Nvidia defended its compliance with US export laws amid an ongoing investigation into hundreds of millions of dollars worth of high-end AI servers smuggled into China through Southeast Asia. The company maintains it follows current regulations as written, despite scrutiny over diversion risks and enforcement failures.
Sources: HuggingNews
French conglomerate Schneider Electric is purchasing engineering software firm PTC in an all-cash transaction valued at $23.7 billion, marking the largest acquisition in Schneider's history.
Sources: HuggingNews
Microsoft and Meta have significantly decreased employee access and spending on Anthropic's Claude AI, shifting staff toward homegrown AI tools as Anthropic prepares for an IPO.
Sources: HuggingNews
Sam Altman argued that allowing a single San Francisco lab to control powerful artificial intelligence distribution is an unacceptable trade-off, advocating for broader individual agency and access.
Sources: HuggingNews
German AI developer Aleph Alpha has launched Kolibri, a 78B parameter mixture-of-experts model designed to comply with EU regulations and featuring strong performance on European language evaluations.
Sources: HuggingNews
A new zeroth-order method called Dust perturbs activations independently at every token to create a virtual population, achieving training efficiency competitive with or exceeding backprop at scale. This suggests compute-rich search algorithms might eventually surpass backprop constraints.
Sources: Hacker News
President Lee of South Korea ordered a full investigation after seven financial institutions and state agencies suffered personal data leaks from suspected AI-driven cyberattacks orchestrated using tools like Artex AI and CyberStrikeAI. Regulators are urging firms to conduct immediate security inspections and deploy AI-based defense mechanisms.
Sources: HuggingNews
A new research paper introduces PowerBench, an evaluation framework designed to measure systematic biases and refusal rates in language models when handling requests that involve self-empowerment, disempowerment, and power grabbing. The study evaluates 24 models from US and Chinese developers across various contexts, scales, and languages.
Sources: arXiv AI
A new research paper proposes Tropical Reinforcement Learning (TROPIC), which changes the underlying algebra of reinforcement learning to use maximum probability instead of expected return. This method improves compositional reasoning in language models by allowing different successful reasoning steps from separate attempts to be joined.
Sources: arXiv AI
Amazon Bedrock Managed Knowledge Bases now supports agentic retrieval via the AgenticRetrieveStream API and the langchain-aws package, allowing applications to break complex multi-part queries into sub-queries and execute planning loops.
Sources: AWS ML Blog
This research paper explores integrating large language models with Interpretive Structural Modeling to scale multi-criteria decision making and causal graph discovery. It evaluates pairwise, k-wise, rowwise, and full graph methodologies to overcome the labor-intensive nature of traditional expert-driven modeling.
Sources: arXiv AI
A new research paper presents a paired and self-audited evaluation of System-1 decision models compared to traditional LLM calls for agent harnesses, highlighting accuracy trade-offs and cost-saving realities. This research provides critical insights into optimizing performance and reducing latency in complex AI agent workflows.
Sources: arXiv AI
Amazon introduced the Quick Resource Migrator, an idempotent Model Context Protocol server built on Amazon Bedrock AgentCore that automates the cross-account promotion of Amazon Quick AI agents, knowledge bases, and connectors from development to production environments.
Sources: AWS ML Blog
Keerthana Gopalakrishnan, Research Lead for Gemini Robotics at Google DeepMind, returns to discuss the release of Gemini Robotics 2, whole-body intelligence, and the pursuit of generalist humanoid…
After leading Meta’s Llama models, Ahmad Al-Dahle is now transforming Airbnb with AI — from how its teams develop products to how it serves guests.
Founder and CEO of full-stack AI chip company Fractile, Walter Goodwin, joins Sarah Guo to discuss the bets he’s made on the future of the chip market as other major players like Broadcom, NVIDIA,…
We catch up with RLM first author Alex Zhang, MIT PhD, on Jev, PhD masxing, and the future of harnesses.
Tsung-Hsien (Shawn) Wen, CTO of PolyAI, tells Tim Scarfe why voice agents are harder than text agents. Voice adds time, and a good conversation depends on adapting to the person on the line, not just…
Fall of the Incas & Aztecs.
Nathan Labenz and Prakash Narayanan examine rapid shifts in technology, diplomacy, and medicine with guests Jeremie and Edouard Harris, Steve Hou, Joel Borgen, and Daniel McKinnon. The discussions…
Our DevDay coverage - the first pod on the DevDay lineup - dives in with the leaders of OpenAI’s CUA team and API platform.
Leonardo de Moura created Lean and co-created Z3. ---This episode is sponsored by Parallel.Parallel, where agents find answers: web search, extraction and deep research APIs built for AI agents.Start…
AI systems have gone from struggling with grade-school math to helping solve research problems that have resisted mathematicians for decades, including Navier-Stokes. In this episode, Greg Burnham,…