But if a patient wants to ask AI a health question, what information or context should they provide to get the most useful answer?

Yeah, that's a great question.

I think it's almost possibly the most important question you could ask, right? So I think one of the, the-- To start, I can say that the more, generally the more information and context you can give AI, the better an answer you're gonna get.

In some ways, it's not that different from going to see a doctor.

If you come in with test results, past records, if you have a list of really good, good questions or in, or a, a few sheets of information that summarize your health history, you're gonna get a lot more out of that visit than you otherwise would.

With AI, I think it's e- especially important to provide context, so helping it understand who are you, how old are you, what are your other health issues? If you have a cancer diagnosis, when were you diagnosed? What-- If you have test results that you feel comfortable sharing, whether that's your MRI reports, CT reports, pathology reports, all that information helps the AI reason more, more carefully about the question that you may ask it.

And why it's important for AI is that AI will almost always give you an answer, no matter what information you give it.

[laughs]

Yeah, please go ahead.

Yeah, so the first thing is memory.

Almost every AI model by most standards They're all brilliant.

They know more about the world than any single human being possibly could.

They were trained off of all of the available knowledge on the internet.

It's almost unimaginable how much knowledge has gone into these AI models.

So they have a lot of knowledge, but their memory is still really limited.

Yeah, so being told what you want to

Mm-hmm.

And part of, part of how you can use this technology safely is to entertain some of that complexity and say, "Hey, I have a headache.

Well, it could be my existing migraines.

Could it be something like sleep deprivation? Could it be something else?" And then the, the more questions that you answer, the better this gets.

And I mentioned that My Doctor Friend has these different models, uh, sort of working together.

One of them is always trying to figure out how confident can we be of the response.

So this is a-another thing that we add to our platform, is the model will always share, "Hey, how sure am I of this answer?" By default, we've designed it so that it is never a hundred percent confident, 'cause unfortunately, that's just rarely the way, um, health works.

Yeah.