A paper argues that recursive self-improvement (RSI) is being pursued by AI companies without any shared view of when it would be safe, and without even a clear framework for reasoning about where it leads. The author also demonstrates that a single high-level prompt can get an AI to train a small LLM from scratch, underscoring how fast the capability barrier is falling.
LLM
大语言模型
Large neural networks pretrained on massive text corpora with general generation and reasoning abilities.
Curated Intelligence
China's National Development and Reform Commission outlined plans to accelerate age-friendly technology, explicitly targeting brain-computer interfaces, smart assistive devices and service robots. A joint policy document calls for deploying embodied AI, large language models, BCIs and exoskeletons in elderly care, backed by national science programs, opening a clear policy-driven demand channel for AI and robotics vendors.
A sophomore computer science student's blog post arguing that AI is ruining her degree has hit the top of Hacker News within 24 hours. She describes syllabi mass-generated by Claude, assignments auto-graded by LLMs, professors offloading teaching materials to AI, and contradictory AI-use policies within the same course, exposing broader governance failures in universities confronting generative AI.
AI companies hunting for training data are unleashing data-scraping bot swarms on academic libraries, slowing or crashing sites meant for students, researchers and the public. One school reported being hit by more than 11,000 searches on Finnish music in a single day.
The mathematics community is distancing itself from OpenAI due to perceived lack of integrity and exaggerated marketing, favoring open-source tools like the Lean prover for their honest and impactful contributions. This shift highlights a broader industry trend towards transparency and verifiable results over hype.
SJTU Antai MBA introduced an AI strategy workshop integrating LLMs and optimization algorithms into real-world procurement scenarios. This initiative addresses the gap where only 7% of enterprises achieve scaled AI deployment, emphasizing the need for leaders who can bridge technical capabilities with business judgment.
Apple has confirmed that the original HomePod mini will not support the new Siri AI due to insufficient chip and memory resources for running local LLMs. Only the second-generation HomePod mini and newer Apple TV models meet the compute baseline, highlighting the rising hardware requirements for on-device intelligence.
Linus Torvalds announced and submitted the first Linux kernel patch series assisted by a Large Language Model (LLM). The patches optimize folio lookup by detecting hot folios and extending RCU protection to cover actual data access, eliminating refcount overhead in fast paths. This resulted in an order-of-magnitude speedup for read-heavy benchmarks, though code review has already identified a race condition requiring further fixes.
Amp CEO Quinn Slack states that their AI coding agent leverages OpenAI models to expand capabilities from code writing to product design and decision support. Addressing the cost pressure of users consuming up to 10 billion tokens weekly, Amp optimizes billing via features like 'Sign in with ChatGPT' to help developers focus on creative execution rather than micro-management.
Rust core contributor Aaron Hsu analyzes the impact of AI agents on programming language design. He argues that as agents increasingly translate high-level designs into traditional languages like Rust with minimal human intervention, the tradeoffs in language design must shift from human readability to machine verifiability and generation efficiency.