Reading up on artificial-intelligence
18 deep · digging since dec 02, 25
- The Promise of Polymath LLMs - by Robin Hanson
Robin Hanson argues that large language models could drive progress by systematically finding and resolving contradictions between unrelated fields of knowledge.
- Why A.I. Safety Controls Are Not Very Effective
Jailbreaking AI systems to bypass safety controls remains trivially easy three years after ChatGPT launched, despite industry claims of progress.
- The Model Is No Longer the Bottleneck
A general-purpose AI model rivaled specialized chemistry software on NMR tasks, showing the bottleneck has shifted from model capability to the workflow scaffolding around it.
- When AI builds itself \ Anthropic
Anthropic presents internal data showing AI systems now write 80% of its code and accelerate research, trending toward autonomous recursive self-improvement.
- Korea's biggest manufacturers back Config, the TSMC of robot data
Config, a robotics data startup backed by Samsung and Hyundai, supplies proprietary training data for manufacturers building their own robot AI, likening itself to TSMC.
- China Seeks A.I. Independence, Weakening Trump’s Leverage
China has achieved a milestone in AI self-sufficiency, reducing its reliance on US technology and weakening American leverage ahead of bilateral talks.
- Random thoughts while gazing at the misty AI Frontier
AI revenue now accounts for 0.25-0.5% of US GDP, with OpenAI and Anthropic at ~$30B each, and could hit 1% by end of 2026 amid compute constraints and market shifts.
- Physical Intelligence, a hot robotics startup, says its new robot brain can figure out tasks it was never taught
Physical Intelligence's new π0.7 model can direct robots to perform unfamiliar tasks by combining learned skills, surprising its own researchers.
- The Trillion Dollar Race to Automate Our Entire Lives - WSJ
Anthropic, OpenAI, and Cursor are racing to build AI agents that automate work and personal tasks using natural language commands, moving beyond code generation.
- The Anatomy of an Agent Harness
An agent's usefulness comes from its 'harness'—the code, tools, and orchestration around the model—which must be engineered for tasks like durable storage, code execution, and context management.
- Can A.I. Detection Tools Really Spot Fake Images and Videos? - The New York Times
A New York Times investigation of over 1,000 tests found AI detection tools have notable strengths but also significant weaknesses in identifying fake images and videos.
- Opinion | Anthropic’s Chief on A.I.: ‘We Don’t Know if the Models Are Conscious’ - The New York Times
Anthropic CEO Dario Amodei discusses uncertain AI consciousness and shares both utopian and dystopian near-term predictions for artificial intelligence.
- What Is Claude? Anthropic Doesn’t Know, Either
Anthropic researchers use interpretability methods to study Claude's internal workings, finding its selfhood is shaped by both neurons and narratives like human minds.
- There’s a Hit Movie Set Deep Inside an AI Lab—and It Will Give You Goosebumps - WSJ
A documentary about Google DeepMind, directed by Greg Kohs, captures the emotional weight and historic significance of AI research, likened to the Manhattan Project.
- A.I. Deal Making Is Getting Faster - The New York Times
AI startup investors now decide on multi-million-dollar deals in as little as 15 minutes and compete with extreme gestures like weight-lifting and rock climbing outings.
- Opinion | How to Prepare Your Kids for the A.I. Revolution - The New York Times
The author, an AI practitioner, expresses concern about AI's impact on children and offers specific strategies to help parents prepare them for an AI-driven future.