Insights from global conversations
OpenAI shares insights from conversations conducted across 22 countries and outlines how these findings will shape its forward strategy.
21 articles
OpenAI shares insights from conversations conducted across 22 countries and outlines how these findings will shape its forward strategy.
Hugging Face optimizes BridgeTower, a vision-language model, on Habana Gaudi2 accelerators. The implementation improves inference and training performance for multimodal tasks.
OpenAI announces the opening of its first international office in London, United Kingdom.
Hugging Face's fourth Ethics and Society newsletter examines bias in text-to-image models. The article analyzes how these systems reproduce and amplify stereotypes from training data, with implications for fairness and representation.
Hugging Face announces major changes to its Open LLM Leaderboard: revision of evaluation criteria, improved transparency and reproducibility of benchmarks. The goal is to align rankings with real user needs and reduce score gaming.
Hugging Face responds to the U.S. NTIA's public consultation on AI accountability. The organization advocates for open standards, model transparency, and safeguards against misuse, while opposing regulatory barriers that would favor large corporations.
Hugging Face releases a guide to fine-tune MMS adapter models for low-resource languages. The approach reduces training data and computational requirements while maintaining speech recognition quality.
Study demonstrates transformers are effective for time series forecasting. Introduces Autoformer, a specialized transformer model that outperforms traditional methods on multiple forecasting benchmarks.
Hugging Face optimizes Stable Diffusion for Core ML, enabling fast inference on iPhone, iPad, and Mac. The implementation leverages Apple hardware accelerators to reduce latency and memory consumption.
Hugging Face enables deploying Livebook notebooks directly to Spaces. Elixir developers can create interactive applications without additional setup, with real-time sharing and collaboration support.
Hugging Face announces updated content guidelines and policy for its platform. Guidelines cover models, datasets, and spaces with clear moderation and compliance criteria.
OpenAI announces API updates: more steerable models, function calling capabilities, longer context, and lower prices.
Hugging Face and AMD announce a partnership to optimize AI models on AMD CPUs and GPUs. The initiative aims to improve inference performance and accessibility of state-of-the-art models across AMD's ecosystem.
Hugging Face investigates whether foundation models can label data with human-level quality. The article examines automated labeling capabilities and compares performance against human annotators.
Hugging Face launches an initiative to integrate the Hub into galleries, libraries, archives and museums. The platform enables cultural institutions to access AI models and datasets, democratizing AI adoption in the heritage sector.
DuckDB now natively accesses 50,000+ datasets on Hugging Face Hub. Users can query parquet data directly without downloading via simple SQL API.
Hugging Face integrates fastText into its Hub. Meta's open-source platform for text embedding and classification is now accessible through the Hugging Face ecosystem, enabling easier sharing and deployment of fastText models.
Falcon, an open-source language model developed by the Technology Innovation Institute (TII), joins the Hugging Face ecosystem. The model is available on the Hub with full support for inference, fine-tuning, and deployment via Inference API.
Hugging Face integrates AI speech recognition into Unity through a new extension. Developers can implement speech-to-text models directly in games and applications without cloud infrastructure.
OpenAI launches a grant program to develop AI-powered cybersecurity capabilities for defenders. The initiative aims to support researchers and organizations working on AI-based security tools.
Hugging Face announces an open source AI game jam. The event invites developers and creators to build games using open source models and community tools.