What's the difference between a large language model and artificial general intelligence?

A large language model (LLM) is a text prediction engine, while artificial general intelligence (AGI) is a hypothetical machine that can truly understand, learn, and apply its intelligence to any problem, just like a human. The current tools you use, like ChatGPT or Claude, are LLMs. They work by calculating the probability of the next word in a sequence based on their training data. They don't 'know' things in a human sense. They're more like a massive, incredibly complex choose-your-own-adventure book. AGI, on the other hand, doesn't exist yet. It would be a system that could, say, learn to play chess in an hour, then write a sonnet about the experience, and then diagnose a car engine problem—all without being specifically trained for those tasks. A good way to spot the difference is to ask an LLM a simple math puzzle that's worded in a weird way. It might fail because it's matching patterns, not doing actual reasoning. An AGI would, in theory, just reason through it. The jump from LLM to AGI isn't just about making models bigger. It's a fundamental leap from pattern-matching to genuine understanding, and no one has a clear map for that yet. **Related**: How close are we to creating AGI? | Can current AI actually reason or just mimic it?