Loading activity…
@random_walker
Princeton CS prof and Director @PrincetonCITP. Coauthor of "AI Snake Oil" and "AI as Normal Technology". https://www.normaltech.ai/ Views mine.
AI
AI as Normal Technology proposes agents that reveal their working steps and let people steer them continuously.
AI
Clement Delangue says it omitted the lack of agent containment and monitoring and Hugging Face's open-model defense.
AI
A coauthor says the paper also offers recommendations for evaluators, AI companies, policymakers and others.
AI
Arvind Narayanan says the collection offers “grounded optimism” about adapting to—and challenging—flawed institutions.
AI
A user argues that safety, control, ethics and skill erosion should be decisive in AI deployment, not secondary concerns.
AI
In a recap of his Ezra Klein interview, the guest argues sandboxes and monitoring deserve more investment alongside alignment.
AI
In a new essay, the authors argue that sincere warnings about AI existential risk may still be mistaken and counterproductive.
AI
A user argues that managing AI agents could pull knowledge workers away from their craft and leave them responsible for mistakes, without the rewards of managing people.
AI
The arXiv description of “Life After Benchmark Saturation” argues that replacing benchmarks with harder versions once accuracy saturates privileges accuracy and misses opportunities for study. An author says “NeurIPS / Claude” rejected the paper.
AI
A post sharing a conversation with Graham says she discusses why her career advice no longer holds in an AI world, including how delegating to AI differs from delegating to a person.
AI
A commenter questions what the benchmark numbers prove, suggests the test setups probably differ, and welcomes more people testing things on their own.
AI
One post argues that AI’s implications for humanity should take priority over concerns about mathematicians’ careers.
AI
The AI as Normal Technology authors argue that loss-of-control incidents are primarily a security story, while warning that controlling AI agents isn't a solved problem.
AI
A user calls for more attention to the group’s work, highlighting a standard covering embedded audits and members including METR, Transluce, RAND, AVERI, Princeton and SecureBio.
AI
One user reports about 75 inquiries for fall 2027 PhD admissions and says they’re long past being able to open all their mail. Their concern: losing a way to reach people without an introduction.
AI
A critique of Dario’s “Pacing AI” letter argues that Anthropic’s model documentation points to coding productivity gains sustaining the pace of progress, not accelerating it.
AI
Retweet by Meta executive highlights release of updated model from Xiangning Chen.
AI
Stanford researcher replies that the term misleads by suggesting ordinary technology.
AI
Tech journalist shares analysis of why robots may lag human workers for years.
AI
Arvind Narayanan argues distinctions remain possible by tracking model evaluation calls versus research assistant uses and funding sources.
