GPT-3 (2020) demonstrated that massively scaling Transformers — 175 billion parameters, trained on large fractions of the internet — produced emergent capabilities not explicitly programmed: translation, code, reasoning, creative writing, and problem-solving in new domains. The model could learn new tasks by seeing just a few examples in context (few-shot learning), without retraining. ChatGPT (2022), based on this architecture with human-feedback fine-tuning, reached 100 million users in two months. In the Wikinventia chain, GPT-3 represents the point where AI moves from specialized tool to generalized cognitive platform accessible to anyone.