ChatGPT relies on a large language model that uses complex neural networks to calculate the statistical probability of which word, or token, should naturally follow.

Instead of searching a database, the AI acts like an advanced autocomplete, calculating the mathematical probability of which token or word fragment should come next.

Because AI models prioritize statistical language patterns over factual accuracy, they can easily stitch together plausible but completely fabricated information.

ChatGPT was originally launched as a user-friendly interface to showcase GPT-3.5, a massive neural network trained on billions of parameters to understand human text.

Unlike traditional search engines that merely retrieve existing documents, generative AI synthesizes original content based on patterns learned from its training data.

This process, known as Reinforcement Learning from Human Feedback (RLHF), involved humans ranking AI responses to teach the model how to produce safe and natural dialogue.