have not read the book but either that’s AI-written or they confused me with Stuart Russell and made some leaps that are a bit off. Some of the first sentence is correct about me; some note. The red lines stuff is really Stuart’s. The fire alarms stuff is sloppy and not
SAFETY
-
AI Proactivity and Personality: Senior Care Robot Case Study
By
–
I’ve been thinking about AI being proactive—tricky to get this right….and how it could fine-tune its personality….an interesting case study: To Stay in Her Home, She Let In an A.I. Robot https://
nytimes.com/2026/02/12/us/
elliq-ai-robot-senior-companion.html?smid=nytcore-ios-share
… via @NYTimes -
RAG as a Control Layer: Beyond Simple Search and GPT
By
–
RAG isn’t “search + GPT”. It’s a control layer:
• limits hallucinations
• enforces evidence
• defines what the model is allowed to know LLMs generate text.
RAG defines truth. -
Human-AI Interaction Framework for Mathematics Assistance
By
–
Yes, we provided 3 things for AI-assisted math:
* Human-AI interaction (HAI) card (photo), inspired by model cards
* Full transcripts https://
github.com/google-deepmin
d/superhuman/tree/main/aletheia
…
* A label for novelty-autonomy, inspired by SAE Levels of autonomy, see #Aletheia paper https://
arxiv.org/abs/2602.10177 -
Elon Musk’s Galileo test: AI must see truth despite false data
By
–
AI must pass, in general, the “Galileo” test: even if almost all the training data repeats falsehoods, it must nonetheless see the truth
-
Agentic AI: Autonomy, Guardrails, and Multi-Agent Architecture
By
–
Chatbot ≠ Agentic.
— Dr. Debashis Dutta (@debashis_dutta) 14 février 2026
RPA ≠ Agentic.
RAG ≠ Agentic.
Agentic AI = orchestrator + memory + planning + tools + feedback, plus multi-agent specialists (retrieval/coding/citations).
The board question: where do we allow autonomy—and with what guardrails? pic.twitter.com/AEnajJZeEPChatbot ≠ Agentic.
RPA ≠ Agentic.
RAG ≠ Agentic. Agentic AI = orchestrator + memory + planning + tools + feedback, plus multi-agent specialists (retrieval/coding/citations). The board question: where do we allow autonomy—and with what guardrails? -
Warning: Dangerous new technology requires isolation and monitoring
By
–
It is the most dangerous (er powerful) new technology ever invented. Put it on its own machine. Be very careful what you give it access to. Watch it like a hawk. Read my security list. Have your Claw read the security list and improve itself.
-
Extreme Risk Warnings and Calls for Caution on AI Development
By
–
And this is going past the bounds of common fucking sense, which says that if you fucking accept that something is 96% likely to kill everyone on the planet, back the fuck off, work on human intelligence augmentation, and sign up for cryonics if you're so fucking scared.
-
ASI Risk Calculus: Accelerationists’ Paradoxical Life Expectancy Arguments
By
–
It is subtweeting a recent claim, favorably reposted by famous accelerationists like Cremieux, purporting to calculate (and speaking favorably of the calculation) that your personal life expectancy is higher so long as ASI has a <97% chance of killing everyone.
-
Rejecting Fatalism in ASI Safety Discourse
By
–
Lads, the point of this post is not, "I analogized ASI to a falling asteroid, therefore I proved it's dangerous." We have a book and online resources about that. This post is about the selfish evil stupidity of "But everyone dies anyways lol."