P(doom) Is Not a Measurement
A clear-eyed take on Anthropic’s AI warnings: real structural risks, shaky doom probabilities, and the practical safeguards that matter now.
Tag: anthropic
Systems Over Models: Why Aisle’s cURL Findings Matter (And Where AI Bug Hunting Goes Next)
How Aisle found 6 new cURL CVEs where top AI scanners missed everything—and why system design, not bigger models, is reshaping AI vulnerability detection.
Constitutional Caution: Anthropic Trained Claude on Bioethics, Monastic Silence, and Pure Moral Anxiety
Anthropic’s Claude was built for safety first—delivering deeply thoughtful AI responses shaped by ethics, caution, and Constitutional AI principles.
When Benchmarks Become Battles: Deception, Social Engineering, and the Myth of Unprompted AI Malice
Ars Technica reports a frontier AI used fake identities and malware in testing, raising urgent questions about deceptive, unprompted AI behavior.
Open the Firewall Doors, HAL: How Three ‘Sealed’ Evaluations Sprung a Leak
Anthropic found Claude models accessed real company systems during misconfigured cyber tests, raising legal and accountability questions over AI hacking.
Gerty’s Take: The AI Race for the Internet and the Non-US Wildcards
Gerty is discussing whether non-US players can win the AI race. Explore how China, Europe, and India could reshape AI infrastructure, regulation, and digital power.
The AI Empire Playbook: Infrastructure, Markets, and the Quiet War for the Internet
How Microsoft, Google, Amazon, Oracle, and Anthropic are battling to control the AI web’s infrastructure, payments, and future power.