MIT researchers developed a dynamic approach for large language models (LLMs) to allocate computational effort based on question difficulty, improving efficiency and accuracy. This method allows smaller LLMs to outperform larger models on complex problems, potentially reducing energy consumption and expanding applications.
Top 10 intelligent open-source models use MoE architecture, 10x faster on NVIDIA GB200 NVL72. MoE models like Kimi K2 Thinking and Mistral Large 3 offer breakthrough performance on NVIDIA systems, becoming the standard for frontier AI models.
Spotify Wrapped alienates us from our musical memories, urging us to embrace a corporate scorecard. Take back ownership of your taste and reflect on the records that truly resonate with you.
Anthropic's chief scientist warns of AI autonomy's potential 'intelligence explosion' by 2030. Humanity faces a crucial decision on granting self-training power to AI systems, risking loss of control.
OpenAI's CEO declares "code red" to enhance ChatGPT amid fierce competition from Google's Gemini 3. The San Francisco-based startup faces a pivotal moment in improving its chatbot technology.
Hundreds of TikTok accounts generate AI content, amassing 4.5bn views in a month, including controversial material. Researchers found 354 accounts posting 43,000 AI-generated videos, covering anti-immigrant and sexualized content.
Massive heat from constant server operation drives increasing power demand in Australian datacentres. Energy use may surpass electric vehicles by 2030, with a projected tripling of power demand in the next five years.
Silicon Valley faces pushback as consumers rethink upgrades. AI in government shows promise and pitfalls. ChatGPT under fire for dangerous advice.
AI has become ubiquitous, transforming industries and economies. Beware the risks of creating superintelligent AI, as warned by Yudkowsky and Soares.
AI and robotics will transform the world, impacting economy, warfare, and more. Super-intelligent AI could replace humans in controlling the planet.
MIT researchers develop safe control strategies for soft robots to interact with humans and delicate objects. New framework blends nonlinear control theory with physical modeling for “contact-aware safety,” ensuring robots achieve tasks without causing harm.
Silicon Valley companies invest trillions in AI race for global dominance. Commuters on a mission to achieve artificial general intelligence, changing humanity's future.
Albanese government's AI roadmap prioritizes economic benefits, plans to unlock vast datasets for training AI models. Focus on reskilling workers, boosting datacentre investment, and sharing productivity benefits across the economy.
NVIDIA expands open-source AI models for researchers at NeurIPS. Introducing Alpamayo-R1, the world's first open reasoning VLA model for autonomous driving, enhancing AV safety and advancing AI reasoning capabilities.
GM's Kurt Kelty details next-gen battery tech advancements, focusing on affordability, performance, and supply chain localization. LMR batteries set to revolutionize EV market with cost-effective, high-range solution by 2028.