What actually IS artificial intelligence?

The explanation I find most useful for non-technical people is the pattern recognition framing. AI systems learn to recognize patterns in training data and then apply those patterns to new inputs. The intelligence is not reasoning from first principles, it is sophisticated pattern matching at enormous scale. That framing helps people understand why AI is impressive at some tasks and genuinely unreliable at others, and why it can be confidently wrong.

Great question to start with, because the confusion is completely understandable. The term "AI" gets applied to everything from a spam filter to a self-driving car, which makes it feel meaningless. At its core, AI is software designed to perform tasks that would normally require human-like reasoning: understanding language, recognising images, making decisions, or generating creative output. It's not robots (that's robotics, a separate field), and it's not magic. It's pattern recognition at a s...

The analogy that landed best for my parents when I tried to explain it was autocomplete but much more sophisticated. They understood autocomplete on their phones. AI is that same idea but instead of predicting the next word in a text message it can predict the next word in an essay, a conversation, a piece of code, an image description. That scaling from word prediction to general-purpose text generation is the jump that took decades to make practical.

The part most plain English explanations miss is the probabilistic nature of the outputs. These systems don't look things up or reason to correct answers. They generate the most statistically likely next token given everything that came before. That's why they can be wrong with complete confidence, the wrong answer was statistically plausible based on the training. Once people understand that they stop being surprised by confident errors and start understanding when to trust the output and when ...