What's the difference between narrow AI and artificial general intelligence?

Narrow AI can do one specific task really well — like recommending Netflix shows or spotting a cat in your photos — while artificial general intelligence would be able to learn and handle any intellectual task a human can, which doesn't exist yet and may not for decades. Think of narrow AI as a calculator. It's brilliant at math but useless for writing poetry. That's where we are right now with all AI tools, including ChatGPT, Claude, and Gemini. They're pattern matchers trained on enormous datasets, and they can generate surprisingly coherent text, but they don't understand what they're saying. A 2024 survey of AI researchers published in the Journal of Artificial Intelligence Research found the median prediction for when AGI might arrive is around 2060, though opinions ranged from 'never' to 'within five years.' The confusion happens because tools like ChatGPT feel general. You can ask it to write a recipe, debug code, or explain quantum physics, and it does all of them passably. But it's not switching between skills the way you would — it's running the same prediction engine on different prompts. AGI would need to form its own goals, learn from experience in real time, and transfer knowledge between completely unrelated domains without being explicitly trained. We're not close to that. **Related**: Will AI ever become self-aware? | How does machine learning actually work under the hood?