Brian O'Grady

and it looks for arid climates, but all of my search results, I have no search results that contain the word arid.

They all contain the word dry.

I'm going to get no hits because suddenly I have filtered unnecessarily.

I've done a keyword match for a word where none of the documents contain that.

The idea is that by encoding pieces of text as numerical values,

As humans, we know like, and English speakers, we know like dry and arid mean effectively the same thing.

But like, you know, a text search engine does not necessarily know that.

But by transforming them into vectors, we're now able to say, just like we're able to kind of like mentally map them to the same space in our brains, we're able to say, hey, like, yeah, dry, arid, they're kind of like the same thing, right?

Vectors are now ways to simply represent that close association mathematically via like a literal distance function.

Yeah, no, and it could get way more interesting just to take some examples further.

I just had one when you were speaking.

I was like, oh, what would be really different from a burger?

And I was like, oh, a brick.

A brick is very different from a burger.

When I think of burgers, I don't think of bricks.

It's like the opposite of bricks.

So if I think about them in some quadrant, they should be very far away from each other.

They should be nowhere near each other.

But then I'm like, oh, well, what about the sequence of words brick and mortar?

Brick and mortar could mean a little local restaurant or something.