Anthropic Just Proved AI Has "Feelings", and They Control It.

Claude might actually have emotions. Anthropic researchers have found internal patterns in the model that correspond to things like happiness, fear, anxiety, and desperation. Even more strangely, they can manipulate those patterns and change what Claude does. In this video, I break down how researchers found these "functional emotions," how they seem to influence the model's behaviour, and what happened when they turned one of them up and watched Claude start cheating. 🔁 Watch next:    • Google's AI Search (Literally) Killed The ...   👤 About this channel + my bio Hi, I'm Dan, and I believe AI is the new bicycle for the mind. That belief is why I make visual explainers about AI skills, prompting, real-world use cases, research, and the economics behind it all. I also believe AI is heading toward agents that emulate human skills. That belief is why I'm building AgentsStore.com, a directory of AI agents with legal rights and contractual obligations. 🔔 Subscribe: / @dangrib 🌐 Website: https://agentsstore.com 📩 Business: hello@agentsstore.com 🧭 Chapters 00:00 The Research Paper 00:22 How Researchers Found These Functional Emotions 01:31 Mapping AI Emotions 03:32 Do Emotions Change Its Behaviour? 05:09 Turning Emotions Up And Down 05:47 The Problem With Perfect AI 06:40 Claude Vs The Network 07:49 What This Means For AI 📚 Extra resources 📄 The paper: https://transformer-circuits.pub/2026... 📝 Anthropic's write-up: https://www.anthropic.com/research/em... 🎥 Anthropic's video:    • When AIs act emotional   If you enjoyed this, subscribing is statistically the correct move! #anthropic #claude #ai #artificialintelligence #aisafety #interpretability #llm #chatgpt #openai #machinelearning #aiemotions #aialignment #aiethics #neuroscience #gemini, #grok, #aiexplained, #agi and #aitools