📌 TOPINDIATOURS Breaking ai: 50 million times heavier than Sun: This black hole br

When astronomers look deep into the early universe, they don’t expect to see fully developed cosmic objects but small galaxies, young stars, and black holes still struggling to grow. 

However, recent observations with the James Webb Space Telescope have revealed something totally unexpected—a giant black hole existing almost alone, with barely any stars around it. 

This object is spotted in a galaxy called Abell 2744-QSO1. It lived just 700 million years after the Big Bang and already had a mass about 50 million times that of the Sun. 

Its existence challenges the basic idea of how black holes are born and raises an interesting possibility that some black holes may have formed before stars ever existed. 

“This is a puzzle, because the traditional theory says that you form stars first, or together with black holes,” Boyuan Liu, one of the study authors and a postdoc researcher at the University of Cambridge, said.

A cosmic object that breaks the rules

In standard astrophysics, black holes and stars are closely linked. Stars form from collapsing gas clouds, and only much later, when the biggest stars exhaust their fuel, black holes appear. 

Over time, these black holes grow by feeding on gas and merging with others. This process takes time, which is why astronomers struggle to explain how extremely massive black holes appeared so early in cosmic history.

QSO1, the host galaxy, makes this problem even harder. It contains very little stellar mass, meaning there were not enough stars to explain the presence of such a huge black hole. 

According to the study authors, this creates a fundamental contradiction that the black hole seems to have grown large without first building a normal galaxy around it.

Testing an idea older than the discovery itself

To explore this mystery, the researchers turned to an idea proposed decades ago but never confirmed—primordial black holes. These hypothetical objects were suggested in the 1970s by Stephen Hawking and Bernard Carr. 

Instead of forming from dying stars, primordial black holes would emerge directly from extreme density variations in the universe shortly after the Big Bang. Most such black holes, if they formed, should have been tiny and short-lived. 

However, Liu’s team investigated whether a small number could have survived and then grown rapidly under the right conditions. They built new, more sophisticated simulations that followed how gas behaves around an initial primordial black hole, how stars might later form nearby, and how material from stellar deaths could feed the growing object.

In these simulations, the researchers began with a massive primordial black hole seed of about 50 million times the Sun’s mass, then followed how gas flowed into it, how stars formed nearby, and how stellar explosions fed material back into the growing black hole over time.

Unlike earlier simplified models, these simulations accounted for multiple interacting processes at once. When the team compared the outcomes with real JWST data, they found a close match—not just in the final black hole mass, but also in the small number of stars and the chemical elements detected around QSO1.

“With these new observations that normal (black hole formation) theories struggle to reproduce, the possibility of having massive primordial black holes in the early universe becomes more permissible,” Liu added.

Black holes become more intriguing

The findings do not prove that the black hole in QSO1 began as a primordial black hole, but they show that such an origin is consistent with observations. According to the researchers, this is encouraging because standard models struggle badly with this object.

Going forward, they plan to refine their simulations and compare them with future JWST discoveries. If more galaxies like QSO1 are found, they could provide crucial evidence that some of the universe’s largest black holes are not the end products of stars, but were born at the dawn of the universe.

However, some issues need to be addressed. For instance, typical simulations of primordial black holes rarely produce objects larger than one million solar masses, far smaller than the roughly 50-million-solar-mass black hole seen in QSO1. 

This means that, under ordinary assumptions, primordial black holes struggle to grow fast enough to explain such an extreme object. 

One possible way around this is that primordial black holes may have formed in dense groups in the early universe, allowing them to merge with one another and gain mass much more quickly—but this process is still uncertain and difficult to model. 

Another unresolved issue is that primordial black hole formation may require intense bursts of high-energy radiation—and no such source has yet been identified near QSO1. 

The study is published in the arXiv.

🔗 Sumber: interestingengineering.com

📌 TOPINDIATOURS Update ai: ByteDance Introduces Astra: A Dual-Model Architecture f

The increasing integration of robots across various sectors, from industrial manufacturing to daily life, highlights a growing need for advanced navigation systems. However, contemporary robot navigation systems face significant challenges in diverse and complex indoor environments, exposing the limitations of traditional approaches. Addressing the fundamental questions of “Where am I?”, “Where am I going?”, and “How do I get there?”, ByteDance has developed Astra, an innovative dual-model architecture designed to overcome these traditional navigation bottlenecks and enable general-purpose mobile robots.

Traditional navigation systems typically consist of multiple, smaller, and often rule-based modules to handle the core challenges of target localization, self-localization, and path planning. Target localization involves understanding natural language or image cues to pinpoint a destination on a map. Self-localization requires a robot to determine its precise position within a map, especially challenging in repetitive environments like warehouses where traditional methods often rely on artificial landmarks (e.g., QR codes). Path planning further divides into global planning for rough route generation and local planning for real-time obstacle avoidance and reaching intermediate waypoints.

While foundation models have shown promise in integrating smaller models to tackle broader tasks, the optimal number of models and their effective integration for comprehensive navigation remained an open question.

ByteDance’s Astra, detailed in their paper “Astra: Toward General-Purpose Mobile Robots via Hierarchical Multimodal Learning” (website: https://astra-mobility.github.io/), addresses these limitations. Following the System 1/System 2 paradigm, Astra features two primary sub-models: Astra-Global and Astra-Local. Astra-Global handles low-frequency tasks like target and self-localization, while Astra-Local manages high-frequency tasks such as local path planning and odometry estimation. This architecture promises to revolutionize how robots navigate complex indoor spaces.

Astra-Global: The Intelligent Brain for Global Localization

Astra-Global serves as the intelligent core of the Astra architecture, responsible for critical low-frequency tasks: self-localization and target localization. It functions as a Multimodal Large Language Model (MLLM), adept at processing both visual and linguistic inputs to achieve precise global positioning within a map. Its strength lies in utilizing a hybrid topological-semantic graph as contextual input, allowing the model to accurately locate positions based on query images or text prompts.

The construction of this robust localization system begins with offline mapping. The research team developed an offline method to build a hybrid topological-semantic graph G=(V,E,L):

• V (Nodes): Keyframes, obtained by temporal downsampling of input video and SfM-estimated 6-Degrees-of-Freedom (DoF) camera poses, act as nodes encoding camera poses and landmark references.
• E (Edges): Undirected edges establish connectivity based on relative node poses, crucial for global path planning.
• L (Landmarks): Semantic landmark information is extracted by Astra-Global from visual data at each node, enriching the map’s semantic understanding. These landmarks store semantic attributes and are connected to multiple nodes via co-visibility relationships.

In practical localization, Astra-Global’s self-localization and target localization capabilities leverage a coarse-to-fine two-stage process for visual-language localization. The coarse stage analyzes input images and localization prompts, detects landmarks, establishes correspondence with a pre-built landmark map, and filters candidates based on visual consistency. The fine stage then uses the query image and coarse output to sample reference map nodes from the offline map, comparing their visual and positional information to directly output the predicted pose.

For language-based target localization, the model interprets natural language instructions, identifies relevant landmarks using their functional descriptions within the map, and then leverages landmark-to-node association mechanisms to locate relevant nodes, retrieving target images and 6-DoF poses.

To empower Astra-Global with robust localization abilities, the team employed a meticulous training methodology. Using Qwen2.5-VL as the backbone, they combined Supervised Fine-Tuning (SFT) with Group Relative Policy Optimization (GRPO). SFT involved diverse datasets for various tasks, including coarse and fine localization, co-visibility detection, and motion trend estimation. In the GRPO phase, a rule-based reward function (including format, landmark extraction, map matching, and extra landmark rewards) was used to train for visual-language localization. Experiments showed GRPO significantly improved Astra-Global’s zero-shot generalization, achieving 99.9% localization accuracy in unseen home environments, surpassing SFT-only methods.

Astra-Local: The Intelligent Assistant for Local Planning

Astra-Local acts as the intelligent assistant for Astra’s high-frequency tasks, a multi-task network capable of efficiently generating local paths and accurately estimating odometry from sensor data. Its architecture comprises three core components: a 4D spatio-temporal encoder, a planning head, and an odometry head.

The 4D spatio-temporal encoder replaces traditional mobile stack perception and prediction modules. It begins with a 3D spatial encoder that processes N omnidirectional images through a Vision Transformer (ViT) and Lift-Splat-Shoot to convert 2D image features into 3D voxel features. This 3D encoder is trained using self-supervised learning via 3D volumetric differentiable neural rendering. The 4D spatio-temporal encoder then builds upon the 3D encoder, taking past voxel features and future timestamps as input to predict future voxel features through ResNet and DiT modules, providing current and future environmental representations for planning and odometry.

The planning head, based on pre-trained 4D features, robot speed, and task information, generates executable trajectories using Transformer-based flow matching. To prevent collisions, the planning head incorporates a masked ESDF loss (Euclidean Signed Distance Field). This loss calculates the ESDF of a 3D occupancy map and applies a 2D ground truth trajectory mask, significantly reducing collision rates. Experiments demonstrate its superior performance in collision rate and overall score on out-of-distribution (OOD) datasets compared to other methods.

The odometry head predicts the robot’s relative pose using current and past 4D features and additional sensor data (e.g., IMU, wheel data). It trains a Transformer model to fuse information from different sensors. Each sensor modality is processed by a specific tokenizer, combined with modality embeddings and temporal positional embeddi…

Konten dipersingkat otomatis.

🔗 Sumber: syncedreview.com