David Oakley, a physicist, discusses how measuring brain function can improve performance, exploring the difference between correlation and causation in neuroscience and how AI analyzes brain data related to pain and addiction.
Bulletproof radio, a state of high performance.
Let's just say that there's clearly a correlation between what the brain does and what the body does. And there are some people listening who probably are fuzzy on their science around correlation versus causation. And just to define terms for you, causation means if you do A, it always causes B. But a lot of the science, including stuff you see from big pharma, a lot of this other stuff is saying, well, we don't really know that A causes B. We just know they happen together. And as a systems guy or a systems biologist or a network thinker, whatever you want to call it, sometimes it's like you could do A or B, right? And then the other one goes. So they move together, but you can't say one causes the other. And I want to go deep with you because you're a physicist, because you guys have unusual brains. And you in particular, because you went from one field into another field, which means you don't think the way someone who went to grad school studying neuroscience would think. So, so you're a disruptive thinker by definition because you had your midlife transition. So talk to me about brains. How do we know what causes what in the brain?
This is something I had to learn because in physics we're trained to look at causation. Yeah, that's what we do. And I've seen so much brain data now, the brain is so complicated that at some point it's easier just to give up on the causation and look at the correlation, make sure the correlations are decent. This voltage signal is so ideal for artificial intelligence to analyze it, but in that you have to give up. You don't know what the artificial intelligence is seeing, but it can see patterns that we can't see for sure. And so if one does an mri, those are always causation based. Where's the blood going? But in our data, we've kind of given up on that in a way. I mean, it's still interesting to go backwards and see what you're saying. But for example, we have a data set, NIH funded study to look at opioid rehabilitation and pain. And so I gave it to one of the students to analyze, and the student came up with an artificial intelligence program that could detect the difference. 90% accurate. 95. So with four seconds of data, I think is what he used, something like that. He could detect people who were addicts versus people that were truly in pain and control also. So a threefold separation. We don't know what the artificial intelligence was looking at, but it found the pattern. So in some sense, the brain is so complicated, maybe that's going to be good enough. Right now we can go back and do the studies, MRI studies they have done. One of our collaborators has done a lot of MRI studies on pain. So I can see certain different pain centers in group settings, but on individuals, it's a different animal. So I can say this is your pain center, but it's not really your pain center. It's the group average pain center. So you're going to be different.
