Brian O'Grady

It's kind of unclear.

Now, a word standalone bat, that doesn't tell you much information, but how people are normally interacting with search bars these days is they're kind of

inputting these more like freeform queries that somehow give context, like what the person means when they say bat.

But the issue with keyword search is that it is not able to understand like the actual context.

It's just sort of doing like a

short of exact match for keywords, right?

And then returning results based on sort of statistics and not really trying to understand, quote unquote, the meaning of the search.

So the real advancement that happened in the late 10s and earlier in this decade

was this ability to take text and encode its meaning structurally into something called an embedding, which is also known as like a vector.

So this is where vector search comes in.

So rather than saying, hey, you know, just look for results that contain, like I type in to, you know, another good example might be, let's say I'm using like DoorDash.

And I want to type in the search, like tasty fast food burgers within five miles.

Right.

Like that might be something I might say to DoorDash.

Rather than trying to go ahead and look at all these words that are in there and try to match them to results.

Like if I look for tasty.

Right.

And then try to match it to a result.

You know, that's kind of odd.

Like what kind of results actually have.