Training an AI detector of consciousness

31:37raw brain waves, okay? >> Just can I identify consciousness if you give me the you know, sort of brain activity, the electrical activity of a brain. So, he made something called a deep convolutional neural network. These are just our normal, you know, CNNs that we hear about, giant network of artificial neurons that is trained through back propagation, you know, supervised learning, which is just we've got about 680,000 10-second electrophysiological recordings. So, 680,000 um recordings of brains electrical activity. These recordings come from humans, monkeys, rats, and bats under

32:17anesthesia, in a coma, and wakeful. So, a swath all across mammals, so that means it's not going to really hone in on some human feature. >> Right. >> It's going to it's going to really start thinking about what is the nature of consciousness itself. >> Mhm. Right? >> Mhm. And um he made three discriminators. So, he made a cortical neural network that was going to get trained on the cortical data. There's a thalamic, and then there's a pallidal. That's the three sort of big um brain regions that we were talking about in that meso circuit model, right? The The cortex is your grid with all the um homes and the podcast studios. The thalamus is the relay station, and the

32:59pallidal station is kind of like governing what's going on between these two, >> Yes. >> right? Yes. Okay. Once he trains it, he validates it. It's pretty good. The AI score for consciousness, which is plotted on the Y axis, >> Mhm. is correlated with the ground truth score of consciousness that it was given to each sample in the test sample, okay? Which means that like the the AI, this deep convolutional neural network, is given a sample that it's never seen before, and it's told to guess how conscious it is based on a number between zero and one. That number is correlated with the ground truth. Okay, so that means it's working. Yes. Okay?

33:40Yes. The other big one that I quite liked is from figure 1J, if you see over there. Mhm. What that shows is that the AI was able to classify fully paralyzed ALS patients Mhm. as conscious. Remember in ALS, we did the story where it's a motor neuron disease, right? All of your motor neurons go away. So, you can still be conscious, but you won't be able to to to fully express that. Mhm. Here, they looked at EEGs recordings from ALS patients, and there's no significant difference between them and healthy cortical activity. >> is actually a fascinating follow-up to our ALS story, because we did a great deep dive on understanding what is what

34:20it actually is mechanistically. >> Yes. Um and it dovetails with what we said earlier, which is just because your motor functions go away. ALS is a perfect example of what we were just talking about. Doesn't necessarily mean that you're there's no one home. >> Yeah. Yeah. That's very okay. >> Exactly. And this was able to find that, right? Without motor output, you can now classify. >> That's That's a >> So, already big win. Figure one. Yes. That is huge. >> Right? That's good. Okay. Now, next, we've we've made a detector. We've made a detector of consciousness. Now that we have a consciousness detector, let's make a brain simulator Okay. that can make my own brain signals. Okay. >> Okay? And this is where we get into the

35:01real why there's AI in the title. Okay. Okay? Because just making a deep convolutional neural network that will classify, that's just a autonomous classifier. >> Right. Right. Is that really Yeah. >> What are you doing, right? You're not going to get into Nature Neuroscience like that. And Daniel knows that. >> And this is for the social clips. If we end up clipping the first half without the second half, >> Yeah. please watch the pod so you get the full story. >> not like that's not really I can't believe that's Nature Neuroscience. Yeah, sweet. No, we can't fit a hour-long thing onto Instagram. >> Yeah. Yeah. Okay. So, now what he's going to do is make a brain simulator >> Mhm. He's going to combine that discriminator that he got, the the AI the

35:42consciousness detector, with something called a generative adversarial network. These are GANs. Traditionally with GANs, I mean, they're kind of used to make like generative AI like photographs and things like that. Yeah. Here's how it works. So, this is very different from like a diffusion model, which is usually kind of used in the in the zeitgeist. GANs kind of didn't have popular They had popularity for a while, and then and then diffusion kind of went up and became the generative sort of paradigm.