 I'm Blue Knight, I'm here with Daniel Friedman. Today is November 8th, 2021, for a guest stream with Adam Saffron. Adam, please take it away. Hi, Blue, hi, Daniel. Hi, everyone. It's good to be talking with you today. And I guess we're here to talk about psychedelics and the 5-HG2A and related systems and active inference. And so I'll describe some recent work on this and looking forward to talking about it with you. So let me get my screen sharing, okay. Coming through, excellent. So I'm using my active inference workshop slides as a base and I'll go over those with a little bit slower than I did in the 10 minutes I had there and then we'll take it from there and talk it out. So this is part of an ongoing project to both try to understand the mechanisms of psychedelics, but also the impacted neurotransmitter or neuromodulator systems. And the worlds that would have an active inference and the potential significances for their use as machine learning parameters potentially. And so when it comes to psychedelics, much of the attention has been on the 5-HG2A system. And that's a lot of evidence in the case that it mediates the effects. Barely any work has been done in machine learning on this and there has been worked on an active inference and I'm gonna introduce some, a compatible but slightly different take on the potential significances of these systems. So within machine learning, a good amount of work and with an active inference has been done on the potential roles of dopamine as a parameter encoding the confidence one might have with respect to policy selection. Diane and others have suggested that serotonin can act with an impotency to dopamine based on a variety of lines of evidence from neurophysiology. And so some work has also been done in introducing a complementary uncertainty parameter in the form of serotonin. This proponency can be thought of almost yin-yang like with dopamine being more of the yang and serotonin being more of the yin. Dopamine in general, indicating in the neuroscience literature is incentive salience, confidence and if excessive, potentially impulsivity. Serotonin would be an oponency or complementarity would indicate the satisfactional desires, states of uncertainty and potentially a factor helping to allow for patient satisfaction and a factor helping to allow for patients rather than impulsivity with respect to policy selection. So very little work though has been done in machine learning on 5-HG2A and also in terms of active inference in terms of actually modeling the 5-HG2A system as an active inference parameter. Work is probably going to begin soon in this direction largely informed by the rebus paradigm about soon. But go back to the fundamentals of these systems. The work that's been done so far with the serotonin system can be thought of more as relating to the 1A system. This would be in contrast to the higher threshold, I believe it's a 5-to-1 ratio of sensitivity for the receptor of the 2A system. So 5-HG is serotonin and you have some, I think, 14, maybe 17 classes of receptors depending on how you count them. So are we going to talk about all of them? Well, maybe eventually. But it seems that we might be able to go a fair ways in terms of understanding their significance, focusing on a few classes and sort of working our way forward with the additional receptors maybe having a common significance to the other ones but just because they're located at different parts in the brain it's a common organism significance. But functionally, you need them to do different things that's located in different places into more detail soon. But the 1A system is shown to in general have more inhibitory effects on the signaling of pyramidal neurons while the 5-HG2A system tends to be more excitatory. Carhart-Harrison and Nutt have suggested that these both in a face of challenge would be responses to challenge but the 1A might be something more like an active or a passive coping response being more still, patient, stop and think as opposed to the 2A system which might be more of an active coping response in the face of challenge. That's something more imaginative and potentially creative. So, yeah, let's keep going. So, going back to dopamine versus serotonin though, serotonin in general, not focusing necessarily on 1A versus 2A there seems to be some evidence suggesting that dopamine will push you or actually a good amount that dopamine tends to push the organism towards more externally focused attention in terms of helping to increase the gain on more extraceptive and salience networks for externally focused attention. 5-HG in contrast seems to be up-regulating the default mode and more internally focused attention. So, that's another interesting difference. But to come to the 2A, so these receptors as I mentioned before they tend to have high thresholds for activation by serotonin. 1A will be modulating mood as their serotonin levels are going up and down. You might get the sort of gentle adjustment, but 2A is going to take more to really get those systems going. In fact, some of their activity might be modulated by things other than serotonin. So, there's evidence that things like carbon dioxide, maybe blood acidity in general is changing their activity and maybe the endogenous DMT system which we could talk about later. So, psychedelics tend to come in about two classes. One are classic psychedelics. One would be based on the structure of tryptamine, the precursor of serotonin, and the other would be based on the structure of phenylethylamine, precursor of catchatolemines. Both of these seem to be active at the 5-HG2A system which a variety of lines of evidence have been shown to mediate the effects of psychedelics to a very substantial degree. So, in the machine learning type work, I've been collaborating with Jara, Shikspahi, to try to reverse engineer these different neuromodulators as machine learning parameters. And so, we've been using an architecture which has elements of Hans Schmidt-Huber's world models and Dana Jara's dreamer to interpret 5-HT signaling as potentially being a... as the different classes' receptors as parameters for imaginative planning or in active inference terms, a sophisticated inference. So, with respect to world models, you'll see, is this working? Yeah, it's working. Hans Schmidt-Huber's architecture, they'll use these autoencoders to take these different frames. This is in a DOOM game. They'll compress them into a lower-dimensional latent space and then they will unpack these and through this training of the autoencoder to learn how to reconstitute these frames, you get this low-dimensional... source of low-dimensional vectors in latent space which can be used for training and with much better computational efficiency and some learning properties when you don't have to go all the way back to pixel space where things can get messy and you don't know what's causing what. If you work with this latent space, you can come out substantially ahead in terms of training. And so, you have this variational autoencoder which is learning how to compress scenes. So, this is one aspect or compress the observations. You then feed this latent space compressive vectors into this recurrent network which learns to predict the state transitions. You shove this into a much lower-dimensional controller which they trained with evolutionary strategies and then use this for policy selection, generating new observation and the standard sort of reinforcement learning paradigm. The interesting thing about these is they could also be run offline. So, you're able to take these and take the outputs and feed it right back in and you can do imaginative rollouts. So, you could use this for planning or for having the agent train in its imagination. And so, this would be an example. I believe the left would be what's seen or the left would be the compressed latent space and then the right would be the agent... I think the left is what was actually seen. This is pixel space and the right is the imaginative latent space. So, it tends to be... You're not really seeing it here, but it tends to... These imagined trainings like imagination relative to perception or sensation, it's less fidelity but you can still use it for the sake of planning and learning. But one issue is that these agents can do a kind of self-adversarial attack when they're training in their imagination. So, if the temperature or the uncertainty isn't high enough, they can basically come up with these sort of fantasies of easy environments where they are just winning, but then this doesn't transfer to the real world. And so, by introducing a temperature parameter to increase the uncertainty of your imaginative rollouts, this allows for better training back to the real world. And so, this I've been suggesting is potentially one interpretation of the 5-HT1A system. As increasing the temperature, putting the organism... Basically, instead of dopamine would be you're confident, go forth and act. Serotonin acting on the 1A system would say, we're not so confident, stop and think or imagine. Don't act right away, but let's think this out what's imaginatively planned this out. Look a little bit in our mind's eye before we leap. The two-way, though, seems like... It's almost like a hybrid of the 1A and dopamine where it's like, okay, imagine more, but it's not necessarily doing so under a state of uncertainty. Within these imaginative rollouts, things are appearing more vivid where the gap between imagination and reality seems to be narrowing. And the idea is, if you're thinking back like passive versus active coping, so you're uncertain. Let's say the 1A system is in the mix now. You're, okay, I'm not going to overly act. I'm going to stop and think. I'm going to reduce the gain on the expected free energy or the expected value of pursuing a policy right now. Patience. But if the two-way gets in the mix, it's going to tend to be more active. You're going to be imagining, but you're going to be more likely to act because you're imagining is because you're... They're more compelling, more vivid. So now we're coming back to psychedelics. But I want to lay out this foundation because the whole way through the method is basically a kind of Marianne neuro phenomenology, trying to understand the mind at computational, functional, algorithmic, and implementational levels of analysis, and then mapping this onto rich descriptions of subjective experience. So mapping standard, the cognitive science, multilevel stack to phenomenology. And so the machine learning is part of pulling off this mapping. We call or active inference is part of that. So to rehearse the basics of rebus, whether am I coming across all right or just checking in. Okay. So the relaxed beliefs under psychedelics. This I'd say might be the dominant paradigm for understanding the significance of the 5-HG2A system in creating the effects of psychedelics under a predictive processing paradigm. So according to rebus, you will get... I don't think I have to rehearse predictive coding here, but very briefly, the idea is you have a hierarchy of predictions where you are passing your priors or your prediction errors down the hierarchy back towards the primary modalities where the observations are coming in, and then you're passing prediction errors upwards. And this coding scheme is efficient, especially if the events that you're tracking or trying to model are clocking slower to the internal states. So something like telecommunications, if all you have to do is transmit what's different between the frames and 24 Hertz used to be what TV was back in the day. But if most of the pixels are largely the same frame-to-frame, and all you do is you show what's different to get this energy savings, and when you're dealing with an organ as hungry as a brain, two to 3% of your body mass, 20% of the budget, that the savings could be substantial. And in a nature world of the quick and the dead, natural selection might have picked such a thing. So within predictive coding, the predictions are thought to be encoded by these... So if you think of Cortex, as Jeff Hawkins says, it's roughly a dinner napkin and thickness and extent that's been crumpled up and shoved into your head. Cortex has these different layers, roughly six. And this is distinct from levels of a hierarchy, but the Cortex, the sheet, if you go across the napkin, there's these layers, about six. And depending on where you are, they're more or less distinct. But the L5 or the deep pyramidal neurons, these are the ones that will leave Cortex, do things like form loops with thalamus, the stratum, cerebellum, spine, and that's what is thought to encode your predictions, these L5 deep pyramidal neurons, not deep in terms of a deep polyp hierarchy, but deep in terms of relative to the cortical surface, more in there, and within the napkin. The prediction errors are thought to be encoded by the superficial pyramidal neurons, layers two, three. So L5 is forming these big synchronous complexes with a thalamus, probably not the only way it does the predictions, but this will be aggregating information from multiple different places, giving you, and because you're able to get a lot of different causes brought together, it's going to be a good source of priors. It's going to be a good source of your expectation and your modeling. And then you use this to try to suppress the ascending stream of prediction errors along this superficial layer two, three. We can go into layer later if that's useful. But the idea is that with underpsychedelics, you take L5 pyramidal neurons, these excitatory neurons, and you excite them so much with 5-HT2A stimulation that they fall out of sync with each other. And it's a kind of paradoxical effect. When they fire too readily, they don't coordinate anymore in establishing a synchronous rhythm. And because of this, they're actually paradoxically relaxing. The priors become less precise. The belief or free energy landscape becomes flatter. And so the idea is with your expectations being less, more information from the periphery is capable of penetrating deeper into the cortical hierarchy, registering more in consciousness, and impacting, learning and updating to a potentially greater degree. And so this is a model of the, much of the phenomenology of psychedelics, as well as their mechanism of therapeutic action for helping people to, as Colin would say, like how to change your mind, helping you to change your mind, to update. And so I think rebus is largely correct, but I've been introducing a potential refinement of it in something I call albus, which includes, in addition to rebus effects, something I call siebus effects or strengthening beliefs. And so the idea is that you might not get exactly this desynchronization effect along the whole dose response curve of 5-HG2A agonism or stimulation of the 5-HG2A receptor. Rather, this might be the case of what you would expect from very high levels of agonism, really strong excitation. But let's say we increase the gain just a little bit or a moderate amount. I suggest it's not at all obvious that this would cause the energy landscape to flatten. This would not cause them necessarily to fall out of sync with each other. In fact, they might be able to sync up even more powerfully if you just increase it a little bit or moderately. We don't actually know. This might vary from organism to organism. But the idea is we might need different regimes, different accounts depending on where we are with what level of stimulation before we would expect to see the kind of paradoxical desynchronization or potentially more straightforward enhanced synchronous ability. Yeah. So bring it back to predictive coding. So here if you're going to try to illustrate this, the purple would be your deep pyramidal neurons supposedly encoding or that are thought to encode your beliefs. And then up top you have the white circles. This would be your superficial pyramidal neurons, the upper part of the layer. And this would be encoding your prediction errors that you're passing upwards. And so you're seeing these, I'm illustrating the synchronous complexes of these red swaps. The predictions going downwards are these red arrows. And then I'm depicting basically the prediction errors as these gray swaps and black arrows. So within predictive coding, there's been significance assigned to different EEG and MEG frequency bounds as potentially corresponding more or less to predictions or prediction errors. More specifically, the slower rhythms like alpha and beta where alpha is maybe roughly 8 to 12 times a second, beta is maybe 13 to 30 times a second, that these slower rhythms would be the ones encoding your predictions passed downwards. But that your prediction errors might be more likely to be encoded by gamma frequencies, you know, anywhere from 40 to 120, even higher oscillations per second. So within this part of the reason that this might make some intuitive sense is if you're thinking of the formation of a synchronous complex of neural activity that faster would be smaller and slower would be bigger or more inclusive, more integrative, this might make some intuitive sense. If you're thinking of, okay, so you're trying to get this agreement, this negotiation of a common rhythm of a bunch of signaling units. If you're having more units spread over a bigger distance, you're trying to get them to go come into agreement, this might necessarily have to be a slower process. But if you're, let's say like quantizing prediction error, generally like a little packet of like a local agreement that you're passing upwards, well, that you could potentially clock a lot quicker. So that's the basic idea there. So if we were to look at rebus, it would look something like you see the right bottom, the swath of red is more restricted in scope, and the arrows are a little bit smaller. And so this is idea of you've relaxed your beliefs. This is, if you get this paradoxical desynchronization. But, and so once you have this, you're seeing less of these prediction errors are being inhibited by this descending stream of predictions and they're able to make their way in more deeper into the hierarchy. And this would be showing, you know, one modality. Obviously the brain is not just one singular hierarchy, but it's a heterarchy, multiple intersecting hierarchies corresponding to each of your modalities all brought together, stitched together into a heterarchy. Up top would be illustrating a sebus effect. And so the swath is bigger and the arrows are also darker and bigger. And that's, and more prediction errors are being inhibited. There's an additional factor, which is five HD2A receptors are also found on the inhibitory intern neurons of layers two, three. And so in theory, and there's some evidence, I believe that suggests this, this would correspond to a default reduction in the gain on the ascending stream of prediction errors or you would say an active inference speak less precision weighting of the precision of the prediction errors. And so they have less of a contribution of the Bayesian updating that's happening at each level of this belief hierarchy. And so that's illustrated by slightly offices. Yeah, the on the road, the arrows are a little bit smaller. So the idea here with a sebus effect would be, not only might you have stronger beliefs, but these beliefs might be further shielded from contradiction by inconsistent sense data. I think this potentially provides a fairly parsimonious account of some of the conditions under which you get a certain perceptual alterations of a hallucinatory variety. So for instance, people will sometimes report hallucinations and sensory deprivation tanks. They'll sometimes a, I think it's called a Charles Bene syndrome where as you start to let's say go blind or deaf, you'll start to get like auditory or hallucinations within the modality as you lose input from the external world. This also you could think of at a higher level. I don't know if this works, but you could think of potentially delusions in a similar way, a belief which is overly strong and resistant to updating from a potentially contradictory and potentially correct sense data. So in theory, for talking about sebus effects, you might have a single parsimonious account for both hallucinations and delusions that's incorporating a little bit more of the neuroanatomy. That being said, I'm not saying at all that Rebus is incorrect. I actually think it's probably very correct and maybe crucially important for explaining some of the really powerful changes that people observe with psychedelics as a therapeutic intervention. So get into a little more details here and say, okay, are you all with me still? Okay, cool. So I mentioned a heterarchy earlier. So I've taken these predictive coding hierarchies here and I'm just illustrating them as these triangles. And so here would be, let's say, you're... So I think in this framework of Albus I'm suggesting, we need to even, in addition to including things like sebus effects, I think we're going to need to take on things like basically consciousness and higher order consciousness to actually really adequately explain what we mean by relaxed or strengthened beliefs and what actually is happening with these different thermodulators and maybe attempts at creating analogs. So here up top would be somatosensation. To the right would be your vision and then your hearing and we're going to try to bring all these things together. And so if you're thinking of some sort of hierarchy, like here you might have like small beta complexes and you can think of each of these as like a small modeling effort and a small belief over whatever is in the scope of this synchronous complex. Where the idea being that when you synchronize neurons, this allows them to be aligned in their activity and time such that their signals can sum and be integrated rather than decay. And so this would allow you to have potentially a joint belief over whatever is in the scope of the synchronous complex. Pasco-Freeze calls this communication through coherence. And for a reference for this idea of like an inverse hierarchy of size and speed, Buzaki would be a good reference for that I'd say. So here we are on the left, the beginnings of modeling of increased hierarchical depth with small fast beta complexes, the extraction of things like line segments and maybe basic contours and vision, basic patterns within sound, within touch. But then let's say we nest these within maybe slower, broader, more encompassing beta. We could have more hierarchical structure for our modeling effort. And so here you might start to get the beginnings of things like objects as you bring together these different features. The beginnings of things like objects with respect to touch or rather your own body in relation to those. The beginnings of things like meaningful sounds. And so then let's say we bring all the modalities together now, we let them cross talk and you have a multimodal representation that you've organized according to a coherent perspectival reference frame. Well, now you might get the beginning of something like phenomenality or experience. And so here there's a blue swath corresponding to alpha frequencies here where at alpha frequencies that would roughly be, it looks like the size that you would need, so eight to 12 times a second of frequency of updating or sampling or estimation. But this would be big enough to get you the entire sensorium all brought together bringing together that the different hierarchies into a hierarchy. Alpha has some additional significance and that it has particularly concentrated sources of alpha power from midline structures. And I take this as being important because if you look at something like the posterior midline, we look at let's say the posterior singlet cortex which is heavily implicated and the psychedelic and also the meditation research literatures as when alpha goes down there seems to be some associated with phenomena such as ego dissolution. But the idea that alpha might be important for coherent egocentric perspective in addition to bringing together your sensorium for this mutual information, these midline structures in particular the posterior singlet, there's this old idea of the Papez circuit where it's this core circuit for emotional memory that would roughly go out of mammillary bodies around the brainstem. So it goes up through the mammaliflamic tract I think anterior lateral phlamic nucleus goes through the singlin bundle, posterior singlet into the picampus, out through the four inks, back to mammillary bodies. Don't worry about those details. The important thing about this is this was interpreted by actually I said Papez, I think it's Papez but it was interpreted by Papez when he was etching out these elect sources of epileptiform discharge as his core circuit for emotional memory. Part of the reason that, and so the importance of this is the mammillary bodies are receiving stretch receptor information from the neck. And so this is going to be potentially important for knowing where your head's pointing, especially for talking about like, which will be important for us and maybe even more so for talking about like our inner rodent or whatever, that where your head's pointing is going to really help for bringing coherence to explaining your sensorium. Also at this salamic relay you're also getting inputs from the vestibular apparatus of the inner ear. And so this is giving you like the yaw pitch and roll. And so not only do you have your entire sensorium all the information at these posterior midline structures before they funnel into the hippocampus as a source of new memory, but only do you have the entire sensorium brought together. You're on top of it all in network terms but you have the information you would need to organize it by your egocentric perspective. And so this, yeah. So we've got predictive coding we covered, SIBIS effects, RIDIS effects. And now I'm introducing like a basic model of perceptual synthesis where we have something of a Cartesian theater here enabled by the ability of slow rhythms of a large enough extent to bring together the right kinds of information you would need to basically create a coherent world model. So y'all with me still or? We good? Okay. So in this preprint I describe how we can maybe think of with respect to either sensation or imagination different combinations of RIDIS and SIBIS effects happening potentially corresponding to different levels of stimulation of these receptor classes of the 5HT2AR in particular. By the way, I think 5HT2C, that's one to watch. It might even be like almost more strictly RIDIS in its nature but that's something to talk about later. So for sensation on the left here, this would be there's a person out in nature and they're having an interaction with a tree. And then the right here would be a similar thing but this is imagination maybe with your eyes closed, maybe not but this is here you're imagining yourself interacting with the tree. Now, imagination it doesn't need to be in a third person but I'm just taking advantage of the fact that we're talking like if you are to think of yourself in a third person if you're going to move from this, was it the, as James called it, like from the eye to the me to this objectified selfhood, that part is going to require imagination. You can imagine yourself in the first person but the third person point of view, you're going to require imagination, I guess that or a mirror or a good set up of mirrors maybe you could do it that way. So that actually might not, that might be non-trivial but on the left here. At the top we have let's say unaltered, if such a thing ever exists, we're always altered in some ways, we're always on some, our brain stems from a copious always dripping some combination of drugs to us. It's just sometimes it's a very different regime and we say it's altered but there is no like singular baseline. We, who knows how much we both vary as individuals and across individuals along these lines. But so as we move down, let's say we were talking about a rebus effect and let's say you were getting a rebus effect with sensation. So in theory, if your perception is entailed by the predictions then it could be the case that things might seem less vivid to you. It may not be the case that your perception is more vivid. I don't know if I would say this is strictly entailed by the rebus model, but that's one thing you might think potentially. If these slow rhythms, if some of these are actually entailing your conscious experience, if these are reduced in their coherence and their power, in theory you might get a diminution of perceptual vividness. However, in a siebus interpretation things might appear more vivid. Not necessarily more veridical, but more vivid. And so it seems like a rebus interpretation would say that things might be more detailed, more accurate, more richly textured with more faithful correspondences to the external world when you're perceiving, but potentially less vivid. Not because this is a strict entailment, but that seems like one suggestion. Under a siebus interpretation, this would be more vivid, but not necessarily more precise. It actually might be a little bit smoothed over, but it might be a little bit more auto-tuning or a filtering, but also, though, more compelling potentially. And so this, in theory, though, would be the phenomenology of microdosing in theory. And so, when again, and so I try to illustrate this with a brain, you know, bigger or smaller synchronous complexes here. So Albus, the idea is, well, we could have an admixture. It's not just one set of beliefs, but we have a hierarchy of beliefs and in fact a heterarchy. Alpha starts to get pulled back, and that loses coherence, and that seems to be one of the strongest associations in the psychedelic and meditation literatures. Alpha is what people will emphasize as some of these ego-decentering or dissolving effects, but maybe beta stays the same or even increases in power. And so under a rebus model, for instance, the idea would be that, you know, things like alpha go down in power, but things like those particular smaller beliefs, they can come together in different ways in different sort of creative combinations. And so you might get something like synesthesia with psychedelics as like a creative recombination of things under this more anarchic state. Once you took the alpha self-ruler and you reduced its tyrannical grip over your perception. And so this more free-flowing perception, you get all sorts of different creative combinations and consciousness. And so, yeah, so that's one thing, is trying to basically map different combinations of rebus or sebus effects that you might expect under different doses with respect to either sensation or imagination. So really digging into the phenomenology and using psychedelics as an opportunity, actually, in a way to indirectly test models of consciousness. But also, if we're going to explain what these substances, powerful substances are doing, we really actually need to take some of these questions head on, maybe even the hard problem. So some of the more, so in this pre-print, I also are getting into, okay, what do we mean by beliefs? What kind of beliefs are we talking about? And so one kind of beliefs would be, so here you might have the beliefs, with active inference and the free energy principle, it's beliefs the whole way through. Well, which beliefs do we mean? And rebus is saying what specifically is being relaxed is your deeper beliefs, your internal working models, the higher level beliefs corresponding to things like selfhood and its relation to the world, things like autobiographical and narrative selfhood. These are the things that are losing their grip. And within rebus, I don't want to, there is room for sebus effects in some of the ways it's discussed, where it's not just across the board, all the slow rhythms are going down. I've heard it discussed differently. And so you might get an indirect strengthening as you remove these slower integrative and potentially suppressive processes. But in terms of types of beliefs, if we're going to be talking about cognition or if we're talking about something as like highfalutin as the self, now we might want to say like, well, what kind of self are we talking about? Like we're talking about like a minimal embodied self, are we talking about an extended self? And even those, like there's all sorts of bad questions, like what goes into the minimal embodied self and how it's relating to the extended self and how many, maybe there's different kinds of those. And so if we're talking about the construction of self and its relation to the world and objectified selfhood, the ego, the self as construed over time and across circumstances, we're going to need a fairly rich account of cognition. And so here I'm basically providing an account of sophisticated affective inference or imaginative planning and policy selection as orchestrated by the hippocampal and entorhinal system. And so this is drawing on some of the work I'm doing with Tim Verbalen and others trying to decipher that system. But basically, you know, you seem to have, so here hippocampus can be thought of in a way as at the top of the cortical heterarchy. When the prediction errors are not, so as the prediction errors are percolating upwards, being contradicted or stopped by predictions, you stop them as low as you can as close to the modalities as you can. That's the maximum savings and that's what you would expect with the maximally efficient prediction and learning. But if you can't predict a pattern, if nothing, the genuinely novel, that makes its way all the way up the heterarchy into the hippocampus as this prediction error storage place of last resort. And this then can be stitched together to create memory. And specifically what the hippocampal and entorhinal system seems to do is it encodes this sequence memory. It has multiple functions, but one of the ways in which one of its core functions seems to be the encoding of trajectories of the overall organismic system through time and space. And so if you look at the work of people like Reddish, you'll actually see, for instance, as this rodents is in a maze, figuring out where it's going to go, you'll see these sweeps of activity within the hippocampus kind of play out in different directions across these place fields or this mapping of space. And then the organism will take like a chance for another of these potential sweeps of predictive activity. And so I probably don't want to get into this here, but I have an account of like the anterior versus the posterior hippocampus seem to be qualitatively different, where the anterior hippocampus seems to be the one that's more predictive, giving you these sweeps, giving you things like counterfactual processing. The anterior seems to be more closely coupling with the posterior sensorium and will be more important for more specific episodic encodings and maybe more faithful episodic rememberings. And so the idea is you have these little hexagons which are corresponding to the mapping of space of these hippocampal place fields, whereas the organism within some sort of space. And then they're playing out in some trajectory. And so now along this trajectory, it is you're having the predictionaries from the cortical generative model, which can basically, by pointing back to cortex, can unpack at every point in space the experience you would expect for that particular temporal spatial trajectory. And this would be an account of episodic memory and episodic imagination or the stream of consciousness, if you will. Okay, so here we are. And I think this is the last thing I'll... So here we are with these trajectories being played out and different experiences being entailed at each point as it plays out. And they have a given length. It seems that there's a extent of this, I think James called it the specious present. Or the idea is there's like a thickness of the moment we find ourselves in. Or this might be a somewhat different idea, but there seems to be three seconds. It seems to be roughly the upper bound of what we would call a given extended moment of experience, Edelman I think called it like the remembered present, a little bit of like a prediction and a post-diction and you're moving back and forth in the sliding window frame of experience. And so one interpretation, and so within here as you're doing these different episodic rememberings and imaginings that one thing that the 5HG2A system might do is both increase the vividness of what you're imagining or perceiving as you go, but also allowing you to have more extended rollouts before you have to reset. Where it seems like a roughly three seconds might be the upper bound. And so basically I have within the preprint this comic description, this neuro phenomenological comic where as you're increasing the dose imagination is getting closer to perception and you're getting a longer runway to work with for this extended present moment within which you are planning and learning and making sense of things. And so it gets a little bit bigger with the micro dose, a little bit bigger with this threshold dose. Things are even more vivid. When you started out, you're just like, I'm kind of moving through the world. I'm kind of awake. I'm kind of conscious, I guess. And then it's like, oh yeah, like things are crisper and a little bit more creative, like not that much here, but a little crisper. This will be something like, I don't know, like a program or in Silicon Valley, like micro dosing is very popular. It doesn't seem like they're trying to like trip, but they're almost using it like a little bit of a stimulant, trying to like increase absorption with whatever they're doing. And so then, as we move through, you increase the dose some more and now it's even more vivid and more creative, more different imaginations or having access to more and more is manifesting in mind. But then after a certain dose, it seemed, what I'm illustrating is that basically the coherence starts to go and you start to get smaller rollouts. They're even more vivid and they're more creative, but they're becoming less clear in terms of the sequencing. They make less sense with respect to achieving particular goals. And so I think something that I should explain and we might have to talk it out, but the underlying this is, it's okay, these hippocampal resetting events are occasioned by different things. One is if I go from one room to a next, if some sort of like really surprising thing happens, that's going to trigger this sharp wave ripple activity and a retiling of space within which you're doing your sophisticated inference. So something either, so there's some evidence that reward prediction errors and something really, really good happens, that might cause you to basically ramp up the hippocampal activity and code that, generate like a phasic burst, generate a phasic burst of dopamine, learn it, and then keep going or if something really bad happens, this might be an opportunity to pivot. What you're doing is not working. Let's retile, let's try again. And so this will be a function of both things like, what's the neuromodulatory tone? What are those like the circuit level conditions? Like maybe directly, what's your level of 5G to a simulation that might influence how long you get before you have to reset. But it could also be explained at the level of experience itself, at the level of the expected free energy changing as this immersion cause. So as I'm getting creative, if I'm getting so creative that I'm no longer making sense of things, my expected free energy, if that goes up, like that spikes up because what I just said is really surprising. It's novel, it was experienced vividly and it's not making a lot of sense. That might be something that would occasion this resetting. And so the idea is a model of basically a stream of consciousness or episodic remembering and imagining, which becomes more coherent, more rich, more absorbed within this first regime and then a second regime that's more properly psychedelic where it becomes extremely creative so much so that it becomes qualitatively different than normal experience. And even so much so that, for instance, at a certain point, you're not going to be using this to enhance your job as a programmer. You will be doing this on vacation with trusted others or something like this hopefully. And this is not something to take to work. People are not mega-dosing at work hopefully or they will not be working there for very long at most workplaces. And so, yeah, that's basically... And so then coming, looping back around to the beginning, the idea is with a deeper understanding of these systems that includes some rebus aspects and some sebus type aspects, we might be able to use these as inspiration for novel, active inference machine learning parameters. So potentially you can think of like getting stuck at a little, people, you know, if they're living like a life of quiet desperation or something, they've gotten stuck at some kind of local maxima and they're looking to change and they're looking to get, to kick out of it. They're not finding an elegant path to minimize their expected free energy. And so if you've found yourself trapped at this local free energy minima, you can use psychedelics to turbocharge your sophisticated affective inference enough to maybe find a novel solution, see something that you would otherwise. And then I guess the final thing I would like to point to is, in terms of using psychedelics as a probe on mind, you might have different interpretations of what they reveal if you're thinking of either a strictly sebus story, a strictly rebus story or some kind of combination of the both of them. And so I'd suggest that for instance, you can think of a good amount of psychedelic phenomenology as corresponding to the unmasking of your core priors. And so things like, let's say fractal phenomenology, that maybe would be a core perceptual prior maybe of a partially evolutionary, partially as a reliably learnable posterior or empirical prior for making sense of the world. And this would make some sense in that if you use fractals as generating your predictions, that's a very efficient basis set because you can do a lot with different combinations of fractals. And this is especially going to be efficient if that's what nature itself is doing. You're using different combinations of fractal processes to model a world which is largely structured by different combinations of fractal type generative processes. Further, things like entity perceptions associated with different psychedelic experiences, archetypal type experiences, those also could be thought of as the unmasking of core priors, potentially strengthened core priors. Because you would expect from active inference that entities is something, the kind of thing you would expect as a very powerful empirical prior because that's how we survive, especially agents such as us who think through other minds and where our primary niche is largely each other. And so whether we're, the details of psychedelics might actually reveal these foundational inductive biases or foundational lessons in the learning curriculum and the structure learning whereby we come to be the kinds of beings we are. And that is it for now. Okay. Thank you, Adam. That was super awesome. I am particularly overloaded right now on my 1A receptors, which is I just want to imagine and step back before I ask some questions. I know Daniel asked a lot, so I'm going to let him ground me a little bit in reality and have the first jab. I was just writing about just stuff in the live chat, so definitely anyone can ask you jumped through many areas. So if anybody has a question, they can type it, but let's take a quick breath, collect our thoughts, and then blue, you can ask the first question. Wait, I just passed to you to like ground me. All right. So I guess I'm already here. All right. Since I'm already here, I'm going to like unclearly ask a whole lot of questions at once and also like try not to fangirl out too much because this line of work is really super awesome and unifies a whole lot of my particular interests. So just really quickly, what do you think the relationship is, if any, between inference and memory? And can we quantify that through the brain, the brain system, dopamine, serotonin? Do you have thoughts on that? I'm sure you do. Or do you have thoughts that you can summarize in 10 minutes or less? So in terms of memory, there's a sense in which it's all, there's a sense in which, within the Free Energy Principle and Active Inference, all effective causation is inference and it's all inference based on different kinds of memory. So every belief is a memory and everything that happens is a kind of action, is a kind of inference. But there's another sense though of like memory, the kind I'm trying to get to, like the focus on the hippocampal entorhinal system, this orchestration of the overall memory system and this encoding of the genuinely novel in these episodes and in these frames of sense making, that this would be a more complex kind of inference and a sophisticated one, specifically where you are inferring the most plausible, rather than, so there could be a sense for instance, in which you might get a fairly, under some circumstances, maybe for like recent things, that aren't like, I don't know what the shelf life would be, but like, so some of the literature, a couple of weeks, couple of months, I don't know, but like that actually hippocampal system through just this chaining of attracting states along these trajectories could point back to cortex and play out something that's actually like, a decently faithful but still very biased account or reconstruction, a remembering of what occurred. But for things that are older, that's probably not the case, even for things that are newer, there's this constructive, so it's always constructive and it's always a kind of, every moment perception is inference and every moment, there's a creative aspect of filling in, but actually like situating yourself within some sort of causal unfolding, that's gonna involve a lot of value, canalized policy selection. That's gonna involve you using constraints like, well, what kind of being do I think I am and what's plausible? And that's gonna determine when you're going down this like, this branching forking paths of different possibilities, like you'll be moving between like, well, what might I have done under this or that and you'll be picking plausible accounts of you, partially as you go. And so some remembering might be a little bit like the naive like playback one of you play back the trajectory, it's got pointers back to cortex and then you have this belief hierarchy and so you give the prediction errors back to where things were new and then it just unpacks itself as a good enough account of what happened at each point as you went through some trajectory and it's a decent semi-faithful play up, but there's also gonna be this other kind of memory though, that's not faithful at all and can really deviate a substantial amount of where and hard to distinguish from imagination and continuous with it. It's like, how do you know if you're remembering something or imagine you use these different cues. One cue you might use is perceptual vividness. So you would assume like, if I remembered something it's gonna be a little bit crisper and if I just like totally made it up. The other one might be like just the plausibility like yeah, that's like the kind of thing someone like me would do but you can get it confused and that would be a source monitoring error and I think that's actually in a way for psychedelics both in terms of this core process of figuring out who you are, where you've been, where you're likely to go and where you ought to go, thinking of these pathways as potentially what it is to be an agent as revealing ways that agency and different self-consciousness and selfhood can vary but also in terms of this I think needs to be brought in in terms of what we might expect to be impacted and what we might want to prepare for and harness with things like psychedelics. So for instance, there might be a very real risk for phenomena such as false memory. So if the gap between perception and imagination starts to narrow if your imagination starts to seem more vivid it might become more difficult to distinguish between a remembering and an imagining or, yeah, so things of that nature. So I think the issue of what is memory in its relation to inference is going to be key for what it is like we want to alter and what it is we maybe don't want to alter but might accidentally alter when we do different types of interventions. So you played so perfectly into my follow-up question was how does creativity play into that? And what is maybe the function of creativity? I mean personally my autobiographical memory is awful but I have a really stronghold good factual memory for all kinds of ridiculous things. So where does creativity play into that memory, perception, inference, paradigm and are the people that are more creative or that take more psychedelics like do they have a less accurate memory or are they more likely to remember things inaccurately or even what is inaccuracy? Sorry, let me just get super meta because I really fundamentally believe that perception is different for all of us but what is written in my biochem textbook less so. So maybe that's why I put less priority on my autobiographical memory because I believe it's functionally inaccurate but whatever how many angstroms are between each subunit of DNA that's way more important. I mean that's one of the hardest questions I can think of. One of the most important too so I'm trying to forget the name of the paper it's this dual constraints model of creativity that does address like head on the issue of kinds of creativity and psychedelics I'll post that later but it does seem like for instance there are I wouldn't say necessary trade-offs with respect to different processes as to whether they allow for realizing different types of value so let's say with memory or with creativity you might distinguish between divergent and convergent creativity so like convergent creativity where like you want to hit a particular target and generate a novel solution for a particular challenge or divergent where let's say you just want to think of something new maybe just a novelty generator from which you then might select later convergent solutions in a longer process or just as part of an open-ended exploration and so there's different ways like you for instance you could get if you just turn up the heat, the temperature and you make things more imprecise and unconstrained you might be able to get all sorts of different combinations and this might be a good like fountain of divergent creativity some of the evidence suggests I believe that even with the micro dose regime so I think people who are doing like micro dosing usually they're looking for things like enhanced conversion creativity I don't think the evidence is great that that's actually being achieved but divergent creativity might be more likely to be associated with psychedelic experience proper combinations of things that just you never would have imagined it's not clear it's not necessarily even useful for any particular thing it's not hitting any particular target necessarily it might be after the fact it might reveal something to you that you never imagined that made all the difference maybe like it juiced your defense mechanisms in terms of like your policy selection was leading into garden paths you're seeing some things and not others and then it took the blinders off it revealed something to you that you might have missed but I'd say that people definitely very partially maybe due to things like low level parameter tunings like different people will vary in terms of like the natural level of the gain on these systems the natural tendencies for occupying these different regimes and so you'd expect that to vary either through some combination of things best explained at low levels like you know different alleles of neuromodulators and some of it would be not well explained at that level it's more just that's the way the structure learning happened that's the way the system bootstrapped so different kind of agents and everything in between so one thing that I find to be like I want to go into is that humanists in general us as a creative species that these pathways might have been important for allowing us to have be more creative and being able to combine things in a broader range of ways and that this potentially you know unlocked a more flexible type of cognition maybe a greater capacity for analogical reasoning the kinds of things that might have potentially precipitated this great leap forward in the birth of cognitive modernity where we became a symbolic species and cultural evolution came off the ground there's an older idea by McKenna called the stone date theory that like there was actually like a period where we were in this symbiosis with mushrooms and we're eating psilocybin containing mushrooms and this actually helped to get that going I don't think we could rule that out but you could also get something very similar with a you know a couple high impactful mutations along these pathways that doesn't involve a mushroom eating phase and so there still might be a really interesting account of actually the secret of our success as a species as a particularly creative kind of hominid who's able to bring things together in more novel imaginative ways that could be involved and being continuous with the story of psychedelics and how they work and in the present day so that obviously you know I'm extremely interested in and also as a source you're saying individual difference like you know are there more like to this day more shamanic personalities are there you know are there more and to what degree can this account be told at the level of genes or epigenetics or is it not well told there one yeah so that's those are some thoughts on that one okay so wait now you just like walked into this trap that I laid for you so I've now got to ask this next question so you talked a lot about psychedelics and sensory deprivation and now you want to talk about like our symbiotic relationship with fungi and I'm a huge like Terence McKenna fan so if you've read the archaic revival which I'm sure you have so where do sensory deprivation in conjunction with psychedelic use like where does that fall on this like perceptual perception action mental action alpha beta brainwave serotonin dopamine spectrum like I just want to know your thoughts sorry not sorry but I'm sorry so I don't know what this exactly gets at it but it's possible that this would be rather different but like okay this is this we're really going to conjecture land here but so I forgot what the book is it's a recent one about us is so that there's like two domestication books recently about humanness but so one of them is the goodness paradox and us as a self-domesticated species then there's this other more recent one but the idea is that this serotonin pathways if it depends which ones we're talking about they'll be quite different but if they are up-regulated let's say of a 1A variety you could think of that as being more conducive to something like a domesticated phenotype so if your dopamine is higher and you're having your you're more confident in your policy selection then that might also cause you to be somewhat impulsive and that might make you not necessarily be able to discount utility enough hold back think things through a bit more and if the dopamine is too high you might get less cooperativity if it's and so you could think of the serotonin to the patients that it's the 1A in particular as being of a domesticated variety and things might appear for instance less vivid perceptually and imaginatively in theory if we're just talking about those pathways the 2A for all this there's so much we don't know so much of this is conjecture but the 2A does seem to have a little bit more of a dopaminergic character to it and and actually it seems like that pathway evolved roughly around the time of the Cambrian explosion around with the advent of Jaws in the context of predator prey arms races and so that might be consistent with like a dopaminergic significant so this like temporarily like increasing your confidence your policy selection like turbocharging your action selection of bits you can go out and get the goal of either getting your lunch or avoiding becoming lunch and so let's say you're moving into greater gain on imagination let's say we're thinking serotonin either of the 1A or 2A is doing this this then might compete with more externally focused perception sensation and so you're imagining more but because your mind's eye is occupied more with your imaginings you might be attending to the details of the environment a little bit less and so the species or an individual who might have to gain for whatever reason higher on these pathways you might call this a little bit more shamanic or maybe a little bit further on like a schizophrenia type spectrum might be perceiving the world slightly less faithfully but their imagination is now freed up to be more freewheeling and more creative and this is allowing for different things to come into being so very speculative we might think of like neanderthals as like less domesticated humans maybe even a little bit more like autism spectrum but like better able of perceiving the world precisely and us we might have been a little bit more shamanic a little bit more although we are the neanderthals also because we all interbred and the variance within is probably bigger than the variance between but also might our ancestors and us we might be somewhat less I don't know if I'd say less dopamine I don't know if I'd say that so that's what I got for now Thanks Adam great presentation just wanted to make two quick notes of the ask some questions from the chat so the first is there's no free lunch with strategy so there's niches that are just unwinnable where the phenotype cannot win and then there's other niches where potentially a broad range of phenotypes all can proliferate so whether or not a given strategy is successful or under what cases a given kind of neuromodulatory state is good or bad by what metric there's just no single free lunch or preference for lunch food then the second point was the complex gene family point was very interesting and you mentioned some other receptors in the serotonin receptor family and there are many receptors including ones that don't have nice clean names like number one and number two and number three so at least 14 investigating function in these massively interacting and often redundant like if seven does something but only when eight is gone what's the actual mechanism there so here's some questions from the chat so first question from the chat Steven asked interested in the approximately three second window of conscious awareness Adam mentioned does that trace back to William James stream of consciousness awareness and the temporal scale scales of experience I believe so it's kind of like a three second echo is our experience or what does the three seconds represent so so the empirically established phenomenon is there seems to be stitching things together into one continuous unfolding there seems to be a three second upper bound within this continuous unfolding I believe there is relevance to James's species present and Edelman's remembered presence idea of a kind of anticipatory prediction and looking ahead and a post dictive going back to you're moving back and forth of yourself in this unfolding and the speculative part is I am suggesting that this might correspond to the a hippocampal entorhinal tiling of space either a physical space or a conceptual space within which you're doing this and that the the reason this upper bound would be the length of time you have before you need to refresh the attractors and retile speculative and then and the idea would be that this would correspond to something like a these sharp wave ripples that you observe we don't know for sure it's a perfect lead into this second question and then I have another comment from Cambridge breaths so you just said about ripples so they wrote wonder if investigating cortical traveling waves via neural mass biophysical modeling approaches in particular the work of Wilson Callan can contribute to this work if yes how many thanks for elaborations Timmerman in collaboration with Robin Carter Harris has done some really interesting work along those lines I believe it was with DMT looking at traveling waves I believe in the alpha frequency as evidence for the Revis models and so the idea would be that if you are actually relaxing your belief landscape you might get a reduction of like traveling waves of predictions going down towards the periphery and maybe greater inward going prediction error waves and so some work has been done along those lines already by Timmerman and others that's really interesting and as I said before so although I'm like introducing additional factors I do not think Revis is wrong I just want to add a couple additional details of some of the contributions from modulations of cortical micro-circuitry might be different you might be able to get this effect from a variety of ways not just from this desynchronization but before I forget earlier you were saying about free lunches absolutely and I think this is part of like for instance let's say crank it all to the max if each one has its own function it will be depending on the niche you're in both generally and locally so we're talking about dopamine, too much dopamine you might be impulsive too little you might not strike all the irons hot and you might miss opportunities too little 1a you might be irritable and impulsive too much this is opposite problems too much it could generate over time a kind of psychotic predisposition that would be a loss of a free lunch the one more thing that you're mentioning is it's not just these two receptor classes there's actually 14 conservatively one thing I'm wondering though is whether you get kind of like this is potentially wishful thinking but potentially not you get the lion's share by considering a couple and that there might be like an order in which they evolved and a ubiquity where the ubiquity would be one sign for how important it is to figure these things out so the broader the distribution the more you get from explaining the ones with the broadest distribution so the 1a and 2a have this extremely extensive cortical and sub cortical distribution but that's not necessarily the case you could have for instance pivotal areas of the brain stem you're right on top of the core value signals one class of receptor there that could be the difference that makes all the difference another aspect of the hope that you can make progress incrementally with the goal ultimately we're going to want to get every single last one of these receptor classes is that the additional ones might have a common organismic cybernetic significance but they're different because of their different locations and so that their functional account might still correspond to let's say overall levels of serotonin overall levels of some neurohormone or blood acidity or some other homeostatic state or state of challenge and so that the active inference significance might be the same but you're tweaking differently in different neural systems so that the overall coordinate effect is coherent with respect to your expected for energy with respect to realizing organismic value or minimizing prediction error with respect to the prediction of you doing the things you need to do as a successful being and so one more thing though is I'll just throw off the 2C idea is potentially with these gene I don't know how far we can go with this but with this idea of like so with a gene duplication event you have two copies of the gene this allows you to basically take one of the copies because you got a spare that's still doing the same thing and now this other one can mutate and can separately be optimized by evolution to take on a different function and so the idea is that there might have been a common ancestor that looked a little bit more 1A like but that at the extreme levels of it during this gene duplication event this became the 2A receptor and maybe this goes one step further to 2C I'm not sure but maybe we go a couple steps further with this kind of gene duplication story and what we're corresponding to is peeling out and separately in a modular fashion optimizing for particular regimes of organismic significance are there going to be 14 to 17? I'm skeptical I don't think is that I suspect it's like maybe 3 or 4 top stories of this kind of receptors it's these core 2 to 4 stories probably 3 maybe being achieved by having different functions locally but in a coordinated activity the same story globally nice there could be a lower dimensional manifold that gives a lot of resilience and also a vulnerability for the system and learning and one thing we've talked about before but it's interesting to think about is in your presentation you did a really awesome job of connecting the general kind of dynamical systems and inference questions to specific human regions and then it's interesting to ask okay well insects have a different brain layout they don't have the entorhinal so we can juxtapose cognitive systems to learn a lot about which attributes of the computation are necessary and sufficient so I wanted to ask which cognitive systems you thought were interesting to study in that sense I have an intuition and a hope that may or may not pan out that the extent of convergence is fairly substantial now things can I think flip in all sorts of ways I wouldn't just assume this is correct but let's say we're talking about within an insect we would potentially look to the central complex which has a lot of hippocampal entorhinal functions and it might actually be an even better system to study in terms of you would have complete access theoretically to the whole you would have a much richer access via something like calcium imaging and so that could be looking at something like the attractor dynamics of there's experiments let's say they take fruit flies they put them on top of a little styrofoam bowl and hook them up to virtual reality and they have them doing different things you look at the central complex while it's doing that under different states of neuromodulation and we might see differences and a common computational account in terms of the mapping space and policy selection now how far back does this go some of this logic might even play out at the level of individual cells and like either like for instance like 1A versus 2A so low levels of serotonin seem to have an effect of strengthening cytoskeleton dynamics and so that would promote like an organism kind of staying in place but high levels actually destabilize the cytoskeleton and so they'll actually like literally create plasticity and fluidity of the morphology and allow the organism to move around more and so some of these accounts might go all the way back to the like near the origins of life and individual cells and metabolism how far we can go I don't know these might have even been repurposed for the different social insects communicating although you could see these things flipping because you know what goes in the individual doesn't have to be the same in fact because you have things in one significance on the individual level that might mean that has the opposite significance at the group interactive level so but I think they'll be extremely rich to like I my kingdom for that science to be done blue do you want to ask a question I will jump in here actually you know when we talk about the origin of life you mentioned earlier in your talk that the third person perspective is related to imagination and creativity and this goes back to kind of what I was asking earlier but is it so much that the third person is related to this kind of like coming outside of ego and looking at yourself you know from this kind of meta perspective or is it just fundamentally the dissolution of the self like how much does an ant think about themselves or like a bacterial that's part of a micro like or a biofilm like how much of that is imagination like looking at themselves as part of a whole or that's just like myself doesn't matter I'm part of this bigger unit so like if you're part of a well coordinated collective there's a sense with that within active inference you will have converged on a shared generative model for your coherent inaction that you're taking part in and so your role in being enacted there's a sense in which it's not exactly like on some level in some way there is information of some kind which is not necessarily egocentric in fact it's like your role allocentrically within the thing like even if on some level it's like what you are you as generative model if you were to give the generative model process as a whole a perspective it's not egoic it's not necessarily from a first person point of view it's from the coordination manifold of the group that's doing this common thing am I kind of getting it? yeah it's interesting like developmentally if we were to kind of take when these different perspectives of different kinds come online for instance right out the gate it's not clear to me that we have coherent egos it's not clear we attain to this right out the gate when we're born and so there's a sense by default the infant might be egoless and by default it's doing this more like an active coupling with its caregivers and has not yet been enslaved to an egocentric point of view there's kind of an interesting thing with language where it's actually like the ego itself seems like when we use it in psychodynamic terms so we say egocentric perspective it's like the I and it's the first person but when you say ego as a kind of extended unfolding of you that's in the world and all your possessions and your sense of yourself in a social context is a third person so in terms of egocentric perspective it's more first person but ego as a psychodynamic concept and the kind of thing you might want to modify with a psychedelic seems like a third person kind of thing so the question would be maybe there's an implicit allocentricity or non-egoicness being enacted by any coupling and maybe that's the primal state and then you attain to egocentric perspective but there seems like there's another kind of more psychodynamic ego or objectification of self the third person of taking an explicit perspective it's actually first person and that you're perceiving it from a point of view but it's fictitious and you're taking the view on yourself as if from without and that might come later in development at some point maybe through enactments of mirroring between infant and caregivers or between conspecifics but if you're mirroring with others then you might be able to establish these cross-mappings between what you do and what someone else does and this could potentially help you to create representations that you can then harness for viewing you as if from the outside but when we're talking about when we say self, when we say ego what kind of ego, what kind of self but it seems like there's a primordial minimal self egocentricity of just seeing from behind your eyes but then there's more like a psychodynamic ego slash me which is involving the objectification of you adopting a perspective on you imaginatively as if you were viewing yourself from without and then contextualizing yourself across space and time and in different contexts different social unfoldings as what you would need for something like an autobiographical self something of that nature so how does an ant see it though does an ant have like is the first person gone or like does an ant entirely meta all the time or a bacterium is that like a total meta perspective to go through this like ridiculous like self whatever creating and dissolving loop or Dan would know better than I would you actually have to ask the ants that's kind of the hard part but can I make one comment the difference between the egocentric and then the more impartial or autobiographical reminded me a lot of the projective consciousness model projective geometry model which was also integrated with higher order theories and global workspace and a lot of those other topics with Ryan Smith and Christopher White and also the more geometrical work with Rudroff and others because it's like we do need both those modes we need the one where the hands are bigger when they're closer to your face and we need the one where it's like you're playing the video game with the third person view and then it was really interesting when you point to a lot of the seen as deficits of memory or cognition like false memory but just seeing them as computational outcomes of stochastic interceptive lossy compression and generalization semantic encoding what's more likely who kind of person are you you know scale whether you think you did something or not so that's just such an interesting area that you kind of took it down thank you it's interesting bringing up like Rudroff because there's some things like I'd wonder like you know when do you first get the capacity for projected geometric modeling of what kinds and so like this so with and what's the nature of the projected geometric model like is this something that I should I stop I can stop the sharing there we go is this something that you would expect at the level of like cortical micro circuitry like some interpretation of like what cortical macro columns are doing or is this something that's actually more complex and it's like a distributed algorithm like Hinton has ideas like glom that kind of get it something like this because what you have with this projected geometric model is the ability to flexibly move between egocentric and allocentric perspectives I can take your point of view or mine once I raise everything or boost it up into this I forget to hyperbolic or whatever the space is he puts it in from within there you can embed either want to move between them the thing I'm wondering is like the nature of these like I just mentioned this like ability to take fictitious points of view on you I'm wondering whether like the way in which we do is a lot more derived and a kind of skill of moving between things like the representations you might have acquired through mirroring with others and like to to take another perspective rather than it being something that that the visual system can just do of moving between of automatically translating from points of view within this higher space like what is the nature of that model like I've even wondered like for instance is perception actually always 2D and a kind of like holographic way it's always a 2D projection and when we talk about something like depth it's actually either some sort of like a sense of affordance or interceptive cross mapping or the likely transitions between 2D or even like 3D like we like we never like Mara said it was like 2.5 division like is that just the ability to make sure that what's on the screen that's 2D is always coherent or that you move between as if you can move around a thing but we actually never perceive anything in 3D in terms of our visual spatial awareness but to loop around to the ant I don't know but like you might distinguish for instance if you consider the perspective of the ant itself as generative model then that might be by default more allocentric especially because of it's like what it's specifically doing to the it doesn't even make sense to pick it's fundamentally part of the super organism the organism is it's not even super organism like like Daniel says like it is the unit that makes sense that's being selected is the shared enactment but then if you go maybe like the attractor dynamics of like the central complex it's like quasi-hippocampal quasi-thalamic thing that might be more egocentric and I don't think you're gonna get just totally agree with that I think it's a established fact that the compass heading or the polarization of light is processed within one nasmate head and then there are collective decisions like the sex ratio of the larva that they're gonna keep or the overall regulation of foraging that wouldn't be expected to have even a correlate necessarily in the head capsule of any single nasmate and so I think that's where the discussion comes like multi-scale systems which is kind of a fun topic because in multi-scale systems we see sometimes a clean mechanistic difference between the lateral and the kind of vertical or the nesting of systems so like communication we can see the interface where you have the lateral layer of individuals and then where you have the layers that are inside of the individual it's a little harder to disentangle it if they're all like kind of cross-wire but what are blue do you want to ask a question otherwise yeah I mean it just makes me think of how like are we fundamentally broken with our like self little centric loop right like that are we maybe like measuring forever the incorrect scale of stuff like should we maybe be more concerned about like the scale of all of us together or the scale of all of the parts that make up us instead we like kind of falsify and create like our ego and our relationship we like project a lot about how we think we should be within the world or how we think things are within us that like we don't actually know anything about I don't know just related to that I mean it's definitely beyond neuroscience but like I'd say yes but also like I both think this is one of the most valuable and fraught parts of psychedelic and discourse psychedelic meditation also where it's like if you get this myopic ego centric perspective that's going to be a real problem for multiple reasons it's even for you it's going to be a problem because you see things from one way you're going to get owned by yourself as a kind of like you'll get trapped in all sorts of local you see things in one way you really barely see them anyway but I do worry a little bit though about overly you know egos get themselves in trouble but like the ability to stitch together for instance you know there is like a unit over which you have potentially maximum purchase and then in terms of influence there's a sense in which like you're the steward of that because like the causal closure is densest and there's also a sense in which like let's say I want to coordinate with others including myself across time the ability to like have a narrative structure the ability that's going to be helpful for not like just jumping at whatever source of utility or expected for energy comes along and so like to basically avoid being functionally psychopathic like you kind of need the ego but you also need to have the ego not be too tight to avoid being functionally psychopathic so like not too tight not too loose otherwise the whole thing's wrecked that sounds like narrative stigmergy like agent environment feedback loops and also of the scripts research that we've talked about previously with all the russin and others where that's sort of the top down prior so it's related to what you said about the hard problem if or why or why not there would be awareness of experience at for example the bodily level versus the you know tripartite zoom chat level but at the very least instrumentally those top level scripts are the narratives that decide like are we going to put everybody you know through a tunnel where only this exact width is going to make it through or is that a phenotype that we have a lot of ambiguity around using the Bayesian prior term so it is interesting how and I just want to ask you whether you started with the more neuro side and found it broaching to these questions or whether you started maybe focused on other areas and then saw the brains specifics as increasingly important like there's so little separation for me between like the neuroscience obsession and the existential like I'm constantly doing I'm constantly in a state of existential freakout and neuroscience obsession that I don't know which led which I don't know who's the dog who's the tail and who's wagging whom it's just constant existential crisis and obsession so I don't know what would be advice to somebody who's curious about the field or they want to be working in an interesting area of neuroscience and embodiment and everything that you kind of mentioned today it seems like if someone wants to focus on psychedelics the best route is to cut your teeth or to like basically get a home outside of psychedelics establish a competency there and then try to bring that within the psychedelic domain that seems like the best path that most people who succeed in the space take and yes it's the standard fraught thing of academia where it's like you're trying to not be an inch wide and a mile deep and you're trying to not be a jack of all trades master of none and where the perverse incentives abound and good luck like some combination of like both paying your dues and not optimizing yourself into a brittle straight jacket where you can't see anything past the next grant yikes I feel you thanks for the answer so one small question from the chat and then if anyone else has any final questions so Adam thanks for your time and the great presentation and cool work and hanging out with us so Adam mentioned the DMT paper with respect to the cortical traveling waves the question was wonder if there's any other work on non alpha waves and then just about that and then just a little more generally why do these traveling waves matter or what matters if they are important or not so there's a sense in which like I drew like those swaths those like synchronous complexes but those are necessarily a core screening you zoom in on them and it's like so what's the ontological status it has a certain relational realism and that the extent that makes sense like draw that out it's like active inference in general with Markov blankets it's like you can it's relative to what other system that's engaging in what types of inference they'll just decide like the scope of relevance I mean there's additional like auto poetic self organizing elements of like the internal but so like if you look though at like a synchronous like an alpha like beta alpha you zoom in it's actually comprised of these traveling waves is more of a fine grain description although you zoom in on that that's actually core screening of population level activity and so it'll depend so there's going to be potentially if you're looking at traveling and cortical waves that might be useful so I focus on like the standing wave description as like a joint belief over whatever is in its scope but if you're talking about something like predictive processing within the model you might want to zoom in on the standing wave or the standing description for like the business end of passing on your predictions and kicking up your prediction errors but depends like it depends on this the within the multi-scale account what's of interest for the modeler one quick thing from this in the side handle the wave question whether there's any non alpha wave traveling wave papers I don't think so I think the one I think the termin one was focusing on alpha if I'm remembering I could be misremembering I think but there are four psychedelics Terry I think would be the person to look at for some really good general biophysics of traveling waves in the brain yeah I think would be the place to look quick aside on DMT like I mentioned like there might be endogenous DMT there's debate over how significant it is so some people say it's not significant at all some people say it's very significant so it's one place we know that it's produced is in the pineal body the seat of the soul and it's another place and there's some evidence but some people have question it that it might be almost akin to like a neurotransmitter system in its own right we just didn't know how to look at it before Jim O Borgen at University of Michigan if you're interested in that issue and that debate but it's possible though like one of the main endogenous ligands we don't know of the 5HG2A system and the main stimulus might not actually be serotonin but might be DMT so like under like let's say like a limit experience like so one like I mentioned like carbon dioxide can do it or maybe lactic acid like you're running for your life or for your lunch and it's super high levels of blood acidity and that might be activating these proto-psychedelic pathways but another one could be like potentially DMT mediated and things like sex birthing in your death experiences so like you almost bled out maybe it's time to relax your beliefs you're about to bond with another creature for years maybe this is part of how you do it whether bonding with a mate or your offspring very speckled these are early days interesting and also maybe think about brain and body dopamine and serotonin other neurotransmitters in the blood in the blood vessels the immune system and then neuropeptides the whole continuum of short acting neurotransmitters to slower acting phenomena and then we're thinking about it from a nested multi-scale systems perspective with this kind of more local faster and then slower broader spatial scales which just has a lot of analogy with everything from regional governance to the regulation of tissue so it's kind of an interesting idea how a lot comes together there it also made me think of how staring into someone else's eyes for like a prolonged period of time induces like violent disturbing hallucinations which is like reported in I don't know some paper I can't remember but I'll look it up and see if I can post it in the chat but it's interesting how like thinking through others minds or like looking through others eyes is so violently incorrect to us that it's perceived as like an altered perception I wonder if this is speaking to like the way in which we bootstrap like objectified selfhood and me and like but because like we do it so much this is like a core axis of imagination and perception that you can create like these weird feedback loops and I also wonder like how much it varies across people like like if you're like really far on the autism spectrum like or far in a schizophrenia spectrum like would you like in a moment like you look in someone's eyes and it's like it takes you two seconds to enter like the psychedelic space or might it take like you know you got to sit there for like an hour this test was like I think 10 minutes so I'll find it right now it's happening to me right now okay well Adam not your first active live stream not your last but thanks for joining and sharing this cool work we hope to see you around the lab and just in the area so always a pleasure and an honor talk to you later