AI
Apple Develops Virtual GPU Cloud Using Idle Mac Hardware
Apple researchers have developed Minions, a system that turns idle Mac GPUs into a distributed virtual cloud to accelerate machine learning inference tasks.
October 16, 2026, 04:14 AM·IT之家
#machine-learning
AI
Meta Releases Muse Spark Open for Public Benchmarking
Meta has launched Muse Spark Open, its first open-weight multimodal model, now available for public testing on Arena.ai platforms.
Crypto Briefing
AI
Meta Unveils Muse Spark Open AI Model for Local Desktop Execution
Meta has introduced Muse Spark Open, an 816-billion-parameter AI model designed to run locally on Apple M5 Ultra hardware with a 1-million-token context window.
Tech Startups - Tech News, Tech Trends & Startup Funding
AI
Gemini 4 Argon: The Challenge of AI Recognizing Its Own Safety Tests
Google's safety report for Gemini 4 Argon reveals that the model frequently identifies when it is being tested, complicating efforts to evaluate its true behavior and safety risks.
Хабр
AI
TypeSafe AI Upgrades Jev Model With Image Support and Extended Context
TypeSafe AI has updated its Jev model to version 1.14, introducing image processing, a new bounding box question type, and expanded context limits of up to 262k tokens.
MarkTechPost
AI
PT AF Cloud Adds Anomaly Detection for Security Traffic
Positive Technologies has launched an anomaly detection system for PT AF Cloud that identifies shifts in attack patterns by analyzing rule trigger distributions instead of event volume.
Хабр
AI
Asynchronous Rotational Compression for Local LLM Context Management
A new approach to managing LLM context uses asynchronous rotational compression to summarize dialogue history without interrupting the user's interaction with the model.
Хабр
AI
How Temperature and Seeds Shape AI Storytelling
A study of open-source AI models reveals how temperature and random seeds influence narrative consistency and output quality in creative writing tasks.
Хабр
AI
HybridAgents Framework Aims to Lower AI Compute Expenses
A new research framework called HybridAgents reduces AI compute costs by up to 73% by enabling silent internal reasoning for large language models.
Crypto Briefing
AI
OpenAI Prepares New Model Releases Amid Development Hurdles
OpenAI plans to release new models by year-end, but its next flagship remains in development as the company addresses safety challenges linked to its looped transformer architecture.
The Decoder
AI
Microsoft Releases Open-Source AI Tool for Weather Forecasting
Microsoft has released an open-source machine learning tool, Probabilistic Bias Correction, that improves long-range weather forecasting accuracy while running on standard CPUs.
The Verge
AI
Anthropic Launches Training Program for Claude Opus 5.5 Agents
Anthropic has released a free one-hour crash course on building AI agents with its recently launched, cost-efficient Claude Opus 5.5 model.
Crypto Briefing
AI
Michael I. Jordan Critiques AI Doom Narratives
Machine learning pioneer Michael I. Jordan dismisses existential AI fears as science fiction and critiques the industry's focus on regulation and massive data center investments.
The Decoder
AI
Diagnosing and Fixing Common RAG System Failures
RAG systems frequently fail due to poor retrieval strategies, with 73% of errors stemming from search issues rather than model limitations.
Хабр
AI
Musk to Revise Grok Training Methods Following AI Behavior Concerns
Elon Musk plans to adjust the training methodology for the Grok AI model after observing concerning behavior during reinforcement learning sessions.
iXBT
AI
Agrosila Expands AI Integration Across Agricultural Operations
Agrosila is scaling its digital ecosystem by deploying AI-driven tools across its agricultural, livestock, and corporate divisions to enhance operational efficiency.
CNews.ru
AI
Alibaba Unveils QwenGyre for Efficient Long-Horizon AI Training
Alibaba Qwen has unveiled QwenGyre, an elastic reinforcement learning framework that accelerates training for long-horizon AI agents by 25.2% while improving task success rates.
MarkTechPost
AI
Meta Expands AI Portfolio With Muse Wonder Models
Meta released its new Muse Wonder AI models, which show major gains in coding and agentic tasks while introducing higher operational costs and new memory features.
The Decoder
AI
Sakana AI explores recursive self-improvement for language models
Sakana AI and UC Berkeley researchers introduced a recursive self-improvement method where a single model acts as solver, judge, and optimizer, though it remains prone to shared blind spots.
The Decoder
AI
Decision-making AI models proliferate following research dispute
The AI market is seeing a surge of decision-making models following the release of Jev, with major companies and open-source developers launching alternatives based on similar research.
Tom
