Steve Brown

Gosh, when you ask the question differently, you get a different answer to you.

When you ask a question from a different point of view, like a gastroenterologist versus a cardiologist, you also get a different answer.

So how do I take this kind of diversity of answers and turn that into a strength?

So I said, well, I have like 36 different agents, all with different points of view.

And when they look at your medical record from different points of view, they might get different answers.

And you might have a different conversation with them.

But when you take those 36 different points of view and you synthesize them and look where answers converge, that also is a really important signal.

If you see somebody that's just an outlier and nobody else, you know, and then when the agents talk to each other, they don't agree, they say, hey, you know, that's...

that's out there, that tells you something.

But when you take an answer and you feed it to the other agents and they say, hey, that's a good idea.

I didn't think of that the first time.

But now what you're doing actually is you're creating a new pathway through this massive collective consciousness of human knowledge.

So every time you do this, a different pathway.

So I want a lot of different pathways and I want to look for convergent

because that's really useful.

If all of human knowledge is compressed into this thing called an LLM, there's a lot of different points of view in that knowledge, and there's a lot of variety in what could be right and what could be wrong.

So the other thing I'm also not asking for like, what's the answer?

Like, hey, here's my data, what's the answer?

Even in the diagnostic side, it's called differential diagnosis.

It's based on, hey, based on all of this stuff, it could be this, it could be that, it could be something else.