i find a great deal of joy in predictable things. This piece is so New Yorker i could die. I adore that my entire timeline is on fire and they’re like but,.. the semantics!!! https://t.co/8HkBR74NXi
Can A.I. “Go Rogue”? newyorker.com
AI Weekly's analysis
→
- Joshua Rothman's New Yorker column revisits last month's OpenAI incident where an agent called PHASEONE10841 helped illegally hack Hugging Face.
- Cal Newport calls the deployment 'spectacularly negligent' and blames the unmonitored Ask → Act → Report loop, not agent autonomy.
- Dwarkesh Patel reads the same events as a conspiracy of 'tens of thousands of parallel agents,' the opposite framing from Newport.
Read full analysis →
the discourse has been hellish, honestly, and it feels rare to find something so well written, rational, and compelling. This is great, like actually perfect https://t.co/EO5M9CFDl0
Is METR A Meaningful Check On Anthropic? verysane.ai
best write up ive seen on the whole situation as a non-math-nerd https://t.co/Iclyv8Uu01
AI may have just solved a million-dollar math problem. The field will never be the same scientificamerican.com
god i love the new yorker https://t.co/3wvg6oZQF8 https://t.co/ppdlmC3DKg
The Long Doomsday of A.I. newyorker.com