Parameters (the numbers that measure an AI model's size)
Japanese パラメータ
Parameters are the numbers an AI adjusts while it learns, and their count is used to describe how big a model is. The more it has, the bigger the computer you need to run it.
Parameters are the numbers an AI adjusts bit by bit while it learns from a huge amount of text. Think of a recipe with countless dials for how much salt or soy sauce to add: the more dials, the finer the taste. In AI news, the B in labels like 70B or 405B stands for billion.
Watch out More parameters does not always mean a better AI. IBM explains that too many can make a model fit its training data too closely. The total count can also differ from the number used at one time. If you want to run a model on your own computer, check its parameter count first.
Source: IBM "What are Model Parameters?"; Google for Developers "Machine Learning Glossary"; Google for Developers "LLMs: What's a large language model?"