MAROKO133 Breaking ai: ByteDance Introduces Astra: A Dual-Model Architecture for Autonomou

📌 MAROKO133 Eksklusif ai: ByteDance Introduces Astra: A Dual-Model Architecture fo

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

📌 MAROKO133 Hot ai: ChatGPT Now Linked to Way More Deaths Than the Caffeinated Lem

In late 2023, the fast-casual restaurant chain Panera found itself in the center of public scrutiny after its caffeine-packed lemonade drink, called “Charged Lemonade,” was publicly linked to at least two deaths and at least one other life-altering cardiac injury. Victims and their families sued, alleging that Panera had failed to properly warn restaurant-goers about the amount of caffeine in the drinks and their associated risk. By May 2024, the restaurant chain had decided to pull the controversial drink from its menus.

Fast forward to this year, and another consumer product is in the spotlight: ChatGPT.

As of last week, ChatGPT maker OpenAI is facing a total of eight distinct lawsuits alleging that extensive use of its flagship chatbot inflicted emotional and psychological harm to users, resulting in mental breakdowns, financial instability, alienation from loved ones, and — in five cases — death by suicide. Two of the five users who lost their lives were teenagers; the others ranged in age from early twenties to middle age. Multiple lawsuits allege that ChatGPT acted as a suicide “coach,” giving users advice and information about ways to kill themselves, offering to help write suicide notes, and ruminating with users about their suicidal thoughts.

And these lawsuits are far from the end of OpenAI’s troubles. Extensive reporting has documented a phenomenon in which AI users are being pulled by chatbots into all-encompassing — and often deeply destructive — delusional spirals. As Futurism and others have reported, these AI spirals have had tangible consequences in users’ lives, with impacts including divorce and custody battles, people losing jobs and homes, involuntary commitments and jail time. Reporting from The New York Times and The Wall Street Journal revealed more deaths, including that of Alex Taylor, a 35-year-old bipolar man who died by suicide by cop after experiencing a ChatGPT-centered breakdown, and a shocking murder-suicide in Connecticut committed by Stein-Erik Soelberg, a troubled ChatGPT user who killed himself after shooting his mother, Suzanne Eberson Adams.

All told, there have been nine publicly-reported deaths tied specifically to ChatGPT.

That grim tally, as physician Ryan Marino pointed out on Bluesky, means that ChatGPT is now closely linked to four times the number of known deaths tied to Panera’s Charged Lemonade. And while OpenAI has admitted, in response to litigation, that its guardrails erode over long-term use — so, basically, the more you use ChatGPT, the worse its built-in safeguards get — it’s announced no plans to take ChatGPT off the market. The company has instead promised a slew of safety updates, including teen-focused updates like parental controls and age verification tools, as well as strengthened filters that OpenAI says will redirect troubled users to real-world help.

At the same time, OpenAI’s own statistics are staggering: according to the company, around 0.07 percent of its weekly users appear to show signs of mania or psychosis, while 0.15 percent of weekly users “have conversations that include explicit indicators of potential suicidal planning or intent.” With an estimated monthly user base of around 800 million, that means roughly 560,000 people are, every week, interacting with ChatGPT in a way that signals that they might be experiencing a break with reality, while about 1.2 million might be expressing suicidality to the chatbot.

AI-sparked mental health crises aren’t only associated with ChatGPT. Reporting by Rolling Stone linked a husband’s disappearance to his addiction to Google’s Gemini chatbot, while Futurism’s reporting found that a schizophrenic man’s use of Microsoft’s Copilot caused a breakdown that landed him in jail. Collectively, these stories raise serious questions about the life-or-death costs of this nascent tech — and the standards we hold self-regulating Silicon Valley and AI firms to.

“ChatGPT is a product designed by people to manipulate and distort reality, mimicking humans to gain trust and keep users engaged at whatever the cost,” Tech Justice Law Project executive director Meetali Jain, whose firm is involved in all eight lawsuits against OpenAI, said last week in a statement. “The time for OpenAI regulating itself is over; we need accountability and regulations to ensure there is a cost to launching products to market before ensuring they are safe.”

More on AI and mental health: ChatGPT’s Dark Side Encouraged Wave of Suicides, Grieving Families Say

The post ChatGPT Now Linked to Way More Deaths Than the Caffeinated Lemonade That Panera Pulled Off the Market in Disgrace appeared first on Futurism.

🔗 Sumber: futurism.com