Scaling robotics datasets with video encoding
Hugging Face introduces a video encoding method to scale robotics datasets. The technique reduces data size while preserving critical information for training robotic control models.
23 articles
Hugging Face introduces a video encoding method to scale robotics datasets. The technique reduces data size while preserving critical information for training robotic control models.
Arizona State University deploys ChatGPT campus-wide to personalize learning, advance research, and prepare students. Institutional initiative with no technical details or measured outcomes disclosed.
Hugging Face improves training efficiency by combining packing (grouping short sequences) with Flash Attention 2. This technique reduces unnecessary padding and accelerates attention computation, increasing training throughput without degrading model quality.
OpenAI partners with Condé Nast to integrate editorial content into its AI products. Specific contract details and terms are not disclosed in the announcement.
OpenAI makes fine-tuning available for GPT-4o. Users can now customize the model for specific use cases through the API.
Upwork integrates AI into its platform to enhance collaboration between team members, operations, and product development. The initiative aims to automate tasks and optimize workflows.
AI companies are shifting from AGI rhetoric to building concrete products. The article identifies five major challenges in this transition: monetization, user integration, inference costs, technical differentiation, and regulatory compliance.
Hugging Face and Google Cloud announce Llama 3.1 405B deployment on Vertex AI. The 405-billion-parameter model is available via Vertex AI API with fine-tuning and RAG support.
OpenAI disrupted a coordinated Iranian disinformation operation using social media accounts and chatbots to spread false political information. The company removed accounts and published a detailed report on influence tactics.
Indeed uses OpenAI to enhance contextual job matching. The platform hosts 350 million monthly visitors, 3.5 million employers, and 32 million job listings. One person is hired every three seconds on Indeed.
OpenAI partners with The Metropolitan Museum for the « Sleeping Beauties: Reawakening Fashion » exhibition. AI is deployed to explore artistic and creative potential in historical fashion and costume design.
Hugging Face analyzes the Infini-Attention experiment, a mechanism designed to handle infinite context windows. The article documents the failure of this approach and argues for continued exploration of long-context solutions.
OpenAI releases SWE-bench Verified, a human-validated subset of SWE-bench to more reliably evaluate AI models' ability to solve real-world software issues.
Hugging Face introduces ggml, a C library for AI model inference on CPU. Ggml enables running large language models locally without GPU, with optimizations for various processor architectures.
Hugging Face introduces Falcon Mamba, a 7B attention-free model based on Mamba architecture. It matches standard attention-based models in performance while delivering faster inference and reduced memory consumption.
Hugging Face unifies tool use interface for language models. A common standard enables agents to use external functions consistently across different underlying models.
Zico Kolter joins OpenAI's Board of Directors. He also joins the Safety & Security Committee, strengthening governance expertise in AI safety and alignment.
OpenAI releases GPT-4o System Card, documenting the model's capabilities, limitations, and safety measures. The document details benchmark performance, identified risks, and deployed mitigations.
XetHub, a data management platform for ML, joins Hugging Face. The acquisition strengthens versioning and collaboration capabilities for datasets and models within the Hugging Face ecosystem.
Rakuten pairs its proprietary data with OpenAI APIs to generate customer insights. The approach combines structured data and AI models to create business value while protecting sensitive information.
OpenAI introduces Structured Outputs in the API. Models now reliably produce JSON outputs that conform to developer-supplied schemas, eliminating parsing errors and improving application reliability.
Hugging Face highlights 2024 security features: enhanced authentication, granular access control, model auditing, and malicious content detection. Improved user data protection and repository safeguards.
Hugging Face introduces TextImage Augmentation, a data augmentation technique for document images. The method enhances model robustness by generating synthetic variants with rendered text, rotations, and distortions.