Loading slide
Attention is built around one small question, asked over and over: when I am trying to understand this word, which other words in the sentence matter most for it?
Watch the question do real work. Read these two sentences, identical but for the last word:
The trophy didn't fit in the suitcase because it was too big.
The trophy didn't fit in the suitcase because it was too small.
In the first, "it" means the trophy. In the second, the very same "it" means the suitcase. Nothing about the word "it" changed. What changed is the word at the end, and you reread "it" in its light without any effort. To get "it" right, your mind leaned almost entirely on two words, "trophy" and "big," and barely on the rest.
Attention is that leaning, made mechanical. For the word "it," some words in the sentence are highly relevant ("trophy," "big") and most are nearly irrelevant ("the," "in," "because"). Understanding "it" means knowing where to lean.
So here is the job we are handing the machine. For each word it is processing, look across the whole sentence and score every other word by how much it matters to this word, right now. Not a fixed rule carved in once, the same scores forever. A fresh judgement made from the actual sentence in front of it, so that "it" can lean on "trophy" in one sentence and "suitcase" in the next. The next slide shows what the machine does with those scores once it has them, and it turns out to be something you already understand.