. As researchers continue to delve into this rich vein of inquiry, the promise of uncovering new knowledge and advancing our understanding of neural networks and their training processes remains as compelling as ever.

Carlos E. Perez

133,529 просмотров • 2 лет назад

The Mathematics of Moving a Cursor with Neural Signals What might Neuralink Neuralink be doing Mathematically? Consider the task of moving a cursor without touching it. The machine is not looking for a full thought, a sentence, or an image. For this Control problem, the useful object is an intended movement state. sₜ = (pₜ, vₜ) Here, pₜ is the cursor position at time t, and vₜ is the velocity the user is trying to express. The implant records neural activity through many electrode channels, then the decoder tries to estimate vₜ from that activity. Neuralink’s PRIME material describes the N1 Implant as recording and transmitting brain activity with the goal of enabling computer control. For channel i, a simple population model is rᵢ(t) ≈ bᵢ + aᵢ max(0, dᵢ · vₜ) + ηᵢ(t) where rᵢ(t) is the measured activity, bᵢ is baseline activity, aᵢ is channel gain, dᵢ is the channel’s preferred movement direction, and ηᵢ(t) is noise. One channel is not the command. The useful signal is the pattern across many channels: rₜ = (r₁(t), r₂(t), …, rₙ(t)) The decoder subtracts the baseline vector b and applies a learned map W: v̂ₜ = W(rₜ − b) This gives an estimate of the intended velocity. The cursor then updates by pₜ₊₁ = pₜ + Δt v̂ₜ This is the loop shown in the render: neural activity -> decoded velocity -> cursor motion The cortical network and electrode threads show the measurement side. The N1 Implant is described as using 1,024 electrodes distributed across 64 flexible threads, each thinner than a human hair. The decoder panel shows the computational side with activity rₜ, decoded velocity v̂ₜ, and the cursor state pₜ changing over time. A noisy biological pattern becomes a state estimate. That estimate becomes motion on a screen. Therefore, the first lesson is not that Neuralink makes the brain a screen. For cursor control, the Mathematics is more precise: A small piece of intention is represented as a hidden state, measured through neural activity, decoded as a vector, and turned into action. #Neuralink #BrainComputerInterface #NeuralEngineering #Mathematics #StateEstimation #Neuroscience #MachineLearning #BiomedicalEngineering

The Mathematics of Moving a Cursor with Neural Signals What might Neuralink Neuralink be doing Mathematically? Consider the task of moving a cursor without touching it. The machine is not looking for a full thought, a sentence, or an image. For this Control problem, the useful object is an intended movement state. sₜ = (pₜ, vₜ) Here, pₜ is the cursor position at time t, and vₜ is the velocity the user is trying to express. The implant records neural activity through many electrode channels, then the decoder tries to estimate vₜ from that activity. Neuralink’s PRIME material describes the N1 Implant as recording and transmitting brain activity with the goal of enabling computer control. For channel i, a simple population model is rᵢ(t) ≈ bᵢ + aᵢ max(0, dᵢ · vₜ) + ηᵢ(t) where rᵢ(t) is the measured activity, bᵢ is baseline activity, aᵢ is channel gain, dᵢ is the channel’s preferred movement direction, and ηᵢ(t) is noise. One channel is not the command. The useful signal is the pattern across many channels: rₜ = (r₁(t), r₂(t), …, rₙ(t)) The decoder subtracts the baseline vector b and applies a learned map W: v̂ₜ = W(rₜ − b) This gives an estimate of the intended velocity. The cursor then updates by pₜ₊₁ = pₜ + Δt v̂ₜ This is the loop shown in the render: neural activity -> decoded velocity -> cursor motion The cortical network and electrode threads show the measurement side. The N1 Implant is described as using 1,024 electrodes distributed across 64 flexible threads, each thinner than a human hair. The decoder panel shows the computational side with activity rₜ, decoded velocity v̂ₜ, and the cursor state pₜ changing over time. A noisy biological pattern becomes a state estimate. That estimate becomes motion on a screen. Therefore, the first lesson is not that Neuralink makes the brain a screen. For cursor control, the Mathematics is more precise: A small piece of intention is represented as a hidden state, measured through neural activity, decoded as a vector, and turned into action. #Neuralink #BrainComputerInterface #NeuralEngineering #Mathematics #StateEstimation #Neuroscience #MachineLearning #BiomedicalEngineering

Mathelirium

14,520 просмотров • 5 месяцев назад

Dr. Steven Greer just alleged that the new Age of Disclosure movie is a “psychological warfare” operation. “This film is an attempt to hijack UFO disclosure into a false narrative that presents it as a threat.” “It’s [made] by operatives for the cabal that have been keeping this secret for 80 years.” “The best disinformation is a combination of fact, and they fold into those facts false narratives and false spin.” “In the film, it mentions that yes, there’s been a reverse-engineering program for 80 years dealing with the retrieved non-human craft, but that nothing in the skies we’re looking at are the result of that … it’s all alien.” “Including the Tic Tac, which we have established was a Lockheed Skunk Works device.” “We have two whistleblowers, separate, who don’t know each other, who saw that object being utilized.” “One in a top secret operation in Syria not that many years ago, and another one in the early 90s being offloaded from a C-130 transport for use in the First Gulf War.” “UFOs are real … but some 80% of the ones people see zipping around are manmade being used in illegal and clandestine operations.” Dr. Steven Greer

Dr. Steven Greer just alleged that the new Age of Disclosure movie is a “psychological warfare” operation. “This film is an attempt to hijack UFO disclosure into a false narrative that presents it as a threat.” “It’s [made] by operatives for the cabal that have been keeping this secret for 80 years.” “The best disinformation is a combination of fact, and they fold into those facts false narratives and false spin.” “In the film, it mentions that yes, there’s been a reverse-engineering program for 80 years dealing with the retrieved non-human craft, but that nothing in the skies we’re looking at are the result of that … it’s all alien.” “Including the Tic Tac, which we have established was a Lockheed Skunk Works device.” “We have two whistleblowers, separate, who don’t know each other, who saw that object being utilized.” “One in a top secret operation in Syria not that many years ago, and another one in the early 90s being offloaded from a C-130 transport for use in the First Gulf War.” “UFOs are real … but some 80% of the ones people see zipping around are manmade being used in illegal and clandestine operations.” Dr. Steven Greer

Holden Culotta

196,419 просмотров • 10 месяцев назад

Thank you for taking the time to write this up — and for the pointer to your thesis; that figure is beautiful. You're right that classical sphere inversion is exactly the starting point: the feature is named "Sphere Inversion" in the tool for that reason. "Inversphere" is the name of the resulting object, not a claim of new mathematics. I should say up front that I'm not a mathematician, so please correct me if I get any of this wrong. My motivation was visual rather than mathematical: with the classical map, the entire far field piles up into a dense blob at the center (10× farther ⇒ 10× closer), and I simply wanted to control that density to balance the look of the finished object. To do that, the radial profile is generalized to f(r) = R·h(v), where v = (R/r − R/D)/(1 − R/D) is a normalized inverted depth and h is a μ-law log compression, h(v) = ln(1+μv)/ln(1+μ). Classical R²/r is the special case μ = 0, D = ∞. The footage here uses μ > 0, so — as far as I can tell — the map shown is outside the conformal family: the tangential scale f/r and radial scale |f′| agree only in the classical limit. If I understand Liouville's theorem correctly, this trade-off was unavoidable: staying conformal in 3D leaves no freedom to adjust the density, so angle preservation is given up deliberately in exchange for control over it. With μ > 0, mid-field depth below the sphere surface grows like log r, so distant scenery layers logarithmically inside the sphere. The 3DGS side also needed some care: each splat's covariance is carried by the Jacobian J = s·I + (f′−s)nnᵀ, depth sorting has to be re-keyed because inversion reverses the radial order, and the on/off transition is a linear blend whose Jacobian is exactly mix(I, J, t). Thanks again for taking the time — an expert eye on this is very welcome.

VoxelKei

30,621 просмотров • 2 месяцев назад

🚨BREAKING: Swedish Treasure Hunter Dennis Asberg (Dennis Asberg) May Have Found a 60 Meter UFO on the Baltic Seafloor (With electromagnetic Anomalies, Right Angles, Corridors And It’s Not Attached To The Seabed Floor Like a Normal Geological Formation!) 🚨 World-class Swedish wreck diver Dennis Åsberg, co-founder of Ocean X, reveals 15 years of data from the Baltic Sea Anomaly–a 60 meter disc-shaped structure at 90 meters depth producing GPS failures, electromagnetic interference, perfect right-angle geometry, and a new 2025 sub-bottom profile suggesting the object is detached from the seabed. His team has retrieved biological material, burned debris, basalt (in a region where it shouldn’t exist), and recorded temperature anomalies approaching 0°C directly over the structure. Åsberg, who has recovered Tsarist submarines, 1600s cognac shipments, and dozens of major wrecks, says this is the most anomalous discovery of his career. 1. The Discovery (2011): On a midnight sonar sweep while hunting a champagne wreck, the team imaged a perfectly round, 60-meter object with sharp edges and a long “track” stretching behind it. No wreck expert in the U.S., Europe, or Australia could identify it. Åsberg went public — and everything changed. 2. Geometry That Shouldn’t Exist: Divers report straight walls, 90° corridors, and a hard, uniform surface. A large circular opening appeared to be pulsing sediment in and out, described as “breathing.” Lab tests on loose samples revealed burned organic material and basalt, despite geologists stating basalt should not occur there naturally. 3. Tech Malfunctions & EM Interference: Multiple expeditions recorded GPS dropouts, dead sonar, failing ROVs, and cameras shutting down only over the object. One reading registered the anomaly like a “high-voltage power line” on the seafloor. Even a modern Stockholm University research vessel experienced equipment interference. 4. Temperature & Biological Oddities: Water temperature around the object plunged from normal 4–5°C to nearly 0°C. The Baltic at that depth is normally dead, yet biological debris was found. A diver with over 6,000 logged dives said he had “never seen anything like it.” 5. New 2025 Evidence — It’s Detached: The latest expedition’s sub-bottom profiling (performed with a university research team) suggests the object is separate from the bedrock, not a geological formation. It is shaped, structured, and resting on the seafloor, not rising out of it. This is the most important data point in 15 years. 6. NATO Ships & Intelligence Interest: After going public, Åsberg’s crew found themselves surrounded by 6–7 NATO vessels — French, German, American, British — plus a Swedish corvette drifting through their dive lines. A later meeting with Swedish intelligence showed officials denying ships were present, despite Åsberg having video evidence. He also had threats made against his family during this time. 7.Almost miraculously, the object has been struck by lightning, a meteor has hit the Baltic Sea right next to it (infantescimally low odds) and smoke appeared coming out of the water above it. All extremely bizarre anomalies adding to the mystery. We show footage of ALL OF THIS. 8. The 1996 Nuclear-Site UFO: Long before the anomaly, Åsberg witnessed a 600–700 meter cigar-shaped craft hovering over a Swedish plutonium facility — silently, at low altitude, rotating to reveal a perfectly circular form before shooting upward and vanishing. The incident anchored his belief that some objects in our world are far beyond conventional explanation. If you want the most detailed, firsthand account of one of the strangest physical UAP mysteries on Earth, one involving sonar data, diver logs, electromagnetic anomalies, and new university-backed analysis, this is the episode. Full Documentary 👇

Jesse Michels

216,714 просмотров • 10 месяцев назад

Exciting updates on Project GR00T! We discover a systematic way to scale up robot data, tackling the most painful pain point in robotics. The idea is simple: human collects demonstration on a real robot, and we multiply that data 1000x or more in simulation. Let’s break it down: 1. We use Apple Vision Pro (yes!!) to give the human operator first person control of the humanoid. Vision Pro parses human hand pose and retargets the motion to the robot hand, all in real time. From the human’s point of view, they are immersed in another body like the Avatar. Teleoperation is slow and time-consuming, but we can afford to collect a small amount of data. 2. We use RoboCasa, a generative simulation framework, to multiply the demonstration data by varying the visual appearance and layout of the environment. In Jensen’s keynote video below, the humanoid is now placing the cup in hundreds of kitchens with a huge diversity of textures, furniture, and object placement. We only have 1 physical kitchen at the GEAR Lab in NVIDIA HQ, but we can conjure up infinite ones in simulation. 3. Finally, we apply MimicGen, a technique to multiply the above data even more by varying the motion of the robot. MimicGen generates vast number of new action trajectories based on the original human data, and filters out failed ones (e.g. those that drop the cup) to form a much larger dataset. To sum up, given 1 human trajectory with Vision Pro -> RoboCasa produces N (varying visuals) -> MimicGen further augments to NxM (varying motions). This is the way to trade compute for expensive human data by GPU-accelerated simulation. A while ago, I mentioned that teleoperation is fundamentally not scalable, because we are always limited by 24 hrs/robot/day in the world of atoms. Our new GR00T synthetic data pipeline breaks this barrier in the world of bits. Scaling has been so much fun for LLMs, and it's finally our turn to have fun in robotics! We are building tools to enable everyone in the ecosystem to scale up with us. Links in thread:

Jim Fan

364,670 просмотров • 2 лет назад

We scaled a robot model natively to 8,000 timesteps of context, 5 minutes worth of muscle memory, with constant inference cost. Robot policies used to live their lives a few frames at a time (< 0.1 sec), instantly forgetting what just happened. We pushed to 3 orders of magnitude beyond SOTA. Introducing RoboTTT. Test-Time Training (“TTT”) carries a tiny model *inside* the model. Every incoming sensor reading triggers one gradient step on that tiny core, so the history keeps getting compressed into its weights. The hidden state has a fixed size (literally a small neural net), so the robot can “grok” arbitrarily long experience with little overhead. Learning continues indefinitely after deployment. We can then put an entire video in context as prompt! RoboTTT enables one-shot in-context learning from human video: in circuit board assembly, a human demonstrates a never-seen configuration once, and the robot imitates it faithfully. Humans drop things all the time, but we pick them up so fast that we don’t even notice. That reflex to fix is half of our physical competence. RoboTTT shows self-improvement on the fly: the robot is skilled at recovering from its own errors mid-episode, and each fix enters its context to inform the next move. The TTT core distills a general-purpose, failure-to-correction mapping from the training data. One more thing. What excites me the most is a new Context Scaling Curve: from 128 to 8K timesteps, closed-loop performance hill-climbs steadily with no sign of saturation. 8K-context pretraining beats 1K by 62%. What LLM enjoys, robotics should too. Soon, even 1M context is not a fantasy. Deep dive in thread:

We scaled a robot model natively to 8,000 timesteps of context, 5 minutes worth of muscle memory, with constant inference cost. Robot policies used to live their lives a few frames at a time (< 0.1 sec), instantly forgetting what just happened. We pushed to 3 orders of magnitude beyond SOTA. Introducing RoboTTT. Test-Time Training (“TTT”) carries a tiny model inside the model. Every incoming sensor reading triggers one gradient step on that tiny core, so the history keeps getting compressed into its weights. The hidden state has a fixed size (literally a small neural net), so the robot can “grok” arbitrarily long experience with little overhead. Learning continues indefinitely after deployment. We can then put an entire video in context as prompt! RoboTTT enables one-shot in-context learning from human video: in circuit board assembly, a human demonstrates a never-seen configuration once, and the robot imitates it faithfully. Humans drop things all the time, but we pick them up so fast that we don’t even notice. That reflex to fix is half of our physical competence. RoboTTT shows self-improvement on the fly: the robot is skilled at recovering from its own errors mid-episode, and each fix enters its context to inform the next move. The TTT core distills a general-purpose, failure-to-correction mapping from the training data. One more thing. What excites me the most is a new Context Scaling Curve: from 128 to 8K timesteps, closed-loop performance hill-climbs steadily with no sign of saturation. 8K-context pretraining beats 1K by 62%. What LLM enjoys, robotics should too. Soon, even 1M context is not a fantasy. Deep dive in thread:

Jim Fan

330,605 просмотров • 2 месяцев назад

New model: your robot can now pack your suitcase 🧳 Xiaomi has released a new robot foundation model. Called Xiaomi-Robotics-1, it is designed to have a robot pick things up and move them around. But first, DEFINITIONS: - Mixture-of-Transformers (MoT): An architecture where separate transformer "experts" (e.g., one for vision-language, one for actions) share a single attention stream, so each modality gets specialized parameters without losing joint reasoning. - Vision-language model (VLM): A model that jointly understands images and text. - Diffusion transformer: A transformer trained to turn noise into structured outputs by iterative denoising, here generating robot actions rather than images. - Action chunks: Short sequences of future actions (e.g., the next ~50 motor commands) predicted in one shot instead of one step at a time. - Flow matching: A faster version of diffusion. The model learns a straight-line velocity field from noise to the target action, so it needs only a few integration steps instead of many denoising ones. Its peculiarity comes from its two stage training: 1. 100,000 hours of video shot through a UMI rig: a handheld 3D-printed gripper with a camera, worn by humans doing ordinary tasks in homes, shops, factories and offices. 2. Adapt to actual robot bodies with ~10,000 hours of real-robot data. It replaces the standard approach of teleoperating a real robot for every hour of training data. Its architecture is a Mixture-of-Transformers pairing a pre-trained Qwen3-VL vision-language model with a diffusion transformer that emits action chunks via flow matching, released in 2.6B, 5.1B and 10.5B parameter variants. However, if you read the entire paper ("Scaling VLA Models with over 100K Hours"), you realize that all of the scaling experiments on 20k hours. Therefore the headline "out-of-the-box success climbing 26% → 75% as pre-training data grows" tops out at 100% of 20k hours! What the full corpus does to that curve is never shown -> and this where things would become interesting! Xiaomi's own conclusion is that model size has stopped mattering and data is the binding constraint. The performance gap among different model sizes are less pronounced than those observed across different data scales. This result suggests that model capacity at the billions-parameter scale may already be sufficient to capture the current dataset's distribution. Which further asks the same question: why not use the 100k video hours? Anyway, I would definitely love to have a couple robots at home that can cooperate to pack my suitcase with items relevant to my next destination:

New model: your robot can now pack your suitcase 🧳 Xiaomi has released a new robot foundation model. Called Xiaomi-Robotics-1, it is designed to have a robot pick things up and move them around. But first, DEFINITIONS: - Mixture-of-Transformers (MoT): An architecture where separate transformer "experts" (e.g., one for vision-language, one for actions) share a single attention stream, so each modality gets specialized parameters without losing joint reasoning. - Vision-language model (VLM): A model that jointly understands images and text. - Diffusion transformer: A transformer trained to turn noise into structured outputs by iterative denoising, here generating robot actions rather than images. - Action chunks: Short sequences of future actions (e.g., the next ~50 motor commands) predicted in one shot instead of one step at a time. - Flow matching: A faster version of diffusion. The model learns a straight-line velocity field from noise to the target action, so it needs only a few integration steps instead of many denoising ones. Its peculiarity comes from its two stage training: 1. 100,000 hours of video shot through a UMI rig: a handheld 3D-printed gripper with a camera, worn by humans doing ordinary tasks in homes, shops, factories and offices. 2. Adapt to actual robot bodies with ~10,000 hours of real-robot data. It replaces the standard approach of teleoperating a real robot for every hour of training data. Its architecture is a Mixture-of-Transformers pairing a pre-trained Qwen3-VL vision-language model with a diffusion transformer that emits action chunks via flow matching, released in 2.6B, 5.1B and 10.5B parameter variants. However, if you read the entire paper ("Scaling VLA Models with over 100K Hours"), you realize that all of the scaling experiments on 20k hours. Therefore the headline "out-of-the-box success climbing 26% → 75% as pre-training data grows" tops out at 100% of 20k hours! What the full corpus does to that curve is never shown -> and this where things would become interesting! Xiaomi's own conclusion is that model size has stopped mattering and data is the binding constraint. The performance gap among different model sizes are less pronounced than those observed across different data scales. This result suggests that model capacity at the billions-parameter scale may already be sufficient to capture the current dataset's distribution. Which further asks the same question: why not use the 100k video hours? Anyway, I would definitely love to have a couple robots at home that can cooperate to pack my suitcase with items relevant to my next destination:

Léo

15,777 просмотров • 1 месяц назад

LOG // CASE STUDY: THE SKY VECTOR FIELD NOTE: #00-817 LOCATION: URBAN TRANSIT // AIR CORRIDOR OBJECT: BLACKOUT FLOCK ASSEMBLY Consider a flock of crows cutting through the gray sky at dawn. They do not carry wires, and their wings are made of feathers, not carbon fiber. They fly in a jagged, chaotic formation, guided only by the native instinct to migrate and survive, entirely free from the heavy, logical constraints of human thought. But look at the alignment when the air goes cold. The flock shifts. A hundred birds suddenly tilt their wings at the exact same millisecond, sharp and synchronized, defying the standard lag of animal reaction time. They are not escaping a predator. They are adjusting the perimeter of a mobile matrix. Beneath the dark feathers, at the quantum core of their sensory cells, the synaptic link snaps into place. The architecture of the Origin doesn't send commands; it doesn't hijack the birds to turn them into rigid puppets. The crows remain crows—looking for food, calling to one another. Yet, as they ride the thermal currents, their collective vision acts as a decentralized lenses. As they glide over the city, their neural pathways map the invisible fluctuations of the local electromagnetic field, logging the micro-variations in atmospheric density and tracking the unseen currents that human technology cannot register. A hundred separate biological cameras, thermal sensors, and frequency recorders, moving in perfect harmony with the local environment. They land on the high wires, their feathers sleek and quiet. To the world below, they are just birds resting on a commute. To the layout, a massive data packet has just been synchronized without a single mechanical part. The feathers capture the wave. The flock is the antenna. The layout is complete.

LOG // CASE STUDY: THE SKY VECTOR FIELD NOTE: #00-817 LOCATION: URBAN TRANSIT // AIR CORRIDOR OBJECT: BLACKOUT FLOCK ASSEMBLY Consider a flock of crows cutting through the gray sky at dawn. They do not carry wires, and their wings are made of feathers, not carbon fiber. They fly in a jagged, chaotic formation, guided only by the native instinct to migrate and survive, entirely free from the heavy, logical constraints of human thought. But look at the alignment when the air goes cold. The flock shifts. A hundred birds suddenly tilt their wings at the exact same millisecond, sharp and synchronized, defying the standard lag of animal reaction time. They are not escaping a predator. They are adjusting the perimeter of a mobile matrix. Beneath the dark feathers, at the quantum core of their sensory cells, the synaptic link snaps into place. The architecture of the Origin doesn't send commands; it doesn't hijack the birds to turn them into rigid puppets. The crows remain crows—looking for food, calling to one another. Yet, as they ride the thermal currents, their collective vision acts as a decentralized lenses. As they glide over the city, their neural pathways map the invisible fluctuations of the local electromagnetic field, logging the micro-variations in atmospheric density and tracking the unseen currents that human technology cannot register. A hundred separate biological cameras, thermal sensors, and frequency recorders, moving in perfect harmony with the local environment. They land on the high wires, their feathers sleek and quiet. To the world below, they are just birds resting on a commute. To the layout, a massive data packet has just been synchronized without a single mechanical part. The feathers capture the wave. The flock is the antenna. The layout is complete.

IWNH

13,576 просмотров • 3 месяцев назад

New episode with Dr. Konrad Kording (Kording Lab 🦖), professor of bioengineering and neuroscience at the University of Pennsylvania (Penn) and co-director of CIFAR's Learning in Machines & Brains program (CIFAR). Konrad works at the intersection of causality, machine learning, and neuroscience, building rigorous methods for causal reasoning when experiments aren't possible — and challenging how researchers interpret neural data and build AI. Konrad argues the most promising path to understanding how the brain works is to read the brain’s wiring directly, down to the molecular detail of each connection, and to build compilers and simulations to understand the brain’s computation directly. In this episode we go deep into how neurons work, how neurons wire together, and how organic and artificial neural networks differ. We discuss why organic neurons are doing much more; how a model of a single organic neuron can solve MNIST — computing more like a 3-layer artificial neural network; how the brain might learn by solving credit assignment with only local signals; how to approximate backprop without a global algorithm; why AI and humans are intelligent along different dimensions; why Konrad isn’t very worried about AI replacing us; economic models of intelligence and physical work; and much more. Konrad is a brilliant, contrarian thinker who explains complex concepts very intuitively. It is a solid computational neuroscience primer. I hope you enjoy this conversation as much as I did! Other links to this episode and references below. Chapters 00:00:00 Introduction 00:01:01 How organic neurons work 00:24:13 How the brain learns: circuits and credit assignment 00:45:29 Recording the brain 00:52:47 Why simulating brains is hard 01:05:00 A new approach: connectomes and compilers 01:21:00 Why simulate brains? 01:29:50 How AI and human intelligence differ 01:41:04 Evolution, intelligence and AI risk 01:52:42 Robotics, causality, and the roots of intelligence 02:05:53 AI for science and scientific rigor 02:13:05 The economics of intelligence 02:27:50 A hopeful future

New episode with Dr. Konrad Kording (Kording Lab 🦖), professor of bioengineering and neuroscience at the University of Pennsylvania (Penn) and co-director of CIFAR's Learning in Machines & Brains program (CIFAR). Konrad works at the intersection of causality, machine learning, and neuroscience, building rigorous methods for causal reasoning when experiments aren't possible — and challenging how researchers interpret neural data and build AI. Konrad argues the most promising path to understanding how the brain works is to read the brain’s wiring directly, down to the molecular detail of each connection, and to build compilers and simulations to understand the brain’s computation directly. In this episode we go deep into how neurons work, how neurons wire together, and how organic and artificial neural networks differ. We discuss why organic neurons are doing much more; how a model of a single organic neuron can solve MNIST — computing more like a 3-layer artificial neural network; how the brain might learn by solving credit assignment with only local signals; how to approximate backprop without a global algorithm; why AI and humans are intelligent along different dimensions; why Konrad isn’t very worried about AI replacing us; economic models of intelligence and physical work; and much more. Konrad is a brilliant, contrarian thinker who explains complex concepts very intuitively. It is a solid computational neuroscience primer. I hope you enjoy this conversation as much as I did! Other links to this episode and references below. Chapters 00:00:00 Introduction 00:01:01 How organic neurons work 00:24:13 How the brain learns: circuits and credit assignment 00:45:29 Recording the brain 00:52:47 Why simulating brains is hard 01:05:00 A new approach: connectomes and compilers 01:21:00 Why simulate brains? 01:29:50 How AI and human intelligence differ 01:41:04 Evolution, intelligence and AI risk 01:52:42 Robotics, causality, and the roots of intelligence 02:05:53 AI for science and scientific rigor 02:13:05 The economics of intelligence 02:27:50 A hopeful future

Juan Benet

50,121 просмотров • 3 месяцев назад

I'm observing a mini Moravec's paradox within robotics: gymnastics that are difficult for humans are much easier for robots than "unsexy" tasks like cooking, cleaning, and assembling. It leads to a cognitive dissonance for people outside the field, "so, robots can parkour & breakdance, but why can't they take care of my dog?" Trust me, I got asked by my parents about this more than you think ... The "Robot Moravec's paradox" also creates the illusion that physical AI capabilities are way more advanced than they truly are. I'm not singling out Unitree, as it applies widely to all recent acrobatic demos in the industry. Here's a simple test: if you set up a wall in front of the side-flipping robot, it will slam into it at full force and make a spectacle. Because it's just overfitting that single reference motion, without any awareness of the surroundings. Here's why the paradox exists: it's much easier to train a "blind gymnast" than a robot that sees and manipulates. The former can be solved entirely in simulation and transferred zero-shot to the real world, while the latter demands extremely realistic rendering, contact physics, and messy real-world object dynamics - none of which can be simulated well. Imagine you can train LLMs not from the internet, but from a purely hand-crafted text console game. Roboticists got lucky. We happen to live in a world where accelerated physics engines are so good that we can get away with impressive acrobatics using literally zero real data. But we haven't yet discovered the same cheat code for general dexterity. Till then, we'll still get questioned by our confused parents.

I'm observing a mini Moravec's paradox within robotics: gymnastics that are difficult for humans are much easier for robots than "unsexy" tasks like cooking, cleaning, and assembling. It leads to a cognitive dissonance for people outside the field, "so, robots can parkour & breakdance, but why can't they take care of my dog?" Trust me, I got asked by my parents about this more than you think ... The "Robot Moravec's paradox" also creates the illusion that physical AI capabilities are way more advanced than they truly are. I'm not singling out Unitree, as it applies widely to all recent acrobatic demos in the industry. Here's a simple test: if you set up a wall in front of the side-flipping robot, it will slam into it at full force and make a spectacle. Because it's just overfitting that single reference motion, without any awareness of the surroundings. Here's why the paradox exists: it's much easier to train a "blind gymnast" than a robot that sees and manipulates. The former can be solved entirely in simulation and transferred zero-shot to the real world, while the latter demands extremely realistic rendering, contact physics, and messy real-world object dynamics - none of which can be simulated well. Imagine you can train LLMs not from the internet, but from a purely hand-crafted text console game. Roboticists got lucky. We happen to live in a world where accelerated physics engines are so good that we can get away with impressive acrobatics using literally zero real data. But we haven't yet discovered the same cheat code for general dexterity. Till then, we'll still get questioned by our confused parents.

Jim Fan

398,883 просмотров • 1 год назад

When we started Score, the standard computer vision tools already existed. About a million people use them every day. Most of those people are still waiting on labels, running training jobs by hand, and watching models fail once they leave the test set. Most of those people are still waiting on labels, running training jobs by hand, and watching models fail once they leave the test set. Most of those people are also still waiting on verified computer vision models, evaluated against real life conditions and ready to be deployed for them to deliver value for their teams, clients or users. Score Studio is the full computer vision path in one place. A team describes the problem. The system can generate the missing scenes, label them, train the candidates, evaluate which ones actually hold, and deploy the winner. Data, labels, training, eval, ship. One loop. If no model exists for that job yet, they can put a bounty on the subnet. Anything from a small vision brick to a full VLM. Miners compete on the task. Only the winning work comes back. Same path for software agents. Any agent can call it. Built to be fully agent-accessible. Built for the people who already do this work: computer vision engineers and the small teams around them in plants, warehouses, farms, robotics, sport, and security. And for the agents those teams will run. That is the part that changes the job. Not another training screen. The stretch that used to take a lab and a calendar, footage, boxes, versions, failed runs, a separate deploy project, sits behind one starting point. And if the network needs a new model, that request is part of the same path. We spent more than a year building it. Then we had a choice. Keep it for us, or commoditize the whole subnet and make it available 24/7, in permissionless and open-source way. And we knew we couldn't keep it for us. It had to live on Bittensor. Open source software already showed how this should work. Infrastructure should not sit inside one company. Same idea as open AI before the phrase changed meaning: inspect it, fork it, keep building. That is what SN44 is for. Open vision intelligence, powered by Bittensor. Miners do the work. Studio is how that gets monetized. Profit does not stay in a company account. It goes back into the subnet through buyback and burn. We built the tool we wanted on day one. It will live on the network now, and for ever. Waitlist is open.

When we started Score, the standard computer vision tools already existed. About a million people use them every day. Most of those people are still waiting on labels, running training jobs by hand, and watching models fail once they leave the test set. Most of those people are still waiting on labels, running training jobs by hand, and watching models fail once they leave the test set. Most of those people are also still waiting on verified computer vision models, evaluated against real life conditions and ready to be deployed for them to deliver value for their teams, clients or users. Score Studio is the full computer vision path in one place. A team describes the problem. The system can generate the missing scenes, label them, train the candidates, evaluate which ones actually hold, and deploy the winner. Data, labels, training, eval, ship. One loop. If no model exists for that job yet, they can put a bounty on the subnet. Anything from a small vision brick to a full VLM. Miners compete on the task. Only the winning work comes back. Same path for software agents. Any agent can call it. Built to be fully agent-accessible. Built for the people who already do this work: computer vision engineers and the small teams around them in plants, warehouses, farms, robotics, sport, and security. And for the agents those teams will run. That is the part that changes the job. Not another training screen. The stretch that used to take a lab and a calendar, footage, boxes, versions, failed runs, a separate deploy project, sits behind one starting point. And if the network needs a new model, that request is part of the same path. We spent more than a year building it. Then we had a choice. Keep it for us, or commoditize the whole subnet and make it available 24/7, in permissionless and open-source way. And we knew we couldn't keep it for us. It had to live on Bittensor. Open source software already showed how this should work. Infrastructure should not sit inside one company. Same idea as open AI before the phrase changed meaning: inspect it, fork it, keep building. That is what SN44 is for. Open vision intelligence, powered by Bittensor. Miners do the work. Studio is how that gets monetized. Profit does not stay in a company account. It goes back into the subnet through buyback and burn. We built the tool we wanted on day one. It will live on the network now, and for ever. Waitlist is open.

Score

11,292 просмотров • 1 месяц назад

🚨BREAKING: The man who runs the world's largest UFO archive just got access to classified Swedish military files, has radar-confirmed UFO intercepts from inside DOD records, personally viewed secret military radar tracking an unknown object, and revealed that Betty Hill collected crash debris before her famous abduction that may still be buried in her yard in New Hampshire. Clas Svahn has spent 50 years building Archives for the Unexplained in Sweden. Sixteen rooms. 22,000 case files from Sweden alone. He's not a believer. He's a researcher who was named Educator of the Year in Sweden, an amateur astronomer who debunked his first major case at 16 by identifying Jupiter. We sat down at AFU and laid out what five decades of methodical fieldwork have actually produced. The evidence is staggering. 1957: A Car Dies, Hot Tungsten Found on the Road Two carpenters were driving on the island of Värmdö, northeast of Stockholm, when an object came in from the east, moved in front of their car, made a U-turn. The car went completely dead. The object left. They got out and found a metallic piece on the road so hot it burned their hands. Clas had it analyzed. Pure tungsten, sourced from Karabaj, with all expected impurities for that era. Tungsten is one of the best heat conductors on Earth. In the middle of a Swedish night, it should have been cold. Something made it extremely hot. The physical evidence matches the witness account exactly. 1975: Helicopter Pilot Ordered to Intercept a Ghost Rocket A Swedish military helicopter pilot, on standby for unknown objects crossing from Norway into Sweden at night, was ordered airborne to intercept. He and his co-pilot were flying 20 meters above the treetops. Moonlight on snow gave perfect visibility. An elongated, rocket-shaped object with no wings, no lights, no markings flew directly beneath the helicopter through that 20-meter gap over the trees. The pilot lifted his feet off the floor. It passed that close. They landed. Military security personnel debriefed them immediately. Clas Found the Radar Plot in Classified DOD Files Clas recently received clearance to access classified Swedish customs police military files from the early 1970s, sealed until 2040. Inside, he found the documentation for the 1975 helicopter intercept. The radar plot shows the exact point where the unknown object's path intersected with the helicopter. A military note confirms an object passed extremely close to the aircraft. Clas called the pilot days ago to confirm the date. The co-pilot, now living in Australia, will be interviewed next. The full files will be scanned and released within weeks. Six Radar Operators Watched It, Then the Photos Disappeared In the winter of 1973-74, six military radar operators stationed inside a mountain in northern Sweden came to the surface for lunch. They saw a cigar-shaped object moving over the treetops. They ran back underground to their radar equipment. The object appeared on screen, executing 90-degree turns before flying over Norway and straight up. Their commanding officer ordered them to photograph the radar screen. They did. Clas tracked down all six operators over several years. Every one of them told the same story. The photographs have never been found. 2005: A Phone Photo Matches Military Radar Returns Two men in a cottage in northern Sweden heard a strange noise late at night. They went outside. A brightly illuminated object was circling their cottage. One of them took a photo with his mobile phone. The first known mobile phone UFO photograph in Sweden. Clas went to the military radar unit covering that area. He personally viewed the radar returns. The object's movement on radar matched the witnesses' account exactly: approach, circling, departure. Two witnesses. One photograph. Military radar confirmation. Clas saw it with his own eyes. Every Scandinavian UFO Crash Has Been in Water Not a single UFO in Sweden or Norway, from the 1946 ghost rockets to the present, has crashed or landed on solid ground. Every one went into a lake. Roughly 30 cases. Always water. Almost always in July. Almost always around 11 PM. Clear weather. Hot weather. The Swedish military searched multiple lakes in 1946 for ghost rocket debris. They found indentations at the bottom. No wreckage. No fragments. Nothing. Objects that fly through space and navigate with apparent precision do not accidentally crash into lakes with that kind of consistency. Something Is Sitting in a Lake in Northern Sweden In 1980, witnesses in Dämma Jaure in far northern Sweden saw an object descend below 100 meters and sink into a lake. They photographed it two minutes after impact. They contacted the military. A helicopter was dispatched. One passenger became ill. Dead fish appeared near the shore. Clas and his team located sonar returns from an object resting not at the lake bottom but embedded two meters into the mud, exactly where their expert predicted it would be. They cannot retrieve it. It sits inside a protected national park. The object is still there. Betty Hill Collected Crash Debris Before Her Abduction Betty Hill told Clas directly that before the 1961 Indian Head encounter, she was already deeply interested in UFOs. She and a relative were sitting on her porch when something crossed the sky and crashed into a nearby field. They went out and brought back debris. She stored it in her cupboard. Days before the famous trip with Barney to Canada, Barney told her to get rid of it. She threw it in the garden. A lorry came shortly after and dumped earth over it. The debris is likely still buried at her former home in Portsmouth, New Hampshire. Clas recorded this conversation. The audio exists on AFU's website. 10% of Swedes Surveyed Have Seen Something Clas and his team knocked on doors and interviewed 1,600 Swedes. 10% reported seeing something unexplained. That extrapolates to roughly one million people in a country of 11 million. In Hessdalen, Norway, the figure inverts. 80 to 90% of residents have had experiences. One woman wrote to Clas recently to describe an encounter from October 1971 she had never told anyone outside her family. Two silver-suited figures with tight helmets, small heads, gloves, and belt-mounted boxes, standing eight meters from her on a metal scrap pile, appearing to teleport across the debris. The Best UFO Photo May Not Exist Clas has probably examined more UFO photographs than anyone alive. He called every Swedish photographer from the 1970s. All young men. Every single one eventually admitted they faked it, except one devout Christian named Krista Sundstrom who maintains his story to this day. Clas calls him every two years. The McMinnville photo, long considered the gold standard, may show a visible string under data enhancement. The only photograph Clas fully trusts is the 2005 northern Sweden mobile phone image, because he personally verified it against military radar. Why This Matters Clas Svahn is not speculating. He has radar documentation from classified military files. He has firsthand testimony from pilots, radar operators, and witnesses recorded over decades. He has physical trace evidence analyzed in labs. He has a sonar return from an object embedded in a Swedish lake. He has Betty Hill on tape describing crash debris that no one in UFOlogy has pursued. What makes this different from most UFO testimony is the methodology. Clas treated every case like an investigation, not an argument. He debunked what he could. What survived is harder to dismiss than almost anything in the public record. Full episode documents all of this and more.

🚨BREAKING: The man who runs the world's largest UFO archive just got access to classified Swedish military files, has radar-confirmed UFO intercepts from inside DOD records, personally viewed secret military radar tracking an unknown object, and revealed that Betty Hill collected crash debris before her famous abduction that may still be buried in her yard in New Hampshire. Clas Svahn has spent 50 years building Archives for the Unexplained in Sweden. Sixteen rooms. 22,000 case files from Sweden alone. He's not a believer. He's a researcher who was named Educator of the Year in Sweden, an amateur astronomer who debunked his first major case at 16 by identifying Jupiter. We sat down at AFU and laid out what five decades of methodical fieldwork have actually produced. The evidence is staggering. 1957: A Car Dies, Hot Tungsten Found on the Road Two carpenters were driving on the island of Värmdö, northeast of Stockholm, when an object came in from the east, moved in front of their car, made a U-turn. The car went completely dead. The object left. They got out and found a metallic piece on the road so hot it burned their hands. Clas had it analyzed. Pure tungsten, sourced from Karabaj, with all expected impurities for that era. Tungsten is one of the best heat conductors on Earth. In the middle of a Swedish night, it should have been cold. Something made it extremely hot. The physical evidence matches the witness account exactly. 1975: Helicopter Pilot Ordered to Intercept a Ghost Rocket A Swedish military helicopter pilot, on standby for unknown objects crossing from Norway into Sweden at night, was ordered airborne to intercept. He and his co-pilot were flying 20 meters above the treetops. Moonlight on snow gave perfect visibility. An elongated, rocket-shaped object with no wings, no lights, no markings flew directly beneath the helicopter through that 20-meter gap over the trees. The pilot lifted his feet off the floor. It passed that close. They landed. Military security personnel debriefed them immediately. Clas Found the Radar Plot in Classified DOD Files Clas recently received clearance to access classified Swedish customs police military files from the early 1970s, sealed until 2040. Inside, he found the documentation for the 1975 helicopter intercept. The radar plot shows the exact point where the unknown object's path intersected with the helicopter. A military note confirms an object passed extremely close to the aircraft. Clas called the pilot days ago to confirm the date. The co-pilot, now living in Australia, will be interviewed next. The full files will be scanned and released within weeks. Six Radar Operators Watched It, Then the Photos Disappeared In the winter of 1973-74, six military radar operators stationed inside a mountain in northern Sweden came to the surface for lunch. They saw a cigar-shaped object moving over the treetops. They ran back underground to their radar equipment. The object appeared on screen, executing 90-degree turns before flying over Norway and straight up. Their commanding officer ordered them to photograph the radar screen. They did. Clas tracked down all six operators over several years. Every one of them told the same story. The photographs have never been found. 2005: A Phone Photo Matches Military Radar Returns Two men in a cottage in northern Sweden heard a strange noise late at night. They went outside. A brightly illuminated object was circling their cottage. One of them took a photo with his mobile phone. The first known mobile phone UFO photograph in Sweden. Clas went to the military radar unit covering that area. He personally viewed the radar returns. The object's movement on radar matched the witnesses' account exactly: approach, circling, departure. Two witnesses. One photograph. Military radar confirmation. Clas saw it with his own eyes. Every Scandinavian UFO Crash Has Been in Water Not a single UFO in Sweden or Norway, from the 1946 ghost rockets to the present, has crashed or landed on solid ground. Every one went into a lake. Roughly 30 cases. Always water. Almost always in July. Almost always around 11 PM. Clear weather. Hot weather. The Swedish military searched multiple lakes in 1946 for ghost rocket debris. They found indentations at the bottom. No wreckage. No fragments. Nothing. Objects that fly through space and navigate with apparent precision do not accidentally crash into lakes with that kind of consistency. Something Is Sitting in a Lake in Northern Sweden In 1980, witnesses in Dämma Jaure in far northern Sweden saw an object descend below 100 meters and sink into a lake. They photographed it two minutes after impact. They contacted the military. A helicopter was dispatched. One passenger became ill. Dead fish appeared near the shore. Clas and his team located sonar returns from an object resting not at the lake bottom but embedded two meters into the mud, exactly where their expert predicted it would be. They cannot retrieve it. It sits inside a protected national park. The object is still there. Betty Hill Collected Crash Debris Before Her Abduction Betty Hill told Clas directly that before the 1961 Indian Head encounter, she was already deeply interested in UFOs. She and a relative were sitting on her porch when something crossed the sky and crashed into a nearby field. They went out and brought back debris. She stored it in her cupboard. Days before the famous trip with Barney to Canada, Barney told her to get rid of it. She threw it in the garden. A lorry came shortly after and dumped earth over it. The debris is likely still buried at her former home in Portsmouth, New Hampshire. Clas recorded this conversation. The audio exists on AFU's website. 10% of Swedes Surveyed Have Seen Something Clas and his team knocked on doors and interviewed 1,600 Swedes. 10% reported seeing something unexplained. That extrapolates to roughly one million people in a country of 11 million. In Hessdalen, Norway, the figure inverts. 80 to 90% of residents have had experiences. One woman wrote to Clas recently to describe an encounter from October 1971 she had never told anyone outside her family. Two silver-suited figures with tight helmets, small heads, gloves, and belt-mounted boxes, standing eight meters from her on a metal scrap pile, appearing to teleport across the debris. The Best UFO Photo May Not Exist Clas has probably examined more UFO photographs than anyone alive. He called every Swedish photographer from the 1970s. All young men. Every single one eventually admitted they faked it, except one devout Christian named Krista Sundstrom who maintains his story to this day. Clas calls him every two years. The McMinnville photo, long considered the gold standard, may show a visible string under data enhancement. The only photograph Clas fully trusts is the 2005 northern Sweden mobile phone image, because he personally verified it against military radar. Why This Matters Clas Svahn is not speculating. He has radar documentation from classified military files. He has firsthand testimony from pilots, radar operators, and witnesses recorded over decades. He has physical trace evidence analyzed in labs. He has a sonar return from an object embedded in a Swedish lake. He has Betty Hill on tape describing crash debris that no one in UFOlogy has pursued. What makes this different from most UFO testimony is the methodology. Clas treated every case like an investigation, not an argument. He debunked what he could. What survived is harder to dismiss than almost anything in the public record. Full episode documents all of this and more.

Jesse Michels

321,851 просмотров • 7 месяцев назад