Select text to annotate, Click play in YouTube to begin
00:00:00
my favorite paper of yours and maybe one of my favorite papers ever of all time is the computational boundary of a self um paper that you wrote in 2019. I started to try to formalize uh the way
00:00:13
that uh one could think about any possible uh cognitive system or any possible intelligence and when I see intelligence I mean William James this kind of definition where it's a competency to reach the same goal by
00:00:26
different means it's a very cybernetic definition it means there's some problem space there's some goal you're trying to reach in that problem space it becomes a navigation task and and it asks you to think about for any given system how
00:00:38
much competency does the system have in navigating its world to reach its goal uh despite all kinds of new things happening various barriers and and and and so on so I started thinking about what if you know if we can't if we can't
00:00:50
rely on origin story uh because I don't think that matters um if we can't rely on composition because I don't think having a brain that looks like ours makes any difference here what do cognitive systems have in common
00:01:03
right and so and so what I settled on is this notion of a cognitive light cone and the idea is this and this is borrowed from uh well it's sort of upside down actually but it's but it's borrowed from from the way they do um minkowski may have been the first to do
00:01:16
it but but these uh these kind of space-time diagrams and physics so what you do is you put you know time is on the vertical axis all of the dimensions of space are are on the on the horizontal axis and what you can Define
00:01:27
is you can Define this cognitive light cone uh as the spatial temporal size of the biggest goal that the system can pursue Michael Levin is a scientist at Tufts University his lab study is
00:01:40
anatomical and behavioral decision making at multiple scales of biological artificial and hybrid systems he works at the intersection of Developmental biology artificial life bioengineering synthetic morphology and cognitive
00:01:53
science if you got this far please hit that subscribe button and I hope you enjoy our conversation Mike thank you for coming on yeah thanks for having me happy to be here such a pleasure I mean I I first learned about your research a
00:02:05
couple of years ago you did a podcast your first of a few with Kurt J mungle that was my first exposure to your work um and I just thought it was absolutely incredible and uh I agree with him when
00:02:17
he said like your work already it's like deserving a Nobel Prize which you've already been able to accomplish and it's just let's just think of the implications of what your work could bear out in the future it's it's
00:02:28
remarkable well thank you thank you um and your lab you're working with is there over 40 people working in your lab oh at the moment there's probably around 32 30 okay 30 to 40 I know yeah it sort of varies probably I'm sure with time
00:02:41
and your work is at the intersection of all these interesting areas I think um to get us kicked off it'd be great for uh sort of the lay audience people who aren't familiar with their work if you give them a brief overview of of what
00:02:54
you're working on maybe your your career a little bit and then we can dive into uh the native gritty I have a ton of questions to ask you about the specifics of your work too but just give us a little high level grounding sure yeah
00:03:05
well uh our work is spread over a number of areas so so we have there are people in the lab who do very basic uh conceptual kind of uh stuff that's almost philosophy and then uh computer there's some computer science and
00:03:19
there's some bench work in developmental biophysics and uh there's some Behavioral Science and things like that and uh we have a we have a mix of of uh biologists physicists bioengineers
00:03:31
computer scientists and so on and there are you know there are there are lots of projects in in areas such as birth defects regeneration uh moving towards regenerative medicine cancer
00:03:42
AI uh uh unconventional Computing and unconventional cognition and things like that and all of it it sounds like a grab bag of 100 different things but it's actually all stemming from one fundamental question that I'm interested
00:03:56
in and that I've been interested in since I was a kid which is really this issue of uh embodied mind you know this this issue of how Minds can exist in the universe how they interact with their
00:04:08
bodies how Minds scale from the Primitive kinds of um metabolic and other competencies of single cells to the emergent mind of the of the body and then of the of the whole
00:04:21
organism in a behavioral sense and uh this the scaling embodiment and uh communication is is at the root of everything so it's at the root of Developmental biology it's it's a way that we think about uh the task of
00:04:34
regenerative medicine is to communicate with this collective intelligence of cells to get it to do various things um and then of course applications to Robotics and Ai and so on so that's kind of the the field of the group you must be having so much fun because
00:04:47
it's at the intersection of all these fascinating areas intersections on intersections you know um it's absolutely just mind-boggling and and I recommend for people who are interested in your work I'll link in the
00:04:59
description um some of your papers some of your uh some of your presentations that you've given I find um I I really love visuals to the company sort of the explanations and uh the sort of regular presentation
00:05:12
you've given in the last couple of years that has everything from the planaria to morphogenesis and explain those kinds of things it's is really super helpful so um that's that's really cool and just as
00:05:25
a shout out the visuals there are a lot of amazing visuals both in the Toxin and the papers and most of them come from one graphic artist named Jeremy gay and he works in Peregrine creative as his company and he is uh phenomenal and and
00:05:38
he helps bring all of our crazy ideas into a a visual form that other people can participate in yeah thank you for pointing that out yeah I was gonna mention him as well because uh those are some of the one of the slides in particular that I wanted to share in a
00:05:52
moment because um to be honest too in preparation for this interview I've only it's been about a dozen interviews so far but this was by far the most difficult interview to prepare for if I'm being honest because
00:06:03
it's extremely it can get extremely technical especially reading your research papers and I believe you have over a couple hundred research papers right um you've published quite a bit um you know Decades of work so and that
00:06:16
a lot of collaborators that you've worked with as well so will will become will be changing scales I think quite a bit during this conversation because you know I want to get um to the details but I think where it's
00:06:29
really interesting is the the biomechanics where they intersect with um say like information Theory and sort of uh jumping between levels and
00:06:41
different scales but they kind of uh say biomechanics is implementation of say it's say a lower or more fundamental loss perhaps but well I think that's something we'll sort of start to unpack
00:06:55
as we go through this and one of the papers of yours though that I think um my favorite paper of yours and maybe one of my favorite papers ever of all time is the computational boundary of a self
00:07:06
um paper that you wrote in 2019 and I think that would be a great place to sort of ground the conversation and because it's got a little bit of everything a lot of your work is referenced in there and we won't go to
00:07:18
the paper to the paper line by line but some of the major topics of conversation I think um will be pulled out of that paper and we can we can dive into the specifics so if you give me a second I can actually
00:07:30
just share my screen here yeah as I just mentioned this this paper of yours it tackles so much um in one could you possibly give us an overview of the paper and um a million questions to ask you in terms of
00:07:42
specifics but should you tell us what's um this paper is about yeah yeah so uh I've been thinking about these things for many many years but what what finally catalyzed me sitting
00:07:53
down and writing this up is that um in 2018 there was a a conference on diverse intelligence uh put on by the Templeton Foundation and we were we were challenged by our
00:08:07
program officer pranabdas to uh think about Frameworks where unconventional intelligences could all be thought about at the same time and so there were people there working on uh ants and
00:08:20
octopuses and uh you know chimpanzees and and things like this and I I uh like with with a number of things I sort of um tend to crank that knob all the way
00:08:32
all the way out and so and so I'm interested in every kind of possible cognitive system so we're talking uh cells tissues minimal matter uh collective intelligence of groups of
00:08:46
organisms uh all kinds of um uh artificial synthetic beings uh software AIS possible aliens like all of it right so so I want to know what what how how
00:08:58
we how we detect how we understand and how we relate to and maybe even create uh diverse and intelligence is what is the space of possible minds and I'm certainly not the first person to ask that question sure so so so we were
00:09:12
thinking about all this and and I started to try to formalize uh the way that uh one could think about any possible uh cognitive system or any possible intelligence and and just to
00:09:25
give it to give it a definition by by intelligence I don't mean the kind of uh second order metacognition that for example humans have where you can sort of think about what you know and so on I mean that's one kind but but I'm talking
00:09:38
about when I say intelligence I mean William James this kind of definition where it's a competency to reach the same goal by different means it's a very cybernetic definition it means there's some problem space there's some goal you're trying to reach in that problem
00:09:50
space it becomes a navigation task and and it asks you to think about for any given system how much competency does the system have in navigating its world to reach its goal uh despite all kinds
00:10:03
of new things happening various barriers and so on and so so this this is so I like that definition because it uh it it's it's independent of the the origin of the
00:10:15
system it doesn't matter if you're evolved or designed or some hybrid uh it doesn't maybe I I think that there are a lot of artificial distinctions that are made in in these various fields that are uh not facilitating progress they're
00:10:27
they're kind of blocking it actually so so I want I want things that are that are as unified as possible I really take this unification task seriously so so I started thinking about what if you know if we can't if we can't rely on Origins
00:10:40
story uh because I don't think that matters um if we can't rely on composition because I don't think having a brain that looks like ours makes any difference here uh what do cognitive systems have in common right what what
00:10:52
does it mean for something to have a code to to have some kind of cognitive uh system and um and so and so what I settled on is this notion of a cognitive light cone and the idea is this and this is borrowed from uh well it's sort of
00:11:05
upside down actually but it's but it's borrowed from from the way they do um minkowski may have been the first to do it but but these uh these kind of space FaceTime diagrams in physics so what you do is you put you know time is on the
00:11:16
vertical axis all of the dimensions of space are on the on the horizontal axis and what you can Define is you can Define this cognitive light cone uh as the spatial temporal size of the biggest
00:11:30
goal that the system can pursue and I don't mean the reach of its sensors and effectors right so so the James Webb Telescope has this incredible uh sensory array that reaches to the edges of of the of the observable universe maybe and
00:11:44
so on uh I that's not what I'm talking about uh I'm talking about I'm not not what you can sense and and where you can act I'm talking about the size of the goal state that you are okay the largest goal state that you are capable of
00:11:56
pursuing so this this frames um much like James and I think I think it was right and like wiener and colleagues before that uh it frames the tasks of of some type of cognition whatever it is on
00:12:10
that on that Spectrum as being fundamentally about pursuing goals of some type and so and so one can sort of start with very simple and so so here it's a tick but you can think about a bacterium or something and you know it
00:12:23
it has a little bit of uh yeah it has the the cognitive like one is quite small because it has a little memory going backwards and has a little bit of predictive power going forwards in time but what it really cares about is the
00:12:36
local concentration of whatever molecules it cares about right at that local level so it's very small the cognitive icon is very small and you can move up to something like a dog so so that has has a longer like going
00:12:47
backwards uh it has a little bit of predictive more predictive capacity going forwards and of course spatially it will know things like one of the goals is to is to keep Intruders out of its neighborhood the the the the
00:12:59
peripheral the perimeter of the house or whatever it is that you know that that's that's what it works on but so so that's great that's way bigger than the than the tick but um or the bacterium but what you're never going to get your dog to do as far
00:13:10
as we know is to care about what happens in the town three miles over a month from now it's it's just not going to happen they're not capable of this as far as we are and and then and then you have very large lichens like humans uh
00:13:23
that may really be one of the goals may be something to do with world peace and uh what's gonna happen you know some people are literally depressed because the sun's going to burn out and so they're working on uh these Technologies
00:13:35
to get us off the planet and and so on so so you could and and of course we all have limitations right so so even though the human lycone is is huge and it does have this interesting feature maybe the first one in this evolutionary chain
00:13:48
maybe not where some of those goals are guaranteed not attainable yeah so so if you're a goldfish and you have goals that reach let's say a half an hour forward that's absolutely attainable you're very likely to live that long and
00:14:00
so your goal most all of your goals are are attainable if you're a human it's very likely that you know that many of your goals are your your cognitive Litecoin is longer than your lifespan you know they're not attainable and that may or may not result in some
00:14:13
psychological pressures that are unique but but I don't know so so our cognitive like ones can be huge but but they're also Limited in that for example and this becomes important in some of our more more recent work uh if you think
00:14:26
about the capacity to I mean what what this is fundamentally about is is care and compassion and if if you think about uh your Your Capacity to care about the welfare of other beings we are quite
00:14:37
Limited in the in the linear range so so you know you might have a certain amount of amount of care about something that happens to some number of people but if it's then you know if you're told that well actually it's not uh it's not a thousand people it was actually ten
00:14:51
thousand people that it happened to you your amount of of Active Care is not going up tenfold we just can't muster that that level right but you could imagine and and you know you could imagine these sort of bodies like beings
00:15:03
with whose like own is large enough where they can actually they can literally care about uh um you know a massive amount of possible uh possible uh sentient beings so anyway so that so that's that's the point of this
00:15:16
cognitively cone and the idea behind all of this and there's a there's a follow-up paper uh from from last year that's kind of the um the the next version of this which is the tame paper the Tammy the technological approach to mind
00:15:30
everywhere which extends this and and really goes into this uh the the mechanics of what this what this is supposed to be what this is supposed to be is um a framework that moves these kind of
00:15:43
questions questions of uh cognition of sentience of uh of of um intelligence and so on from the area of philosophy where people have a lot of philosophical feelings and preconceptions about what things can do
00:15:56
and what things can't do and it really uh really stresses the idea that you you can't just have feelings about this stuff you have to make testable claims and so so every Claim about a system of
00:16:08
some level of of Competency you say this thing doesn't have this level of cognition or it does wherever you are on the Spectrum and it has to be a spectrum it's not binary that's the biggest thing that trips everybody up is trying to be
00:16:20
binary about these things but wherever you are in the spectrum that is a testable claim so somebody has picked that somebody has to pick out a problem space uh say what you think the goal is say what you think the system is capable of doing to reach that goal and then you
00:16:33
test it and then you see if what you've got facilitates Discovery experiment uh further research and so on so all of these so so I you know for me the goal of all of this is to uh
00:16:45
make make a testable um Frameworks that that Advance a research program they're not the not just just philosophy sure sure so fascinating and thank you that was a great a succinct summary of the um of
00:16:59
the paper and I'd love to um the bottom right corner here this um collection of compound intelligences and I think it's one of the one of the points you made uh
00:17:10
you made in a few podcasts is that all intelligence collective intelligence um we're a collection of cells and I think that's such a wonderful point because we do kind of get lost in the um in the philosophical debates let's say
00:17:24
in terms of thinking of it as a singular kind of pointed Focus but this especially this graphic in particular where you have um you know the the cell and then actually the components within the cell I believe those are the um the the pink
00:17:37
um diamonds there that mix up the organism the organism makes up the colony so you can see these things are are made up they're sort of nested but one thing I would love to get your um uh just the explication on a little bit
00:17:50
here is how I believe it's from the paper that most biological systems consist of multiple Nets itself and how is that different than say one say integrate itself so I
00:18:03
think you actually point out it's not one as implied by integrated information Theory could you tell us a little bit about that sure yeah um yeah so so this this uh collective intelligence thing is
00:18:14
is crucial because a lot of thought in uh in in philosophy uh so going all the way back to uh Renee Descartes and before that you know Descartes which I think he gets a bad rap for a lot of things I I like him actually but but um
00:18:28
yeah so if you use them in your slide sometimes yeah yeah he's great well right very some very helpful stuff there but but one of the things that I think you would see was wrong about is that he was he was really interested in the
00:18:41
pineal gland because he said there's only one of these in the brain right and so that makes sense as a kind of a a Nexus for like this this unified human human cognition and so on but but actually if he had access to uh some
00:18:55
some good microscopy he would have found out there's not one of anything if you look into this pineal gland which you see is a whole bunch of cells and when you look into those cells you see a whole bunch of molecular networks and down it goes there isn't really one of anything and so what we have to ask
00:19:08
ourselves is uh where does this large scale because we we certainly as as an organism you certainly have a light cone that does not belong to any of your pieces so you
00:19:20
have goals and preferences and hopes and dreams in uh spaces that do not belong to your individual cells and tissues and organs they work in in physiological space and transcriptional
00:19:33
space and and when you're an embryo they worked in morphogenetic space you you work in three-dimensional space and maybe linguistic space and maybe some other things but um so so certainly there's a kind of scale-up process by
00:19:45
which these competent subunits and that's one thing we have to understand is that we are made of very competent subunits we are not I I give have another talk sometimes it's called um why robots don't get cancer and right
00:19:58
and the reason the reason that our current Technologies don't ever get cancer is because they tend to be made of pretty passive parts so you hope that your robot is has some degree of of intelligent performance but the parts themselves never go off the reservation
00:20:11
and do something you know something uh something else and so but but but with us that's a definite risk because that's the uh that's the that's the price you pay for for for having those scaling
00:20:23
mechanisms you uh you Pro you you you become um susceptible to failure modes where where the individual Parts they they have agendas uh and and there are these these evolution is given us some
00:20:36
some Hardware which is very good at uh scaling up these these cognitive light bones and pivoting them into different spaces into different problem spaces but that has that has failure modes and I think I think it's it's really
00:20:48
um critical to ask that question how the properties of the compound intelligence and the goals right because there are novel goals that appear and so on how those relate to the properties of the parts what Dynamics facilitate the
00:21:02
scale up how do they go wrong because at the level of tissues for example that's that's cancer that's basically a disorder of this of this cognitive scaling it results in cancer um and and and all kinds of other other implications up and down up and down the
00:21:15
chain and so you know the last thing the last thing you mentioned was uh was about um IIT and uh really the the one thing I would say is this and that I I don't we
00:21:28
haven't really gotten into Consciousness at all which is fine I think it's really important to demarcate actual talk of actual Consciousness from from the stuff that we're talking about which is which is um cognition intelligence uh
00:21:39
performance computation and so on and uh uh I think with with respect to those kinds of things I think it's pretty clear that um the kinds of things that very that that
00:21:53
uh for the the kinds of uh ingredients and the kinds of Dynamics for for which brains are sort of famous that and and this is why people think this brains are associated with with some with intelligence and so on those those kinds
00:22:06
of processes go on in every tissue of your body right so so almost I I I'm not aware of um uh any uh real distinction uh that one could draw between the the
00:22:18
kinds of mechanisms when people talk about magnetic fields and they talk about Quantum events in microtubules and they talk about the electrical signaling and various parts uh you know different kinds of integrated um uh spiking all that all that kind of
00:22:31
stuff happens in other parts of the body and so I think we have to be really open to this idea that treating the rest of the body as Clockwork whereas we are uh kind of okay with with attributing various higher level cognitive
00:22:44
properties to brains I think that's a real mistake and I think we're leaving a lot on the table in terms of biomedicine and other things sure by by making that assumption yeah may I ask how did you come up with the name scale free
00:22:55
cognition um that possibly could have been so so that comes from from a paper that I did with Chris Fields uh I don't recall if he was the one that came up with that title or if I was I don't remember at this point he and I do so much together
00:23:10
I don't know but um but the reason it's good here is that we we as humans are very um uh kind of uh uh focused on a very much
00:23:24
like we are in the in the electromagnetic spectrum and so on you know we only see this like tiny little little uh you know bandwidth right uh we are the same way for intelligence and so
00:23:35
we are very good at recognizing kind of medium-sized objects moving at medium speeds in uh three-dimensional space and look there that's a crow and this is a this is an octopus and that's a that's an orangutan we can recognize
00:23:48
intelligence but but there are intelligences all around us at various scales that we just find very difficult to think of in the same way so the bacteria in your gut that are solving
00:24:01
physiological problems and navigating not just physical space but physiological space and very clever ways that's an intelligence but it's very small and literally it's very small and it's and it's hard for us to recognize
00:24:13
those uh the various organs that make up your body people will immediately say Well they're not they don't have a cognitive of um self I have a cognitive self and they're just part of me well yeah that's because that's because we
00:24:25
and they would say the same thing about themselves if they could talk right so this is a this is just basically a first person perspective from a cognitive system uh and we're very bad at recognizing the different scales going going in the opposite direction
00:24:38
um one could for example and this is this is also something that uh that that Chris and others and I are working on the paper is not out yet but this idea that one could look at a whole evolutionary lineage or in fact the
00:24:51
process of evolution itself as this giant distributed agent where every particular animal is a hypothesis that that agent has made about the outside world and like all of our ideas that are
00:25:04
being sort of turned over by active inference by this active inference process these hypotheses may be supported and they may survive and give rise to other such hypotheses or they may be unfit and die and you'll never hear from them again and so so that's
00:25:17
another you know that's those kind of ideas and I'm not saying it's it's right or wrong but sure but those those kind of ideas you can't even begin to have those kind of ideas and consider those hypotheses if you were tied into the
00:25:30
fact that all cognitive systems are roughly uh the size and scale of us and you know and they have to have the brains and they have to work on a time scale that's you know uh to to Rapid and that begins to look too weird and too
00:25:43
stretched out and well that can't be a mind you know but it's all relative it's it's it's all completely relative but we have this Vantage Point like we used to with when we thought we were the center of the universe and all that we have this this scale vantage point that just
00:25:56
um really obscures uh the reality of things cool yeah because I was wondering about the the name that how you named it there because it seemed like it striked me as more scale dependent cognition because it's reliant
00:26:09
on the space and time scope but perhaps scale free means I don't know it's like a different approach it's like a macro look at it versus yeah I see I see what you mean right so so so you're right in that in that this approach does
00:26:22
emphasize looking at Scales it doesn't but the point isn't to get rid of scale what I think what I though when I say free I mean there's sort of it's kind of a play on words there's two ways to think about this one is it's it's free
00:26:34
in the sense that you it's not tied to any one scale so you can sort of jump up and down and it still works right it's these these invariants so so the same thing is true and that we see this all we see this all the time it's I I think
00:26:46
physicists use it this way too it's it's something that's true on one scale and then you you turn some control knob whether it's um size or time or something else and exactly the same thing is true at this other scale right
00:26:58
that same Dynamic so that's one but but I also I also like this idea to me what free uh really specifies there is freedom in the sense of freedom for the system to do what it's going to do and freedom for you as a scientist to not be
00:27:11
locked in in one particular um lens of looking at this system that's how that's how I see things I see things as you know sort of on on unfettering our our ability to to recognize these
00:27:23
diverse intelligences gotcha cool and what we're going to get into morphogenesis and by electricity um in a minute but before we do that we do want to touch on two concepts that are in this paper
00:27:35
um some surprise and info taxes um and I was somewhat familiar with surprise before that the concept of surprise from um it's like information Theory if a taxes was a little bit was a little
00:27:48
bit newer to me but uh one of the things I didn't realize was this concept of likening house these agents they want to minimize uh their surprisal level which is a concept from like the free energy
00:28:01
principle well it's a bunch of other places but and you've of course published with um with Carl friston as well I'm talking to him next month which will be great I can't wait to talk to him about that stuff but um but I didn't
00:28:13
equate it to a stress so like stress and surprise should we think about those two as equivalent so yeah so it's good that you're having uh uh Carl on I'm certainly not going to try to butcher his uh his uh his information by trying
00:28:27
to recap it here but uh let me let me uh sneak up on it from from a from a slightly different uh perspective because I think I think I think stress is is very important uh one way one way to think about uh the
00:28:41
behavior of these Collective intelligences of cells and we're talking about let's let's think about um regeneration or development which is basically just regeneration from a single cell same thing what you have there is a kind of which which what what
00:28:54
the system is doing is it's navigating in morphe space it starts off in this it's amorphous space is this a virtual High dimensional space of all possible anatomical configurations that a particular genome could possibly reach
00:29:07
and so so you start off in one corner of this morphospace as a single as a single egg a single cell and then you sort of navigate to this to this other region which is an ensemble of states that we recognize as the correct Target
00:29:19
morphology for whatever creature that you are um that navigation process is has quite a bit of intelligence to it in the sense that uh not just this is key I'm going to take a quick quick detour here is
00:29:31
that yeah yeah so but one way people critique these ideas and say oh well all you're doing is is painting the word and Intelligence on any complex process right anything that happens you say it's intelligent because look how complicated
00:29:43
it is in a gap from here to there so that isn't it at all this is this is really critical in you you first of all you cannot estimate uh intelligence from observational data it has to be Interventional experiments
00:29:56
and what you're looking at is not [Music] um you're not looking at emergent complexity that's very easy there are many systems that start off with very uh kind of uh simple local rules and then
00:30:09
something very complex comes out the other end that's not intelligence intelligence is the ability to get where you're going to reach that goal uh despite novelty and perturbations despite its problem solving basically so this is so this is key so so okay so so
00:30:21
you're this you're this collective intelligence because you're a group of cells very early on you become a group of cells and you have to uh you have to you have to reach uh reach wherever it is that you're going okay so so so the so the
00:30:34
part of this that's intelligent is the part that uh when you deviate it from what it's trying to do it finds other ways of doing it so let's just uh let's just um uh use a simple example we got it we got a salamander and in that salamander there's a very particular
00:30:47
limb that forms with certain number of fingers and if you amputate that lip and you can amputate it anywhere from the shoulder to the to the fingertips uh the cells will immediately grow uh they will they will start to to to to replicate they will differentiate they'll do all
00:31:00
these different things and and then they stop this is the most amazing part is that they stop when do they stop they stop when a correct salamander limb has formed so the way to one one way to think about this the traditional way to think about this
00:31:13
is as feed forward emergence so so yes every cell has a bunch of local rules dictated by DNA and biochemistry and you just sort of roll them forward and whatever happens happens and and it just so happens that it's a that it's a salamander limb so so that's one way of
00:31:27
looking at it that's I I think that way of looking at it is is perform missing the boat in a profound way because it limits our understanding and our ability and regenerative medicine but anyway the um the the view that I uh recommend is
00:31:40
that you look at this as a uh kind of homeostatic capability of this collective intelligence what it what it has is it and and by homeostasis I mean the the thing that happens in your thermostat where there's this Loop where
00:31:53
it takes a measurement of something let's say temperature in the case of your thermostat it makes a decision relative to a memory a set point of what the right temperature should be and then it takes actions to get to to get closer to that set point right so you can you
00:32:06
can think of you can think of regulative development and regeneration and cancer suppression and metamorphosis in exactly the same way there is a an error minimization process that tries to work
00:32:20
to keep the system in the correct uh anatomical region of space and when it's deviated by injury by mutations by teratogens by whatever uh it tries to get back there as best as it can and I I you know I give lots of
00:32:32
talks about all the amazing examples of of what it can actually do uh in order to achieve that just all the weird weird new solutions that it comes up with to various problems so so in order to do that one so so how do we think about I
00:32:45
mean the key thing about that Loop that that control Loop and this is you know for engineers this isn't absolutely nothing new that that that control Loop the key is that it has a set point and
00:32:56
uh why why does it keep uh why why do these cells act why do they spend metabolic energy um remember they used to be independent organisms individual individual uh cells uh care about little tiny single cell
00:33:11
goals they care about metabolic State they care about local concentrations of things they don't care about these large goals of um uh organ shape and organ size and things like that so why do they keep trying to do this and one way to
00:33:24
think about this and so this is an idea that I sort of push forward in that tame paper and then since then we've done a lot of uh experimental work on it which is not published yet which you'll be seeing later this year and some some uh some good
00:33:37
um computational modeling of it some of it which is published as of as of a couple weeks ago so one way to think about this is that what keeps the system Trying to minimize error is stress now
00:33:49
stress is a really good word because because stress has been used by um Engineers when they talk about uh various materials so so they talk about stress meso mesoscale stress and
00:34:01
materials it's talked about by a psychologist that talked about by evolutionary ecologists in terms of ecosystem stress of course behavioral scientists in terms of single but but where people don't really talk about stress very much is at this middle level
00:34:14
which which we focus on which is cells and tissues and organs and I I think I think that that this this concept of stress is quite General scale free if you will and that the deal with stress is that what one of the things and
00:34:27
there's a couple of major things is just one of those one of the things that helps scale the cognitive light cone and move a set of a set of subunits from caring about from a pile of cells that
00:34:40
care about individual cell level things to a collective that cares about much bigger things like hey we don't have the right number of fingers currently right and no individual cell knows what a finger is or how many you're supposed to have but the collective does and and the
00:34:53
thing that keeps them the motivated to to kind of try to get back to where they're going is stress and so what you need for this in that system what you need for this uh is for every cell to
00:35:06
um to to to be able to be stressed by and to spread its stress to other neighbors in in response to conditions that are really um failures of alignment at a larger
00:35:19
scale so for example so just just to give you an example if if um if we were to take a uh an eye and stick it on the tail of a tadpole right and and sort of you know transplant it
00:35:30
you could have a situation where no individual cell is damaged nobody's injured nobody's poisoned there's no DNA damage there's no reason for any individual cell to to be stressed but the collective is is is is at the wrong
00:35:44
location an anatomical amorphous space there's a geometric problem and so or or for example after amputating the the limb you know after all the cells heal and there's you know the skinny ever everything heals no individual cells should be stressed
00:35:58
because everything is fine on the Single Cell level but there's a real mismatch between there's a Delta between the the pattern memory that the system has that's that set point and where the collective is now so so I think and
00:36:10
there's there's a lot more than than we have time to go and see here but but but that spreading of stress and the linking of individual cell stress response mechanisms which used to be all about my
00:36:22
you know migration and and and and various physiological responses used to be about single cell stresses things that could happen to an amoeba and stuff like that uh and they're now getting harness to much bigger states of Affairs and that's the beautiful thing about
00:36:35
these homeostatic Loops is that the cybernetic kind of aspect of it it's a it's a it's a it's a set of wild cards um measure something what do I want to measure well it could be anything from your the quality of your DNA to uh the
00:36:49
length of the tissue in which you find yourself to your position in the hierarchy of the tribe you know some sort of social thing any of that can be the thing you may measure and then there's a set point to which you compare and again that could be that could be almost anything and then there are some
00:37:01
actions you could take which again could be almost anything so what evolution does is it uses this basic uh loop this this basic concept of this Loop and then it just progressively harnesses the three sort of main parts of that Loop
00:37:13
and points them at different sizes different scales different types of things and and that that I think is a is a major aspect of the kind of multi-scale Competency architecture that makes this the whole thing so robust and
00:37:25
interesting that's incredible the um so for stress one thing I think I'm not sure if it was from this paper another paper of yours but and you you you sort of uh you
00:37:38
didn't mention it's an exported error signal yeah so I think as you put it in the paper um export of stress molecules which have no ownership metadata yeah um can you
00:37:50
explain that sorry I'm not a computer yeah that'd be helpful yeah yeah there's so so there's two there's two pieces that there's there's two kind of two big components to what I think of as what underlies the scaling of these
00:38:03
collective intelligence is the first the first is this is the spreading of stress so so imagine this is this is I'm gonna I'm gonna give you an example but but really this is about all kinds of contexts not just not just cells moving
00:38:15
around but imagine that there's a there are a bunch of cells and there's one cell at the bottom that that according to the correct pattern needs to be up at the top now the cell can sense that it's far away and it wants to minimize this
00:38:28
distance and so it's going to try to migrate but it can't because the other cells are sitting there and the other cells have no incentive to move and let it through because they're perfectly happy where they are their stress is very low
00:38:40
so so what you what you've got here is a system where kind of like a like a cooling magnet right where you have all these domains and as you cool the thing they try to find a globally acceptable uh um configuration where everybody
00:38:52
where the total amount of stress meaning geometric frustration right the alignment is is the or the free energy is another way to put it uh is is as optimized as possible but what will happen is that sometimes it's impossible
00:39:04
for everybody to be happy right you can you can set up situations where the geometric frustration is such that you always have some kind of um lack of uh alignment somewhere so so the other cells so basically the
00:39:17
temperature that system temperature not physically but the in terms of free energy the temperature of that system is low uh the other cells are happy they're not going anywhere the barrier to get them to move is is very high because if they move they because they become less
00:39:29
happy so so now you get a problem so so now why uh what what could what can Evolution do about this well one thing you can do is uh the molecule that and and of course the level of stress how do
00:39:42
we know when something is stressed out well there are markers so there are specific molecules that are used as a um as a physical uh implementation of stress so so so you know heat shock proteins and then other other markers
00:39:55
that are like this is my level of stress right and then there's other there's hormones and whatnot so so now one thing you can do with this is you could keep all the stress to yourself and then this is the only cell that knows he's in the wrong location and nobody wants to help
00:40:08
him and and that's it that's how the system stays or you could be a little bit leaky so if you're a little bit leaky and and you're letting some of those stress molecules propagate to your neighbors then something interesting
00:40:21
happens uh they become less happy now why the why are they less happy well they don't know because stress is stress the molecule itself that's that metadata I was talking about that that that's that molecule doesn't by itself tell you what the problem is it just says you're
00:40:34
there's something's wrong right and so what that means is that the cells in this vicinity uh the temperature starts to rise they get a little bit more plastic and this actually um uh there are a variety of uh uh
00:40:46
systems were actually where for example um that I just saw a beautiful talk uh a Week Ago by uh um somebody who studies uh in in drosophila like like literally cells that are in the wrong location they make
00:40:59
they make the cells around them physically more fluid they get they literally get a little more plastic and when they get a little more plastic they're more willing to let other things happen for example that cell is now going to be able to get through them and
00:41:12
again I realize this is this is a an example that's very kind of physical but but the same idea in all sorts of all sorts of spaces Vice what you're doing by spreading that stress specifically in
00:41:24
a way that looks like all of their all of the other uh agents stress is you are making your problems be their problems it's it's a way to it's a way to get alignment it's a way to get coordination
00:41:38
towards a common goal it's a way to uh uh uh get this kind of unification this kind of unity that's needed for a com for for a complex multi-component system to achieve larger goals so so that's that's one thing the
00:41:52
spreading of stress to to to get and and and notice notice something interesting it doesn't require altruism it creates altruism these other cells are not helping you out because because they want you to feel better they want to
00:42:05
feel better and and the best way for them to feel better is for you to stop stressing everybody out that's that's you know that's that's a real kind of a simple way to put it but that's actually the dynamic that happens is is it's not that that they're there to help you it's
00:42:18
that you you the system is set up in such a way that that lowering the overall stress requires them to to to to to uh to act in a way that lowers your stress now now the other business the other part of this is the is the other
00:42:31
metadata a piece of this um which is which is this one of the things we study in in uh this kind of uh scaling up is this notion of uh this this notion of
00:42:44
communication via Gap Junctions now well now what are Gap Junctions Gap Junctions are these these little proteins that form um like a like a like a submarine hatch kind of thing that in this like like a little um aqueous pour that can
00:42:57
open and close in the cell membrane and when two cells are next to each other they can dock those those uh those those Gap Junction can Dock and then what happens is that uh small molecules and these are very we're not talking proteins but like very small molecules
00:43:10
signaling the eye current and and so on can can go from one to the other now I've made I've made the um uh the claim before that that these Gap Junctions are kind of magic in terms of uh uh this this this cognitive scaling
00:43:24
and and here's here's one reason why and they're not the only thing the the not the only way to do this but I think Evolution exploits them heavily um imagine that you're a cell and uh you've
00:43:36
got there's a cell the cell a and cell b and they're sort of near neighbors and man something happens to selei uh and uh cell a has uh some sort of some internal uh uh uh um consequences of that let's
00:43:49
say calcium has gone up or some kind of you know some kind of uh memory Trace molecule of that happening uh and what it can do is it can send a signal over to cell b now now the typical way that these signals work is that there's a
00:44:01
secreted molecule of some sort it goes out from the cell surface it lands on the surface of cell b there's a receptor up there grabs the recent you know binds to the receptor and now cell b now here's the thing with that kind of
00:44:13
communication um it's very easy for cell b to know that that's that that information did not originate internally that came from outside and with information that comes from outside you have to be very
00:44:27
cautious with it because that could be a parasite trying to hack you it could be uh somebody else trying to you know uh so so basically once you know that it comes from the outside you are free to ignore it you are free to respond to it
00:44:39
you may you may um Let It alter your uh model of what's going on in the world in other words you may learn from it or you may not um you got all those options but one thing you know very clearly is which information is yours that's all the
00:44:52
stuff inside of you and what's coming from the outside now imagine what happens with gap Junctions now cell A and B are now linked by Gap Junctions um something happens to cell a uh that generates a memory Trace let's say a a
00:45:05
calcium spike a calcium train of some sort that propagates through the Gap Junction lands in cell b now what's cool about this is that that calcium Spike or whatever it was doesn't have any
00:45:19
metadata on it that says where it came from whatever that signal is it is what it is and so when cell b gets it uh it comes directly into its internal milia it appears inside right these Gap
00:45:31
Junctions basically sort of make a connection that that um lets the two internal environments largely communicate and what that does is provide a memory wipe in the following an identity wipe in the following for
00:45:44
memory in the following way the memories that you now receive through those Gap Junctions the traces of various events you don't know as cell b you don't know that they're not yours there's nothing the the you know they're just like the
00:45:57
rest of the memories you have so so what it does is it's a kind of mind meld it means that it's now very difficult for that system to say whose memory belongs to whom was that was the was I the one that got poked and now I have this
00:46:09
calcium signal or not well you can't even you can't even do that calculus because you are so connected to be the ASL A and B are so connected they're now one system with respect to the memories they have so now this is a great scaling
00:46:21
because it means that you can now uh you now have a larger computational uh capacity because because now there's more and more networks and more states you have a bigger spatial area that you're sensitive to you can take bigger
00:46:34
actions you can start to exert biomechanical forces as a cell sheet so you can do stresses and tensions that are hard for individual cells to to have on that scale and you can so so so that that mind meld is a is another way to
00:46:47
scale to a higher level uh system because as long as the individuals can maintain I'm me and you're you then there's capacity for cheating and for um defections in the game theory sense and
00:46:59
and all kinds of calculus about well well what's good for me versus what's good for the system once you have this kind of mind meld all of that largely goes away you you can't defect against the cell that's gap Junction to you but
00:47:11
you can't poison it because it'll come back to you immediately it's it you know it it it scales up um cooperation it scales up a computational ability and it scales up
00:47:22
um the the goals that the system is able to pursue now uh I'll just stop after one one more point which is that I've painted a really Rosy picture of this Gap junctional communication I mean it prevents cancer it's I mean it's it's
00:47:35
great for many things but one thing that people often ask and I want to come right out and sort of uh say this is people often say well that's tremendous so so in a scale freeway let's all let's
00:47:47
all Borg ourselves together into this giant uh Collective and we'll wipe our individual identities and that will be you know just fantastic right uh and and I want to be very clear that that is not what I'm saying I actually think that
00:47:59
works out very poorly um and uh what's what what we need to keep an eye on here is uh for whose benefit do such connections
00:48:12
exist because you will then get an individual that a larger individual who who yes will do more impressive things in other spaces but you know as each of us knows who you know if you've done any kind of uh uh you know contact sports or
00:48:26
anything you're going rock climbing you scrape a whole bunch of skin off of your hands I had a great day rock climbing well you're here this your skin cells didn't they had a terrible day they're dead now and in fact and in fact right and so so that happens all the time we
00:48:38
don't think at all but typically about about these little little lives that makes us up so make us up so that's that's what happens when when you when you join a collective so it's you got to be careful it's not it's not a Panacea
00:48:51
by any means sure wow there's so many fascinating different places we'd go just based off what you said the um we want to go to next here I think one of the things that would be cool and
00:49:03
something that you highlight a little bit I don't think I think it's something that's um underappreciated in biology is just the importance of the membranes so
00:49:15
um I was I was watching a um a talk on your uh your Michael Evans academic content YouTube channel that's yours right I just want to make sure sometimes it's hard to tell you know sometimes people can start up but and it's someone else's account but I just want to make
00:49:28
sure that's yours but I watched um the very first video there by Bob gattenby oh yeah and um he stated in his talk and this really stuck out to me that about 99 of the
00:49:41
Shannon information in a cell is in the membrane and transmembrane gradient which um seems like a lot you know for someone who is a light completely a person in
00:49:52
this area can you give us a sense of why there's such this this disparity because I mean within the cell there are all these other sub cell components and they have their own membranes why is it that so much of the Shannon information is on
00:50:06
the outer boundary of the cell yeah well a couple things um Bob is amazing you should you should have them on uh yeah very very smart guy lots of interesting things to say
00:50:18
um I I think that uh and I don't I don't recall the you know the exact calculations that he used to reach that conclusion I can so I can sort of reconstruct why uh one one thing about information is that uh Shannon or or
00:50:32
otherwise is that what's very Central is the role of an observer uh and we're back to this concept of surprise this idea that as an observer how much surprise did you get out of a particular message that you received right or how
00:50:44
predictable was it that those kinds of things and so and so in the I I could I could imagine some kind of an observer in the cell that would uh get way more information from the things that are
00:50:57
going inside versus the membrane but I think what Bob was talking about is the fact that when cells interact with each other so this kind of multicellular context there's an interface by which they uh
00:51:10
communicate and control each other so cells are hacking each other all the time and that interface is the cell membrane and not just because physically that's what's on the outside but it's because in that cell membrane are a lot
00:51:22
of the interesting so so biochemical of course uh bioelectrical because that's where all the ion channels are and uh uh biomechanical because that's how you're gonna you can exert stress um
00:51:34
forces on things uh that's where all the signals come come in and out so so I think what he's saying is that that's the Nexus of the information exchange between cells and the outside world and
00:51:46
every cell is some other cells in in the body every cell is some other cell's external environment which is also kind of a key thing to think about when you think about the origin of cells and so on is that you need to you need to set a boundary somewhere it's not given to you
00:51:59
from from scratch so I think that's what he's talking about that that that com that control and communication interface of cells got it okay yeah that was um that just stuck out to me that
00:52:13
I mean I understand the importance and that will sort of get to that the um the the ion flows and dab Junctions and a little bit and the importance of of of those one thing that I think and
00:52:26
of course there's issues with Shannon information doesn't tell you about the meaning right there's certain issues there but one thing that did stick out to me about this paper I believe is in this paper that you said
00:52:40
um let me think about these light cones that expanding The Horizon is what enables Shannon information to acquire meaning because data data becomes causally linked to distant and past experiences
00:52:52
and acquires implications for future expectations so can you help like decouple that a little bit yeah yeah and that sounds like I'm almost positive that one was from uh that that's a Chris Fields sentence in that one I'm almost
00:53:06
positive and that one was from a paper that Chris and I did on on meaning and where meaning comes from and I you know the the and you should have him on too Chris Chris is also um absolutely remarkable so yeah I've come
00:53:18
across his research before yeah he's great yeah you should you should definitely talk to him but um uh the what I think what I think is is critical there uh and what uh what what
00:53:28
Chris and I uh have been working on a lot is uh the centrality of this concept of the Observer and it's this idea that and and I actually think that's one of the things Shannon information gets it
00:53:41
gets right and that it doesn't assume that meaning is uh I mean yes it misses out on meaning but but when you're trying to capture meaning you you start to get this idea well there is one meaning and it's our job to kind of
00:53:54
Discover it and I I think a better frame is that there are multiple observers which have to imbue uh the data with with meaning it's on them to interpret it and it means different things to different observers everything is
00:54:07
Observer relative and in fact I don't know if you've seen if you've seen this paper but there's a really cool recent uh well there's a really cool piece of research by Josh bongard one of my collaborators uh as a student atusa on
00:54:22
uh uh poly computation and then and then Josh and I wrote a review of and and some new ideas about this this concept of poly Computing and this idea that one way to squeeze more computational
00:54:35
power out of a piece of uh biology or in fact out of anything is not to change the thing but to change your perspective as an observer in the sense that multiple observers can be looking at the
00:54:47
exact same physical process and and get different computational results out of it depending on what their perspective is and what their interpretation is so it's this and it's got this it's it's an idea that has this weird match to you know some ideas in in
00:55:00
um uh kind of for psychology and so on that the changing your perspective on things is is a tool that you have in addition to actually trying to change the thing the thing itself um and so and so that's that's part of it
00:55:13
is that we have that we have to we have to think about that meaning is very much relative to some to some Observer which by the way might be the system itself an observer doesn't mean that there's some sort of there's not always some sort of external thing that needs is there to
00:55:26
give to give something meaning the system itself can be sort of it's like this strange loop as hofstadter puts it um that can do that for itself if it's a complex enough system yeah yeah
00:55:38
really interesting the um and just so you know after the fact I probably won't have this slide up the whole time I'll probably have the side by side of us just because uh it's probably up for you now but uh just so you know I'll probably put it up and put it uh have it
00:55:51
down and put it back up at different different times of the conversation but one thing before because I do want to talk about morphogenesis a little bit going to buy electricity but this light cone idea and I think I think this brings in something you've said
00:56:03
about cancer cells or at least your theory of how cancer cells look they have sort of the wrong scope of what what they themselves are um can you explain that a little bit sure well one one so so a lot of cancer
00:56:16
research uh focuses on why why do we get cancer right like like what causes it and I think that um the right way or a better way to think about the the kind of the origin of it
00:56:29
is why is there ever anything but cancer why isn't it all Cancer all the time because if you think about if you think about individual cells as as which are our as our ancestors their basic state
00:56:44
is is not to sit quietly as a disposable organ somewhere their basic state is to reproduce and go where life is good we are made of pieces that want to behave like a tumor they want to uh go wherever
00:56:57
they want they want to reproduce as much as they possibly can they want to hijack other cells around them to help them do whatever they need to do this is that's the basic state the real mystery is is is is what what keeps that uh under
00:57:09
control and I think this was this was well recognized by people kind of in the at the dawn of um the Developmental and cell biology but it's sort of it sort of lost uh lost Steam for a while but it's but but but it's a very powerful idea because then
00:57:22
you ask yourself well uh if we really take this this cognitive glycone thing seriously we look at individual cells and we say so what's happening here is that there are a variety of signals and those may involve
00:57:35
the stuff that we've been talking about so gab Junction stress propagation you know with all all kinds of other things um what they're doing is they're partially erasing the identity of these individual cells to enable the
00:57:47
collective to work towards a very large goal like making a nice liver or a kidney or whatever and then you start to think through well what would a failure mode look like so you get some you get some sort of oncogene that uh let's say
00:58:01
um uh shuts down all the Gap junctional communication for that cell and immediately what happens is once that cell is electrically isolated from its neighbors it can now start to have an individual identity and that leads to a
00:58:14
calculus of individual little tiny cell goals and and that cell can now as far as that cell is concerned now the rest of the body is just external environment so what happens with these electrical networks and of course chemical and
00:58:27
biomechanical networks is that they set the boundary between a cell and the outside world if if you're a cell and you're connected to a whole bunch of other cells the boundary of yourself is huge it might be the whole body you know and then everything else is is sort of
00:58:40
external environment but as soon as that Gap Junction breaks your your yourself is now tiny it's just it's just a single cell so so that boundary between self and world uh grows and shrinks it grows during during development and grows
00:58:52
during evolutionary um uh you know evolutionary the course of evolution and then it shrinks during during cancer where and that's really important to to to think about that these these cancer cells are not any more selfish than
00:59:06
normal cells you know in a lot of game theory you treat the cells it's more selfish they're not more selfish they're exactly as selfish as every cell they just have a tiny little self so all the other cells are also being incredibly selfish but their self is huge and so they're all working towards this common
00:59:19
purpose that we see of us as an organ or in fact the whole organism um but it's not because they're not selfish it's because their par their identities partially erase through their tight information linkage to other cells
00:59:30
and so cancer is exactly what you would expect from a breakdown of that process as soon as you're disconnected and we see this we see as soon as you get an uncle you get a strong oncogene the first thing that happens is it shuts down um uh Gap junctional communication you
00:59:43
get a weird voltage uh that's a depolarized voltage state and then and then you start doing this calculus well where am I better off well I'm going I'm going to go down there and then I'm going to proliferate as much as I want and this is metastasis and and
00:59:55
then and then yeah then your goals start to start to deviate from the goals of the rest of the collective yeah yeah I love that you said you pointed out the uh cancer cells no more selfish than another cell just has a the
01:00:07
wrong scope of what it is kind of it doesn't have the proper um that it's light cone is is deformed let's say but yeah and and it's it's shrunk and the key is again so so so to
01:00:20
make good on this on this promise at the beginning that this wasn't going to be just just philosophy what we've been able to do is to take that weird way of thinking about it and turn it into a therapeutic strategy so so once you start thinking about that we think okay
01:00:32
so maybe we don't need to kill the cells and maybe the problem isn't the the the molecular Hardware maybe what we can do is is physiologically reconnect it to the network and it should go back to doing its normal thing and that's in
01:00:45
fact what we've done so so we have we have a whole body of work on cancer in the Frog model and we're now moving to uh glioblastoma in human cells uh is to is to convert the cell with a with a really
01:00:58
strong nasty oncogene like a k-ras mutation force it to be in proper electrical state with its neighbors and the cell normal and they normalize there's no tumor they go even though the oncogene is blazingly strongly expressed they
01:01:11
just go back to doing what they what they normally do because they're now again they're part of this they're part of this morphogenetic Collective will we eventually you think be able to electrically communicate with say a
01:01:23
cancer cell and just change the information going into the cell via say electrical currents and tell it to do something else like is that a decent explanation of where we're going towards or is it a little different than that
01:01:36
it's it's it's it's it's reasonable um but I'll twist that just a little bit here's yeah here's where here's where I think it's going um down at the lower level so if you zoom
01:01:49
in yes I think that's what we're looking at is is aberrant electrical and other kinds of signaling between cells that's the target however uh I don't think that uh the way that we're going to be
01:02:01
manipulating this is by focusing do our intervention efforts directly at that layer and the reason I don't think that's what we're going to do and I'll tell you what what I do think we're going to do uh is that it's kind of like
01:02:15
uh it's kind of like programming a computer the way the way you programmed a computer in the 40s and 50s is you had to physically interact with the hardware you literally had to sit there and rewire it right and that's and that's what all of modern molecular medicine is
01:02:28
about it's a crispr and genomic editing and and and rewiring Pathways and protein engineering like all that stuff is down at the level of the hardware I think what what can what the information technology revolution has
01:02:41
taught us is that if your Hardware is good enough and I think the biological Hardware is definitely good enough to do this the the better way to do this is to communicate with the system using its own built-in interface not to try to
01:02:52
rewire it but to but to um uh reset the set point and and and and take advantage of its built-in competencies so so reprogrammability so so what I think we're going to be doing is I don't think we're going to be
01:03:05
telling every cell what to do in either cancer or regenerative medicine context what I think we're going to be doing is uh basic basically of a rough way of saying is retraining it
01:03:17
um we're going to be we're going to be using high level stimuli and and I think I think it's uh AI recent recent examples of AI uh and various techniques in machine learning show great promise
01:03:29
and being that that translation layer to uh to to to do very high level top-down interventions where the details are all handled by the system itself this is how we've done a lot of our work on bioelectrics we can make cells we can
01:03:42
make um bodies form an eye in weird locations or an extra leg or or you know repair various structures and it's never by telling the individual cells what to do it's never by differentiating stem cells It's Always by communicating with
01:03:56
a collective intelligence with a fairly simple signal looks like a subroutine call it just says build an eye here or regrow whatever normally goes here that that's the kind of stuff you don't want to be down at the level and I say this
01:04:09
to my to my to my um students uh you don't you don't get out your soldering iron everyone every time you want to go from from Photoshop to Microsoft Word right I mean you could but but we don't do that why don't we do that because because your system has a much better
01:04:23
way where you can take advantage of these built-in multi-scale controls and that's what I think we're going to be doing oh that's great thank you for correcting that's that's really helpful to have
01:04:34
that um sure you know there's actually there's actually a really interesting piece of this that that ties into uh some some um uh some medical uh uh issues that are that have been coming up
01:04:46
for for years we just had a talk by by uh Fabrizio Benedetti who's this amazing Placebo researcher and he has this quote which I just love and I'm now using it all over the place which is um he said
01:04:59
uh drugs and words have the same mechanism of action and he's exactly right by looking at by looking at how bodies respond to uh the context of being given
01:05:12
a diff specific drug or being given a placebo whether the whether the doctor was wearing a white lab coat or wasn't or whether he told you you were getting a drug or whether he told you you were getting the best drug or what you know all of those things have have massive
01:05:24
implications for the therapeutic outcome and that's because we are part of this multi-scale system where high level cognitive expectations of what's going to happen filter down to the molecular details of how your cells actually react
01:05:37
and this is what makes um a hypnodermatology work is what makes the placebo effect work and it's also less people think that this is some like weird exotic effect that's how you get
01:05:49
up out of bed in the morning because when you're laying there and you decide to stand up out of bed what's happening is that high level executive cognitive uh function that says you know what I think I'm going to get up out of bed out
01:06:00
it has to tr has to Traverse down to change the molecular properties of your muscles of the cell membrane and your muscles in order for you to do to take these voluntary actions it that that cognitive intent has to change the way
01:06:14
your uh voltage is spread across the cell membrane it has to affect the micro level physiology so this business where where top level cognitive States affect molecular biology is not some some weird
01:06:28
you know mind brain medicine woo-woo stuff it's it's a it's the daily reality of being a behaving um a behaving living multi-scale agent that uses uh bioelectricity mostly but
01:06:41
but also a bunch of other stuff to cross levels like that to go from from cognitive goals to molecular States oh that's fascinating I never thought about it that way huh yeah next time tomorrow morning when I wake up I'll uh
01:06:53
think about the I gotta be the executive and actually get up and not snooze a bunch of times right the uh one thing that I want like and you actually just touched on and it's another slide that um I guess the the elevated
01:07:06
um function of bioelectricity versus these these other components that are happening um within the cell um I think it's it's a great I think it's a great like sort of one slider in terms of like looking at more morphogenetics
01:07:19
and um kind of breaking out the component parts and there might they're very very well might be other sort of components here but these are kind of the major ones right to get the ECM I think it's extracellular Matrix is that right
01:07:32
attention can you give us just like um I guess a quick um in a rundown of sort of what each of these kind of component parts do and then like I said like and you have mentioned here of course on the right
01:07:45
here the bioelectricity has its own it's the computational layer so it's got sort of a special component yeah this is this is a very early slide I I used it uh extensively probably 10 years ago I'm
01:07:57
sorry I said well no it's fine it's still it's still valid but but see as as I was rolling out to this whole this whole uh kind of research program starting in I don't know gotten 98 or 99
01:08:09
or so uh I I there were certain things you couldn't say at the beginning that you have to you have to say things a certain way and it's much easier now now that the data are there it's it's easier to to say those things now what would
01:08:22
this thing this this um this diagram is um basically just designed to get across a couple things uh first of all that there is this morphogenetic field that impinges on every cell and it's made of
01:08:35
multiple components so there's the biochemical component or layer and then there's the biophysical and so so ECM and tension are both kind of part of the biomechanical layer they're the stresses tensions pressures the you know those
01:08:47
kinds of gradients um and then there's the bioelectrical layer so so that was that was the first thing I was trying to get across and that this this information field is critical from development to to adulthood to cancer suppression to
01:08:59
Regeneration to to all of that the other thing I was trying to say is that bioelectricity doesn't do the whole job by itself because I've written many papers saying how amazing bioelectricity is and then and then people will often
01:09:12
read that as me saying that well that's the only thing that matters and I'm absolutely not saying that of course all these other things matter and they all work together um and then and then the the last thing which which I only started emphasizing
01:09:24
fairly recently is that what what's it what's really interesting and again this isn't something I would have ever said early on but but but now I think I now I think I think it can be it can be said is that what's really
01:09:38
interesting about bioelectricity isn't um that it's yet another piece of physics that we now have to pay attention to it's more than that it's actually the uh the excitable medium the
01:09:50
the the uh medium in which the the information processing and the Primitive cognition of that collective intelligence is stored so again that sounds very very weird and and sort of mystical but it's basically exactly what happens in
01:10:05
Neuroscience so Neuroscience if you ask the question what's the cognitive glue that takes a whole bunch of individual cells I mean neurons are perfectly good individual cells and then there's other cells in the brain of course uh what's the cognitive glue that binds all of
01:10:19
those cells together so that I emerge and I have some thoughts and and and uh and plans that don't have anything to do you know that they don't exist at the level of e cells well the answer that we've now known for for a really long time is Ah bioelectricity that's what it
01:10:32
is and so all I'm saying is yeah that's right and where did the brain learn that trick it got it from from all the evolutionary history before we had brains and neurons even of using the
01:10:43
exact same thing to store the content of the thoughts that the collective intelligence of the cells was having about anatomy in morphe space and then eventually Evolution sort of pivoted that to be all about three-dimensional movement and things like that so so
01:10:58
that's what that's what that slide's about okay gotcha yeah sorry it's a if it's a dated slide but uh that's an important concept yeah it's got a lot it's a very dense it's got a I think of
01:11:09
um on different components to it and I want to get to so on the bioelectric um level there was I think it was some in the 30s or 40s I'm trying to remember
01:11:21
Harold Saxton Burr yeah who who had a lot of these ideas a hundred years ago um and he was sort of on the right path what um what kind of privilege information do
01:11:35
they have like what do you know anything about him or uh what his deal was yeah yeah that's a great question and I apologize I've just got about a few more a few more minutes um yeah yeah so so the story with Burr
01:11:49
is interesting well first I'll just say I'll just say this this whole idea of privileged information there's there's a few examples and Burr is one of my favorite examples but but there are others like um I'll let this guy this guy Lama tree who was uh around in the
01:12:02
1700s who published this book that said man was a machine and he just imagined you know what machines did he have access to back then you know water clocks maybe was the best machine you could have seen and just to have that that foresight to to extrapolate from
01:12:16
that and to be able to say yeah and so all these great artists and philosophers what they do yeah that could probably be explained that way right just like this unbelievable um ability to to out paint from what you
01:12:28
see into what what the future might be so so Harold Burr had this and and the thing with um uh he wrote he wrote this this book all the the fields of life and he uh he had one of the first good uh
01:12:40
voltmeters and he went around and uh he measured things everything from cells tissues embryos tumors uh maple trees psychiatric patients everything you get his hands on he would do would take
01:12:53
these measurements he didn't have any ability to do functional experiments they didn't really understand at the time what the uh kind of biophysics was that was underlying it at the cellular level they didn't know anything about ion channels uh but on the basis of
01:13:06
these measurements he described what how he saw the fear the the the future you know kind of development of this area of bioelectricity and I gotta say uh he he had a crystal ball I mean it was
01:13:19
unbelievable it's just how that simple with that one tool that he had and nothing but uh nothing but some some crude uh you know extra cellular measurements he reconstructed and he was able to see I mean everything that he
01:13:32
said in that book has been basically slowly uh being being realized to be true uh and then there's some more stuff that you know we'll find we'll we'll see um how it goes but um that that book was was hugely influential uh for me
01:13:46
um I think I think he was amazing I think he was he was a Pioneer to be able to to see all this stuff and uh of course of course the other part of this is you know I sometimes get emails from people who say what are you doing all this work on electricity this was all
01:13:59
done in the in the in the 40s it's all done well okay there's a difference between kind of uh roughly having the idea that something like that should be possible and actually getting it to the point where it you know it's
01:14:11
quantitative and it works and it's a therapeutic so so I'm not saying that I mean all the amazing people that have worked since him which were also you know kind of my hero so so you know Lionel Jaffe and and all of his uh
01:14:22
postdocs um You Know Rich Morgans and Ken Robinson and and all those folks uh uh were were amazing and and they all did valuable work and now we're sort of able to to go from there but uh but yeah
01:14:34
he had these he had these very fundamental ideas uh in the 30s and 40s it's really remarkable it's incredible you have to do more digging on his uh on his work and then and dive into a bit
01:14:46
more and one final thing just before you go the one of the final paragraphs of the scale free cognition paper you you write about Zen Zen Buddhism uh which I thought was great and
01:14:59
brought me back to reading by hofstadter which he has some is that in there um anything you want to sort of leave us on on on that point I think um I can go back to the light cones here
01:15:12
because I think that's it's it gives you a good visual for what you're kind of getting at there yeah the only thing I'll say about that and and I'm no I'm no kind of Zen expert or anything but but I am fortunate enough to be working
01:15:24
with a set of um scientists interested in uh certain ideas in the Buddhist tradition and we have a paper uh out now and we've got a couple of other papers in the works on the
01:15:38
I'm kind of conciliency of this cognitive Litecoin idea with the idea of enlarging one's ability to feel compassion and to active compassion I don't mean feeling emotions about things
01:15:51
I mean actively you know working towards the welfare of all beings and so so to me this is this is really fine there's some good there's some good diagrams in there that I think you like um there's some there's some really fundamentally
01:16:02
important aspects here because uh again one thing that thinking about this diverse intelligence does for you is that it really makes it clear how the kind of simplistic distinctions that we've been making about
01:16:17
who or what we should feel compassions towards is uh it really needs a revising and um you know at the end of what many of my talks now I do a whole thing on uh kind of the space of possible beings and
01:16:30
all the different Weird Hybrids and hybrids and cyborgs and everything else that's going to exist and what it's doing is it's completely wrecking the kind of primitive um uh criteria that we used to use on
01:16:42
which to develop our ethics and that that all has to get redone and I think that the goal of all of this stuff about enlarging your life going by you know technologically increasing our intelligence and our capability and I'm
01:16:54
all for it I think I think that's we've got to do that but the purpose of all of that isn't just to to screw around and have more technology the purpose of it is to uh enlarge our our ability to feel compassion towards other beings and and
01:17:07
uh yeah you know we don't have time for all that but but there is a there is a whole paper on it and there's more there's more coming oh really what's the paper called um I will uh you can here
01:17:20
I just left a link to it and oh I want to read it myself of course too yeah and the tame paper as well yeah so it's called so it's called Uh biology Buddhism and AI care as the driver of intelligence and it's this idea that
01:17:33
care and compassion and the intelligence are mutually uh reinforcing feedback loop so that as you as one grows the other should grow too and that's that's um yeah anyway so so I'm gonna I'm I'm putting it on the uh on the chat here so
01:17:46
you can uh oh thank you yeah there's a bunch yeah so this is done this is done with uh Thomas Dr Olaf Witkowski Elizabeth and Bill Dwayne and there's there's a couple of more um you know they're the Buddhist experts not me so there's a but there's a lot more on this
01:17:59
coming that's right in my Lane my audience is gonna love that um I can't wait to dive into that paper and yeah well you can thank you you know you could have them all you could have them on at some point and just talk about that specifically like that if folks
01:18:11
care about that here then uh that's true yeah that's a real real interesting discussion to be had yeah I'll try to and everything that we discussed as well well these let this page specifically some of my favorite papers of yours your
01:18:24
talks I'll link in the description of this video for the audience watching is there anything else you want to um to tell people where they can find you or uh next yeah the only absolutely
01:18:36
everything we do is on my academic website which is www.drmike11.org uh one word Dr Mike 11.org and everything is there all the papers the talks the inspiration everything everything yeah wait we
01:18:51
didn't get a chance I wish I really wanted to talk to you because I know you're a big mid-journey fan yeah um and you have some stuff on there we are over our I think um styles or interests in art overlap quite a bit
01:19:03
like the the Codex surface it's hard to pronounce Hieronymus Bosch's work so this is surrealist I think we have like a lot of and I think you had something a post about Dwarf Fortress which is like a
01:19:16
game simulator and maybe one of your kids was was getting into it which I also had gotten down the rabbit hole about um like last week separately completely and also my my computer almost froze a
01:19:27
few days ago because I had so many tabs open going off rabbit holes of your work and your research um there's so much fascinating stuff so what we can do we can do we can do another one if you want we can we can talk about that and AI you know we
01:19:39
didn't get to Ai and some of that stuff so yeah wow there's thank you I appreciate that so much yeah there's a lot of questions I kind of over tend to over prepare over I have a lot of other questions I had active inference so I talked about uh Kristen about that but
01:19:52
um there's a lot of other stuff and so I really appreciate that and appreciate your time Mike this was amazing I learned so much I'm sure the audience did too and I love your work and uh well I'll be in touch thank you thank you so much yeah I appreciate it talk to you
01:20:05
soon
End of transcript