Brian O'Grady

So the idea is

Text is simply a representation of shared language and meaning that humans are able to communicate to each other.

So text is a simple representation of this meaning that we are somehow communicating to each other with language.

If we think about how could a computer somehow learn that same meaning and be able to communicate it, what you can do is you can simply say, well, let's simply, instead of representing language and meaning as text,

let's represent it in bits and bytes, AKA numbers, like floating points and associations of floating points.

So like vectors.

So really an embedding and vector search comes down to this ability to take language and meaning.

And rather than encode a language and meaning in like words and text, you simply encode it as numbers.

And it's just another form of representation of meaning.

And it's one that you can now perform mathematical operations on that are very convenient.

Yeah, and, like, we can even go further to say that, you know, like, if you think about this idea of just, you know, what things are close to each other, right?

We can go kind of further.

A good example, actually, that we may have missed on, unrelated to burger, but if you're just doing, like, let's just say you're just doing full text search.

And let's say, like, very close synonyms.

Let's say the word dry versus the word arid.

They mean like the exact same thing practically.

Right.

Like they're just slight variations on the same underlying meaning.

Right.

But if I do a search, you know, in Google and I'm like looking for, you know, arid, something arid.