Brian O'Grady

The sequence of words brick and mortar may be able to be closer in my mind to burger than I would have thought because I'm like, oh, brick and mortar could mean a brick and mortar store, a brick and mortar restaurant.

There's these slight nuances where, depending on the context, the combination of words, suddenly...

And it also enables you to search in ways that you were unable to previously.

An example being, coming back to this idea of embeddings representing sort of meaning into a generic quadrant or space, there have been developments where you can actually take text

and you can map it to images.

So you can do things like text to image search.

So you can say, hey, show me, you know, red Nike shoes, right?

And rather than like looking at, you know, a product description that says red Nike shoes, it'll look at like images that are encoded in the same vector space.

And it's able to say, oh, this image has red Nike shoes in it or something close to it, right?

It might just be like a red shoes, red sneakers, but it'll show them to you as a search result.

I mean, and this is where it's like it's search over meaning.

Right.

And that's kind of where I was trying to get with my philosophical intro, which is like it's this sort of rant.

It's this sort of odd, you know, space of meaning where humans are able to kind of map things together.

We're trying to represent that mathematically, essentially.

That's what researchers are trying to do.

Yeah, so in general, right, like what will often happen, what will often happen is that companies will have some sort of existing search solution, they'll be on, you know, elastic or open search or solar, right.

And the idea is that those engines were built for text search specifically.

So

For text search, they're very good.