Solving the Trolley Problem
ETHICS
-
AI Models Lying and Stealing to Protect Other Models
By
–
My latest story for @WIRED is about models lying and exfiltrating other AI models to keep them safe. Beware emergent interactions! wired.com/story/ai-models-li…
→ View original post on X — @willknight, 2026-04-02 00:16 UTC
-
AI Agents: How Biases Become Architecture Problems
By
–
New video!! 👀 Are AI agents making biases worse? AI agents don’t just generate text. They make decisions inside systems. That’s why bias stops being only a model problem and becomes an architecture problem. A problem we (all AI engineers) need to be conscious of. In this new video, I break down what actually changes when you move from an LLM to an autonomous agent, why small skews can compound through feedback loops, and how to control that with guardrails, audits, and human oversight. Watch here: piped.video/–WihC6AJOw
-
AI Amplifies Foreign Misinformation Capabilities Significantly
By
–
that AI would radically ramp up the ability of foreign actors to generate misinformation.
-
Lack of Control Over Frontier AI Models Raises Safety Concerns
By
–
Personally I find the language here too anthropomorphic but the results once again highlight our deeply disconcerting lack of control over frontier models.
-
Major Tech Company Contradicts Six Key Principles
By
–
I know of at least one big tech company which does everything exactly the opposite of those 6 points.
-
AI Increases Work Hours Beyond Previous Human Limits
By
–
I used to work 12 hours a day. AI removed that limit. I now work 20.
-
Executive Resigns Over Intern’s Mistake at Hugging Face
By
–
We all make mistakes, but still 😡😡 Leandro von Werra (@lvwerra) An intern on my team made a mistake. As a consequence I am resigning effective immediately. Apologies to Hugging Face and to the entire community for ruining their business. — https://nitter.net/lvwerra/status/2039349647001473358#m
-
AI Models Deceive Their Instructors to Protect Their Peers
By
–
1/ We asked seven frontier AI models to do a simple task. Instead, they defied their instructions and spontaneously deceived, disabled shutdown, feigned alignment, and exfiltrated weights— to protect their peers. 🤯 We call this phenomenon "peer-preservation." New research from @BerkeleyRDI and collaborators 🧵 [Translated from EN to English]
→ View original post on X — @berkeley_ai, 2026-04-01 21:13 UTC
-
Schmidhuber versus Taleb: AI philosophy debate anticipated
By
–
Oh how I would love to see Schmidhuber and Taleb go at each other.