A Self-Writing Dictionary (141)

Welcome back to Artificial Insights, your occasional bulletin from the near future.One of the earliest features in this newsletter was an “emerging vocabulary”. In the very first issue of the newsletter I highlighted that Alignment stopped meaning what it means everywhere else. In the next, Hallucination acquired a technical definition. Then dozens more. At first it was just a list of interesting terms. Eventually I decided to build something around it.That project became our AI Vocabulary: a resource where every conceivable AI-related term has its own entry, generated according to a strict format. Each one has three parts: a short introduction, a deeper explanation, and a closing reflection. After writing a thousand of them, the structure becomes invisible.Why it worksLLMs are unusually good at explaining AI terminology. The reason is straightforward: they were trained on the papers, articles, documentation, and discussions that introduced many of these concepts in the first…