AI

Computers used to be machines that only did exactly what they were told. Artificial intelligence (AI) adds human-like judgment on top of that — recognizing a cat in a photo, understanding speech, guessing what word comes next in a sentence.

Here's the important part: the AI you read about in the news today isn't a machine that "thinks like a human." It's closer to a program that learns patterns from enormous amounts of data and then produces very plausible-looking output based on those patterns. That's why it can be shockingly good at some things and absurdly bad at others.

In the news, "AI" gets used broadly. Chat services like ChatGPT, the large language models powering them, image-generation models that create pictures — all of it gets lumped together under the single word "AI." The other entries in this glossary unpack these different branches one by one.

"OpenAI unveiled a new AI model" — here, AI shows up both in the company's name (OpenAI) and in the product (AI model). You just have to tell them apart from context, and most of the time it's referring to "generative AI."