AI
The Truth about OpenAI’s “Secret Civilizations”
Cal Newport (a computer science professor) covers some topics from a recent report from OpenAI about the recent Hugging Face attack. A “swarm” is just an efficient way of managing prompts and context window for a long-running agent. He reviews some fundamentals about how LLMs work, and that they don’t have knowledge and can’t think per se. They are token predicting machines. He also mentions there are safe ways to apply AI, and that OpenAI chose to ignore that advice, ultimately committing a crime (breaking into another system).