MAROKO133 Eksklusif ai: If OpenAI Loses This Trial, It Could Effectively Be Eliminated in

📌 MAROKO133 Hot ai: If OpenAI Loses This Trial, It Could Effectively Be Eliminated

Little love has been lost between Elon Musk and Sam Altman.

The two billionaires have been openly feuding for many years now, despite founding OpenAI together over a decade ago. Musk left the organization in 2019 over disagreements with leadership — and the falling out has only grown since then.

Musk has filed several lawsuits against the ChatGPT maker, most recently alleging it had breached its fiduciary duties by turning what was once a non-profit “charity” into a profit-maximizing corporate behemoth. OpenAI denies these allegations, arguing that Musk was “motivated by jealousy” after being pushed out for demanding to merge OpenAI with Tesla and assume majority control.

The stakes rose exponentially this week as a trial for Musk’s lawsuit against the company, initially filed in 2024, kicked off in an Oakland courthouse. Musk is demanding that OpenAI to undo its recent conversion into a for-profit entity, sack Altman and his board, and $130 billion in damages, which his lawyers refer to as “ill-gotten gains.”

The anxiously-awaited legal proceedings could have vast implications not just for OpenAI — which is rumored to be plotting an IPO — but for the AI industry as a whole. OpenAI is shackled to most of the industry’s biggest players through multibillion-dollar contracts, meaning that if it were to make concessions, lose its status as a for-profit company, and its CEO, it could send major ripple effects across an already shaky industry.

If Musk were to win — a decision that could materialize at the end of an estimated three weeks of legal proceedings — the effects could tear apart the already-widening cracks in Silicon Valley’s all-in, trillion dollar bet on AI, effectively popping the “AI bubble” which experts have warned about for years now.Analysts have long grown wary that the industry’s unprecedented levels of spending are making any hope of an eventual return on investment a long shot, meaning that a massive legal setback for OpenAI could therefore set off a powder keg of pent-up anxiety.

A loss for OpenAI could also set a dangerous legal precedent.

“The broader question of whether AI labs founded as charities can lawfully pivot into commercial enterprises would be settled, at least in California,” as University of Sydney media and communications researcher Rob Nicholls argues in a piece for The Conversation. “This has potential implications for Anthropic and other mission-driven peers.”

Besides, plenty of damage has already been done. The trial “has already pried open Silicon Valley’s normally sealed boardrooms, surfacing diaries, Slack threads and HR memos that paint an unflattering portrait of OpenAI’s governance,” Nicholls noted.

It’s an especially precarious time for OpenAI, which has desperately consolidated its efforts to focus on its core offerings, including ChatGPT and a coding tool, by killing off distracting “side quests.” The company is still bleeding billions of dollars every quarter, despite earmarking a whopping $600 billion in AI infrastructure expenses over the next four years.

Raking in enough cash to cover its astronomical spending spree could prove extremely difficult as well. As the Wall Street Journal reported this week, CFO Sarah Friar has warned internally that the company could fail to grow revenue fast enough to “pay for future computing contracts.” The company already missed its own internal user growth and revenue targets for 2025.

Where that leaves OpenAI’s plans for an IPO is anything but clear, with Friar reportedly butting heads with Altman on when to go public, a disagreement both denied in a statement to the WSJ.

Now that Musk and Altman’s bitter feud has broken into the public in entirely unprecedented ways, the winner of the ongoing trial could take it all.

“If Musk wins, it could result in the defeat of a key competitor in the race to AGI,” law professor and UCLA exec Rose Chan Loui told the BBC. “Whoever wins that race will have a lot of power.”

Musk has less to lose than Altman. But a victory could certainly bolster his position. He recently folded his (very much for-profit) AI startup xAI into his space company SpaceX, which is also expected to go public in the coming months at a record-shattering valuation of $1.75 trillion. By declaring victory over his archnemesis, Musk could aggregate even more influence over the industry.

At the same time, nobody knows what will be left of that industry if it turns out that OpenAI’s plans to shed its non-profit roots were illegitimate and Altman is forced to resign. If OpenAI were to fall, many other companies closely tied to its continued success could follow, a domino effect that could have a lasting impact “felt for many years to come,” per Nicholls.

A loss for OpenAI could hit an already exposed nerve as investors have long grown wary of an industry hellbent on spending as much money in as little time as possible without ever establishing a feasible, long-term business model. A high-stakes, popcorn-chomping “clash of the titans” lawsuit likely isn’t going to dispel that fear.

More on OpenAI: OpenAI in Shambles as IPO Looms

The post If OpenAI Loses This Trial, It Could Effectively Be Eliminated in Its Current Form appeared first on Futurism.

🔗 Sumber: futurism.com

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

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