AI Digest
← All articles

Research

12 articles

DeepMind Unveils Decoupled DiLoCo for Fault-Tolerant Distributed AI Training

DeepMind has introduced Decoupled DiLoCo, an advanced distributed training method designed to make large-scale AI model

DeepMind · May 8, 2026

DeepMind Unveils AI Co-Clinician to Transform Healthcare Delivery

DeepMind is pioneering a new approach to healthcare through the development of an AI co-clinician system. This research

DeepMind · May 8, 2026

IBM's Granite 4.1 LLMs: Inside the Architecture and Training Process

IBM has unveiled Granite 4.1, the latest iteration of its large language model family designed specifically for enterpri

HuggingFace · May 8, 2026

HuggingFace Adds Anti-Gaming Measures to Open ASR Leaderboard

HuggingFace has introduced new safeguards to its Open Automatic Speech Recognition (ASR) Leaderboard to prevent what the

HuggingFace · May 8, 2026

AMD ROCm Enables CUDA-Free Medical AI Training with MedQA Dataset

Researchers have successfully fine-tuned a clinical AI model using AMD's ROCm platform instead of NVIDIA's CUDA framewor

HuggingFace · May 8, 2026

EMO: New Pretraining Method Unlocks Emergent Modularity in AI Experts

Researchers have introduced EMO, a novel pretraining approach for mixture of experts (MoE) models that enables emergent

HuggingFace · May 8, 2026

OpenAI Releases MRC Protocol to Enhance AI Supercomputer Networks

OpenAI has announced the release of MRC (Multipath Reliable Connection), a new networking protocol designed specifically

OpenAI · May 6, 2026

Google DeepMind Launches Project Genie for AI-Generated Interactive Worlds

Google DeepMind has unveiled Project Genie, an experimental research prototype that enables users to create and explore

DeepMind · May 6, 2026

DeepMind Unveils Gemini Deep Think to Accelerate Scientific Research

DeepMind has announced Gemini Deep Think, a new AI system designed to advance mathematical and scientific discovery. The

DeepMind · May 6, 2026

DeepMind Unveils Gemini 3 Deep Think for Advanced Scientific Problem-Solving

DeepMind has announced Gemini 3 Deep Think, an updated version of its most specialized reasoning mode designed specifica

DeepMind · May 6, 2026

Decoupled DiLoCo: DeepMind Advances Distributed AI Training with New Resilience Framework

DeepMind has announced Decoupled DiLoCo, an enhanced version of its Distributed Low-Communication (DiLoCo) training meth

DeepMind · May 6, 2026

HuggingFace Releases Training Tools for Multimodal Embedding and Reranker Models

HuggingFace has announced new capabilities in its Sentence Transformers library that enable developers to train and fine

HuggingFace · May 6, 2026