WEBVTT

00:00:03.537 --> 00:00:10.497
This is the Convergent Science Network podcast. Leading researchers in the domain

00:00:10.497 --> 00:00:16.737
of neuroscience, brain theory and technology are interviewed by Paul Vershoor and Tony Prescott.

00:00:20.357 --> 00:00:24.537
This is Paul Vershoor with the Convergent Science Network podcast together with

00:00:24.537 --> 00:00:26.357
my colleague Tony Prescott here

00:00:26.357 --> 00:00:30.617
at the 2018 Barcelona Cognition Brain Technology Nausea Summer School.

00:00:31.077 --> 00:00:33.837
And we're here with Stuart Wilson. Welcome, Stuart. Thank you.

00:00:34.357 --> 00:00:38.977
And Stuart, you spoke this morning about self-organizing models of brain and behavior.

00:00:40.077 --> 00:00:45.577
So why do you think self-organization is such a useful concept to think about brain and behavior?

00:00:47.877 --> 00:00:55.117
So I think that self-organization is one half of how the natural system works.

00:00:55.577 --> 00:00:59.957
So I think that forms

00:00:59.957 --> 00:01:06.957
that are generated by natural systems and that evolve and become the things

00:01:06.957 --> 00:01:12.377
that surround us in the natural world do so as a combination of the intrinsic

00:01:12.377 --> 00:01:17.957
properties of self-organising systems and the forces of natural selection.

00:01:18.137 --> 00:01:22.557
And I think that we need to think about both of those things and how they interact

00:01:22.557 --> 00:01:26.177
to fully understand patterns

00:01:26.317 --> 00:01:29.137
that we see around us in the natural world so how

00:01:29.137 --> 00:01:32.517
would we define self-organization for practical purposes

00:01:32.517 --> 00:01:35.197
um so there are

00:01:35.197 --> 00:01:38.177
lots of definitions of self-organization from different fields

00:01:38.177 --> 00:01:41.437
uh from physics thermodynamics people talk

00:01:41.437 --> 00:01:44.917
about phase transitions and uh from the

00:01:44.917 --> 00:01:48.857
world of sort of complex systems um people

00:01:48.857 --> 00:01:51.757
talk about um sort of

00:01:51.757 --> 00:01:56.757
edge of chaos dynamics I define self-organization

00:01:56.757 --> 00:02:00.457
as where you

00:02:00.457 --> 00:02:08.137
have a system of individually simple components interacting in individually

00:02:08.137 --> 00:02:15.997
simple ways such that collectively they generate a pattern that is less simple

00:02:15.997 --> 00:02:18.237
than those individual interactions.

00:02:19.057 --> 00:02:25.517
Okay. And so you started your talk referring to this paper by Jonas and Corling

00:02:25.517 --> 00:02:29.757
about what a neuroscientist could understand the microprocessor. Yes.

00:02:32.186 --> 00:02:36.646
I always took it a bit more like, okay, that's funny, but we've been talking about that for decades.

00:02:37.786 --> 00:02:43.366
So why do you find that useful? It's okay, Tony.

00:02:46.566 --> 00:02:52.126
You're poisoning me with your coffee. Hey, you're up to me.

00:02:54.086 --> 00:03:01.826
So why did you feel NEMA was a good start? I don't agree with everything in

00:03:01.826 --> 00:03:06.006
that paper, but I think that it represents a really interesting kind of thought experiment.

00:03:06.586 --> 00:03:14.066
I think it's important for modelers in particular to think about the level at

00:03:14.066 --> 00:03:17.366
which we're modeling the systems that we're interested in.

00:03:18.386 --> 00:03:22.386
So for me,

00:03:22.526 --> 00:03:26.306
the right level of modeling for asking the kinds of questions that I'm interested

00:03:26.306 --> 00:03:31.366
in is to actually try and construct the simplest kind of model that can account

00:03:31.366 --> 00:03:36.686
for the complexities of the thing in the natural world that you're trying to explain.

00:03:37.746 --> 00:03:47.746
And I think that sometimes by pursuing models which are as complicated in their

00:03:47.746 --> 00:03:52.566
formulation as the system that you're trying to explain.

00:03:53.746 --> 00:04:00.366
I think that you can lose some level of understanding in constructing those kinds of models.

00:04:00.706 --> 00:04:06.946
I think we need a mixture of both, but the thing that I'd like to try and guide

00:04:06.946 --> 00:04:13.246
my thinking is to create models which are as simple as possible to explain complicated things.

00:04:14.166 --> 00:04:18.126
But the other message could be you say, okay, we want to do hypothesis testing.

00:04:18.206 --> 00:04:19.746
If you don't have a hypothesis, you're lost.

00:04:20.086 --> 00:04:23.286
Yeah. Right? Yeah, exactly. that which is for something that's

00:04:23.286 --> 00:04:26.066
not necessarily that new as an insight that's right um

00:04:26.066 --> 00:04:30.106
but and the other thing that's interesting is of course that you step into the

00:04:30.106 --> 00:04:35.226
self-organization uh boat if you want yep which of course uh when i was your

00:04:35.226 --> 00:04:39.386
age there was a big emerging thing right that yeah when artificial life comes

00:04:39.386 --> 00:04:45.246
off in the late 80s early 90s was a big hope ultimately people like Stuart Kaufman,

00:04:45.386 --> 00:04:47.686
being very vocal about that.

00:04:47.886 --> 00:04:53.906
And also in your work, at least in your talk, you confess to sort of have taken

00:04:53.906 --> 00:04:55.886
a lot of ideas from Kaufman.

00:04:56.066 --> 00:05:00.826
So do you really see a continuity of ideas from 80s till now,

00:05:01.006 --> 00:05:02.466
or do you think there were some transitions?

00:05:04.266 --> 00:05:10.466
I think there have been fashions in the way that people have thought about these ideas,

00:05:10.466 --> 00:05:13.326
is and a lot of that has been determined by computing power

00:05:13.326 --> 00:05:16.126
that's been available uh things have fallen in

00:05:16.126 --> 00:05:19.546
and out of fashion with neural networks and um

00:05:19.546 --> 00:05:24.826
and an artificial life at different times and so on um i think that what i've

00:05:24.826 --> 00:05:31.066
tried to do so i was i was quite young when when you were reading those when

00:05:31.066 --> 00:05:36.146
you're reading those books um and and the first time i've read them them, they really stuck.

00:05:36.366 --> 00:05:40.326
So Stuart Kauffman's description of how the natural world works.

00:05:42.081 --> 00:05:45.001
Self-organizes and then selection operates on that that has

00:05:45.001 --> 00:05:47.841
stuck with me as a kind of way of thinking

00:05:47.841 --> 00:05:50.561
about the world right the way through you know my

00:05:50.561 --> 00:05:55.081
my education and and now to the to the point where i'm able to to do some research

00:05:55.081 --> 00:06:01.401
on that stuff um and so i think i think i've been i've been not influenced by

00:06:01.401 --> 00:06:08.621
uh fashion so much as you know i've just been i've been fascinated by James

00:06:08.621 --> 00:06:09.961
Gleick's description of chaos,

00:06:10.221 --> 00:06:14.881
Stuart Kaepernick's description of self-organization, and it's just stayed with me.

00:06:15.561 --> 00:06:18.701
Well, for me at the time, it was more like a primogy, right,

00:06:18.861 --> 00:06:23.481
the origins, and a multiframe, applying it to the brain, right, dynamical system.

00:06:23.881 --> 00:06:27.881
And then, but you're on a real, I'm to respect to behavior, right?

00:06:27.961 --> 00:06:29.621
This is to that's more and more for me.

00:06:30.321 --> 00:06:34.361
Yeah, I'm not quite as old as Paul, but at least a week.

00:06:34.381 --> 00:06:41.521
You might not say it. But I do also remember being enthusiastic about these approaches.

00:06:41.801 --> 00:06:47.981
But I think looking back, we can say that they haven't had the impact in neuroscience

00:06:47.981 --> 00:06:49.441
that we expected them to.

00:06:51.181 --> 00:06:57.821
And I did a review chapter for the Living Machines book about this.

00:06:57.821 --> 00:07:05.321
And really, the number of papers that take this approach and use it in a serious

00:07:05.321 --> 00:07:10.581
way to try and understand brain evolution and development is really rather small.

00:07:11.241 --> 00:07:16.721
And the people that have been pioneering in that area have, you know,

00:07:16.721 --> 00:07:20.641
did some work and then moved away. You know, I think they find it very challenging.

00:07:21.801 --> 00:07:26.081
So, you know, what is the scope now for, I mean, why is it challenging and what

00:07:26.081 --> 00:07:30.181
is the scope for taking on the enterprise again and doing it now?

00:07:34.563 --> 00:07:38.823
I think that's a very hard question. Well, there must be a reason why you think

00:07:38.823 --> 00:07:40.683
it's worth picking that up.

00:07:41.343 --> 00:07:45.323
So I think that what is really difficult

00:07:45.323 --> 00:07:51.383
about this stuff is that the work that was done originally was so concrete and

00:07:51.383 --> 00:08:01.083
so well done that I think you have to have a level of confidence with mathematics

00:08:01.083 --> 00:08:05.063
and with physics in order to make progress in those fields.

00:08:05.403 --> 00:08:12.683
But I think that for somebody who has biological questions in their mind,

00:08:12.763 --> 00:08:15.963
who wants to ask about fitting these

00:08:15.963 --> 00:08:20.003
pattern-generating systems to real data

00:08:20.003 --> 00:08:22.743
you know examples of real structures that we

00:08:22.743 --> 00:08:26.023
see in the natural world i think that the often the

00:08:26.023 --> 00:08:29.623
people who are interested in those or who have the competencies in

00:08:29.623 --> 00:08:32.883
those uh two things and are not necessarily

00:08:32.883 --> 00:08:37.503
speaking the same language um and so uh you know i'm not a mathematician by

00:08:37.503 --> 00:08:40.843
training i'm not a physicist by training my background was in sort of psychology

00:08:40.843 --> 00:08:45.943
and then computer science um and and and it it is difficult to be in the middle

00:08:45.943 --> 00:08:50.563
where you where you want your models to be constrained by biological facts.

00:08:52.083 --> 00:08:59.683
But in order to do new things or to have new insights, you need to understand

00:08:59.683 --> 00:09:02.443
the maths and the physics. And I think that's why.

00:09:03.658 --> 00:09:08.218
That's why some of the kind of enthusiasm maybe that comes from the potential

00:09:08.218 --> 00:09:12.298
application of these ideas might have kind of gone in and out of fashion.

00:09:12.398 --> 00:09:19.498
I give you my thought after writing that chapter is that it's difficult to get

00:09:19.498 --> 00:09:24.758
the right level of description when you do this kind of work because these initial

00:09:24.758 --> 00:09:27.358
models by Kauffman and others were very abstract.

00:09:27.798 --> 00:09:30.718
And it goes back to Turing as well, obviously.

00:09:31.878 --> 00:09:36.098
Lindenmeier and people like that. So you've got this fantastic early work which

00:09:36.098 --> 00:09:38.998
is extremely abstract but shows the power of these general principles.

00:09:39.518 --> 00:09:45.458
And then people try to apply it to understanding a particular biological system like the brain.

00:09:45.918 --> 00:09:50.538
Then you encounter all this rich wealth of data, and you don't know which bits

00:09:50.538 --> 00:09:53.818
are going to help you go from beyond these general principles.

00:09:54.398 --> 00:10:01.158
And the problem is that if you don't take enough from the data,

00:10:01.258 --> 00:10:04.318
then your model is under constraint, so why are people going to take it seriously?

00:10:04.598 --> 00:10:09.898
If you do take too much from the data, then your model is overcomplicated and

00:10:09.898 --> 00:10:11.818
it won't do what you want to do.

00:10:11.918 --> 00:10:15.978
So this is a fine line that you have to walk. So what's your strategy?

00:10:16.218 --> 00:10:20.098
I think that's absolutely right. So the strategy in the project that I'm working

00:10:20.098 --> 00:10:24.418
on, which is a sort of collaboration between myself as a computationalist and

00:10:24.418 --> 00:10:28.258
Leah Grubitzer and Kelly Huffman in the States, who are biologists.

00:10:30.718 --> 00:10:36.538
We're sort of asking the question about cortical evolution and development on two levels.

00:10:37.358 --> 00:10:45.398
One is to try and recreate in simulation the kinds of patterns that are out

00:10:45.398 --> 00:10:51.718
there in the natural world, and to calibrate that to data from experiments that those kinds are.

00:10:52.595 --> 00:10:58.855
Conducting so that we can end up with a model that describes basically what

00:10:58.855 --> 00:11:02.175
we think has happened in the biological world.

00:11:02.395 --> 00:11:07.475
And then separately, what I'm really interested in, and we all are,

00:11:07.615 --> 00:11:12.375
is then asking a question that you've thought about as well,

00:11:12.495 --> 00:11:22.155
Tony, which is what is the the design space in which evolution and development have been interacting.

00:11:22.515 --> 00:11:27.435
So on the one hand, we kind of want to ask, how can we account for those patterns

00:11:27.435 --> 00:11:32.475
that you see that are out there in the biology, but also what possible pattern

00:11:32.475 --> 00:11:35.675
forming, what are the constraints on pattern formation.

00:11:37.055 --> 00:11:42.095
What is the design space that evolution and development have been working on?

00:11:42.155 --> 00:11:46.235
And I think more of the abstract level of description, goes into that second

00:11:46.235 --> 00:11:51.435
pursuit, mapping out the design space of evolution and development,

00:11:51.695 --> 00:11:55.075
but then calibration to specific experimental results.

00:11:55.175 --> 00:11:57.495
How does this gene affect this patterning?

00:11:57.895 --> 00:12:06.495
That's the half of what we're doing, which is much more grounded in what a biologist can measure.

00:12:06.955 --> 00:12:11.255
To also add a little bit to Tony's historical summary.

00:12:11.675 --> 00:12:14.955
If you go back to the early artificial life, genetic algorithms,

00:12:15.435 --> 00:12:19.955
neural networks, whatever developments, it was very metaphorical.

00:12:20.515 --> 00:12:26.855
And it created very simple models, like cellular automata.

00:12:26.955 --> 00:12:29.975
People are really obsessed with cellular automata, the other genetic algorithms,

00:12:30.315 --> 00:12:34.915
that did things that if you just close your eyes a little bit,

00:12:34.915 --> 00:12:36.795
and you sort of squint at them,

00:12:37.475 --> 00:12:42.075
they might look a little bit that things might look like in some biology book

00:12:42.075 --> 00:12:43.255
if you also squint your eyes.

00:12:44.367 --> 00:12:49.567
And that gave this illusion of control, right?

00:12:49.647 --> 00:12:56.427
And now, 20 years later plus, it hasn't really panned out, right?

00:12:56.487 --> 00:13:00.907
So if you really take issue of criteria for a theory to be able to explain when

00:13:00.907 --> 00:13:04.707
you can control, their progress has been much less, right?

00:13:04.767 --> 00:13:09.987
But now in your work, you try to improve that by really taking very specific

00:13:09.987 --> 00:13:12.087
constraints that you want to look at.

00:13:12.087 --> 00:13:17.227
Also, you're talking, so you take specific brains and you want to model how

00:13:17.227 --> 00:13:21.627
the maps in the cortices of these different brains could develop.

00:13:22.847 --> 00:13:29.607
But then maybe it's important to now leave this bit of metaphorical biology

00:13:29.607 --> 00:13:31.947
behind and go to real science.

00:13:33.527 --> 00:13:39.007
So then the question becomes, what then makes a good model? So because you can

00:13:39.007 --> 00:13:42.507
also already say you get the variability across species, across mammals.

00:13:43.127 --> 00:13:49.527
But if I take two exemplars of the same species and I look at the borders of

00:13:49.527 --> 00:13:51.667
these maps, it can be rather different.

00:13:52.207 --> 00:13:55.347
So it's not even standardized across these individuals.

00:13:55.747 --> 00:13:59.947
So then the question also first becomes, what's really our benchmark here?

00:14:00.127 --> 00:14:01.167
What are we shooting for?

00:14:01.387 --> 00:14:05.547
So what's the benchmark that allows a model to be good enough? Yeah, right.

00:14:06.468 --> 00:14:10.988
I don't think I have an opinion rather than an answer to that,

00:14:11.088 --> 00:14:15.748
but I agree with the perspective that you're taking when you phrase that question.

00:14:17.108 --> 00:14:23.448
It reminds me of the Rosenbluth and Weiner, the best model of a cat is another

00:14:23.448 --> 00:14:25.068
cat, preferably the same cat.

00:14:25.808 --> 00:14:30.968
And that's really, again, a sort of a thought experiment.

00:14:30.968 --> 00:14:37.568
The point of that is that if you try and recreate all of the detail of the thing

00:14:37.568 --> 00:14:42.108
that you're interested in, then you end up with a model which is as complicated

00:14:42.108 --> 00:14:43.848
as the thing that you were trying to understand.

00:14:44.288 --> 00:14:48.548
And no progress has necessarily been made in the construction of that model.

00:14:48.668 --> 00:14:49.468
You haven't learned anything.

00:14:50.408 --> 00:14:56.528
So for me, the thing I'm interested in is creating the model which is as simple

00:14:56.528 --> 00:15:00.448
as it possibly can be in order to account for the impact. But that's the wrong

00:15:00.448 --> 00:15:01.528
side of the equation, right?

00:15:01.568 --> 00:15:06.008
So what I was asking for, what's your empirical benchmark against which you

00:15:06.008 --> 00:15:08.628
will now compare your model to say the model is good enough?

00:15:09.068 --> 00:15:10.568
What's this empirical benchmark?

00:15:11.368 --> 00:15:19.108
So it will be a model that can recreate the variability in cortical boundaries,

00:15:19.428 --> 00:15:25.128
shape and size, that you can see across all of the species that have been catalogued

00:15:25.128 --> 00:15:27.228
to date by the accrual survey and the colleagues.

00:15:27.228 --> 00:15:35.888
And once we can do that, and I'm pretty confident that we can do that with fine-tuning of parameters,

00:15:36.728 --> 00:15:41.028
because I think that the model that I described earlier today is general enough

00:15:41.028 --> 00:15:42.968
that that should be possible.

00:15:43.528 --> 00:15:48.588
What I will then try to do is to remove all of the components of that model

00:15:48.588 --> 00:15:54.348
that are not required in order to account for the variability and at the point where.

00:15:55.341 --> 00:15:59.221
Removing that bit destroys the ability of the model to account for the data.

00:15:59.481 --> 00:16:01.121
That's where I would have gone too far.

00:16:01.481 --> 00:16:08.521
And so my benchmark will be something that can recreate all of the patterning that we have measured.

00:16:10.261 --> 00:16:16.321
And then I will refine that model to remove as many of the implicit assumptions in it as possible.

00:16:16.441 --> 00:16:21.561
So you're saying I fit the model and then I protect against overfitting by minimizing the model.

00:16:21.661 --> 00:16:25.821
That's right. But you also know if I have enough monkeys sitting behind enough

00:16:25.821 --> 00:16:29.481
computers whittling the parameters of any model, you can get your fit.

00:16:29.901 --> 00:16:36.001
So would it be relevant to also look at, for instance, the temporal dynamics

00:16:36.001 --> 00:16:38.001
of development that you also have to match?

00:16:38.201 --> 00:16:42.201
Let's say there will be a certain period of time that an organism needs to develop

00:16:42.201 --> 00:16:45.161
a map. Would that be a relevant constraint to insert?

00:16:46.721 --> 00:16:49.921
Or let's say the DNA specification

00:16:49.921 --> 00:16:54.621
of unblanked parameters must also so it cannot be more than a certain number

00:16:54.621 --> 00:16:59.801
of of basis right yeah so yeah the article strains from Gothenburg in because

00:16:59.801 --> 00:17:07.181
if you only stick to one level of description it might still be undetermined yeah yeah I think yes,

00:17:07.881 --> 00:17:11.201
I think that um.

00:17:13.265 --> 00:17:18.685
It goes back to one of your suggested components of your model is prediction.

00:17:19.045 --> 00:17:23.425
That's not the only thing that models are for. They help you reveal simple things

00:17:23.425 --> 00:17:26.805
as complex and complex things as simple. No, it's explaining predict control. Okay.

00:17:27.245 --> 00:17:30.345
So prediction, I think, is the benchmark, right?

00:17:30.485 --> 00:17:35.525
So can I calibrate the models, formulating them as simply as possible,

00:17:35.645 --> 00:17:38.705
to data that exists from experiments that have been done?

00:17:38.705 --> 00:17:49.225
And can I then run a broken model, generate a prediction about the consequence

00:17:49.225 --> 00:17:51.945
that you'll see in the shaken size of a cortical area,

00:17:52.585 --> 00:17:59.985
and will that be borne out by the biology when we recreate that simulated experiment in the lab?

00:17:59.985 --> 00:18:09.105
And I think that's when I'll know that the modeling's in a good place, I think. Okay.

00:18:09.825 --> 00:18:12.985
So now we have sort of a day where I want to go, right?

00:18:13.025 --> 00:18:17.405
So we have mammals, we have neocortex, we have the development of neocortex,

00:18:17.565 --> 00:18:18.825
different maps of neocortex.

00:18:19.325 --> 00:18:21.705
Where of course there's a layer of Cougar chest, and it's just a nice idea.

00:18:21.965 --> 00:18:26.165
But how this is sort of modular and this kind of co-evolved with the periphery.

00:18:26.165 --> 00:18:30.505
And now in some sense you're saying well these cortical maps emerge because

00:18:30.505 --> 00:18:35.245
you have different gradients essentially that guide the.

00:18:36.825 --> 00:18:40.785
Organization of global circuits or how let's say thalamic projections will iterate

00:18:40.785 --> 00:18:46.285
the cortical map and how cortical neurons will connect to each other right yeah so,

00:18:47.305 --> 00:18:51.365
I could not trivialize this okay well big deal because the students are saying

00:18:51.365 --> 00:18:55.605
well I will need as many gradients as I have sub maps and then I'm done.

00:18:57.709 --> 00:19:04.609
Yes, but that's a departure from what I was trying to explain in the talk earlier.

00:19:04.609 --> 00:19:10.149
So in the talk, what I was suggesting is that if the patterns of gene expression

00:19:10.149 --> 00:19:16.369
across the cortex are themselves generated by a self-organizing process,

00:19:16.449 --> 00:19:20.069
like a reaction-diffusion system, which is what I was thinking of earlier,

00:19:20.469 --> 00:19:31.849
then only some patterns will be possible given a cortical boundary shape and

00:19:31.849 --> 00:19:33.669
a given choice of diffusion constant.

00:19:33.669 --> 00:19:37.469
So there were kind of two free parameters in the model that I presented earlier,

00:19:37.649 --> 00:19:39.349
at that level of description at least.

00:19:39.449 --> 00:19:45.509
The boundary shape range over which chemicals are signaling to one another.

00:19:45.849 --> 00:19:52.569
And what Alan Turing's original analysis, which I'm kind of inheriting into

00:19:52.569 --> 00:19:57.389
thinking about the system that I'm working with, what that shows is that only

00:19:57.389 --> 00:19:58.829
certain patterns are possible.

00:19:59.229 --> 00:20:04.289
And so I'm no more able to sort of handpick.

00:20:04.929 --> 00:20:09.949
There's not an infinite space of possible starting conditions for the model that I run.

00:20:10.289 --> 00:20:18.109
There are pretty good proofs out there about what modes these systems will like

00:20:18.109 --> 00:20:21.209
to be in, what are the low energy states for a self-organising system.

00:20:21.469 --> 00:20:25.549
And my bet is that that is what Natural Selection has been working with.

00:20:25.689 --> 00:20:30.269
It has been finding those low-energy states and cobbling them together in ways

00:20:30.269 --> 00:20:34.149
that enable an animal to better adapt to its environment.

00:20:34.329 --> 00:20:38.089
And I think at that level of description, you know, it's not a free-for-all

00:20:38.089 --> 00:20:39.469
in terms of parameter change. Right.

00:20:39.829 --> 00:20:43.729
So I really appreciate that in your inner model, because there's also the transition.

00:20:44.169 --> 00:20:50.909
I mean, previous models that you described very much rely on that prior of predefined

00:20:50.909 --> 00:20:54.149
gradients, and then in some sense you can get away with a lot, right?

00:20:54.269 --> 00:20:57.469
And also so you can get something that looks like quite easily.

00:20:57.889 --> 00:21:02.089
But indeed, I think you changed that game quite a bit by removing that prior

00:21:02.089 --> 00:21:04.869
now and making it part of the self-organizing process.

00:21:05.949 --> 00:21:12.189
But now, what do we really know about the temporal dynamics or the spatial dynamics

00:21:12.189 --> 00:21:14.669
of these gradients in the developing brain?

00:21:15.249 --> 00:21:19.709
How rapidly are they expressed? How rapidly do they infuse? How stable are they?

00:21:20.349 --> 00:21:24.169
Yeah, I think I'm going to get myself into trouble trying to answer that question

00:21:24.169 --> 00:21:29.569
because the truth is that my this is where my knowledge yeah um one thing that

00:21:29.569 --> 00:21:32.669
i find that i find really interesting in the context of,

00:21:33.664 --> 00:21:39.544
put that, is inheriting from a paper by Giacomo Antonio and Jeff Goodhill in

00:21:39.544 --> 00:21:40.464
Post-Computational Biology,

00:21:40.784 --> 00:21:47.824
2010 I think, where they summarised a bunch of biological facts that were already

00:21:47.824 --> 00:21:49.344
out there in a nice model.

00:21:50.644 --> 00:21:56.604
And what they describe is a kind of minimal network of five genes,

00:21:56.604 --> 00:22:04.864
so FGF8, EMX2, PAX6, QTF1, SP8, all wonderful names.

00:22:06.144 --> 00:22:12.164
And what they observe is that at embryonic day 8 in a mouse,

00:22:13.004 --> 00:22:20.704
only FGF8 is expressed in only the anterior pole of the developing cortical tissue. issue.

00:22:21.444 --> 00:22:28.904
We don't know how FGAF8 comes to be expressed there, but when it is,

00:22:29.064 --> 00:22:35.064
you then have a kind of cascade of interactions that flip these other genes on and off,

00:22:35.204 --> 00:22:40.844
such that they end up forming a patterning of gradients, which could potentially

00:22:40.844 --> 00:22:44.184
be used as a coordinate system for guiding thermocortical innovation.

00:22:44.664 --> 00:22:50.184
That happens over the range of maybe 10, 15 days or so, I think,

00:22:50.224 --> 00:22:52.264
in the development of the mouse cortex.

00:22:53.084 --> 00:23:00.084
So I think what happens is you've got some kind of signal that's extrinsic to

00:23:00.084 --> 00:23:06.044
that network of those five genes that triggers an event which leads to a self-organising

00:23:06.044 --> 00:23:10.004
process from which these complementary gene expression gradients fall.

00:23:10.864 --> 00:23:15.164
And yeah, that's over the course of a few days.

00:23:16.104 --> 00:23:19.064
The other sort thing that i'm interested

00:23:19.064 --> 00:23:22.324
in uh from the context of your question is um

00:23:22.324 --> 00:23:25.484
kaufman's original uh description of

00:23:25.484 --> 00:23:28.784
how these uh how these gradients unfold in

00:23:28.784 --> 00:23:31.884
the uh the embryonic development of the

00:23:31.884 --> 00:23:42.044
drosophila uh egg um and what he imagined is that that as the tissue grows the

00:23:42.044 --> 00:23:46.244
relationship between between the boundary shape and the range over which cells

00:23:46.244 --> 00:23:50.244
are communicating by diffusion or chemical signalling,

00:23:50.364 --> 00:23:59.284
that flips the self-organisation into between a set of predefined modes.

00:24:00.304 --> 00:24:07.984
So when the tissue is small relative to the diffusion size, the mode will be a kind of low mode.

00:24:08.084 --> 00:24:11.604
You'll get a gradient from front to back, and that gives you the distinction

00:24:11.604 --> 00:24:14.484
between the animal's head and its tail.

00:24:14.744 --> 00:24:21.764
And then as the tissue, as the egg grows, that same process now likes to be in a mode where,

00:24:22.444 --> 00:24:30.704
to have more modes imprinted on the tissue, and that gives you then a separation of the, sort of.

00:24:32.116 --> 00:24:35.236
Lateralization of the body and then as the thing continues

00:24:35.236 --> 00:24:38.276
to grow the relationship between chemical diffusion

00:24:38.276 --> 00:24:41.596
and tissue size uh keeps

00:24:41.596 --> 00:24:44.936
uh flipping from into successive

00:24:44.936 --> 00:24:49.636
modes and from there you get the kind of more fine-grained structure of the

00:24:49.636 --> 00:24:56.556
uh of the animal being specified um and so i think there is a there is a place

00:24:56.556 --> 00:25:04.076
for thinking about the time course of these developmental mental events unfolding,

00:25:04.416 --> 00:25:09.896
which is neatly captured by that kind of reaction-diffusion formalism,

00:25:10.016 --> 00:25:17.116
which I haven't really touched on in my work yet, which has mostly been about

00:25:17.116 --> 00:25:21.296
considering how spatial patterns form.

00:25:21.996 --> 00:25:26.736
But now you know the abandoned brain, is it like more or less five radians or

00:25:26.736 --> 00:25:27.856
is it more? What's the set?

00:25:28.896 --> 00:25:35.356
In descriptions I've seen from people who are more biologically informed than I am, there is...

00:25:35.356 --> 00:25:41.576
So, Ermin Traut's claim, whose work I piggybacked off today.

00:25:42.896 --> 00:25:51.796
Was that a minimal circuitry would involve three genes, which is FGF8, EMX2, and PAK6.

00:25:52.996 --> 00:25:57.336
Other descriptions have defined a minimal gene interaction network,

00:25:57.496 --> 00:25:59.916
which encompasses five.

00:26:00.496 --> 00:26:06.456
When I talked to Leah Kruvitzer about this, she's not impressed by the claim

00:26:06.456 --> 00:26:07.596
that only five genes are involved.

00:26:07.896 --> 00:26:10.296
I think there are at least...

00:26:12.050 --> 00:26:26.850
And of course, the genes that regulate F and efferent expression are also sort of players as well.

00:26:26.930 --> 00:26:28.930
And are these genes, again, controlled by master genes?

00:26:29.790 --> 00:26:33.230
Is it really a regulated expression pattern, or they work independently?

00:26:34.410 --> 00:26:38.890
I don't know. my claim.

00:26:40.410 --> 00:26:44.930
Because I would predict in your case you would need some sort of controlled expression,

00:26:45.570 --> 00:26:48.530
of your gradients right if they wouldn't go off

00:26:48.530 --> 00:26:56.270
independently to very different kinds of map organization so there have been

00:26:56.270 --> 00:27:01.410
experiments done and the way that you arrive at a kind of description of there

00:27:01.410 --> 00:27:07.750
being three main genes that are involved old EMX packs and FGF8,

00:27:08.050 --> 00:27:12.010
the way you arrive at that is to do knockout experiments.

00:27:12.530 --> 00:27:21.110
And if you knock out packs 6, there's not much of an impact on FGF8.

00:27:21.310 --> 00:27:26.210
There is a bit, but you'll still get patterning. If you knock out EMX2,

00:27:26.530 --> 00:27:31.590
you'll get no interaction between doing PAX6 and FGF8, but you'll still get

00:27:31.590 --> 00:27:33.230
patterning. It'll be a little bit disturbed.

00:27:33.410 --> 00:27:42.530
If you knock out the transcription factor FGF8, the morphogen FGF8,

00:27:42.650 --> 00:27:46.130
you'll get no patterning or very disturbed patterning.

00:27:46.250 --> 00:27:53.670
And so I think the picture of exactly how these genes interact with one another

00:27:53.670 --> 00:27:56.210
to affect patterning on the cortex is,

00:27:56.250 --> 00:28:05.250
you know, that's a question that developmental neurobiologists have been working on and they're...

00:28:07.496 --> 00:28:12.156
They have been creating a picture of this gene interaction network,

00:28:12.316 --> 00:28:17.816
which I'm taking from the textbooks and representing in my equations at some level.

00:28:19.576 --> 00:28:24.276
But what I'm trying to do a little bit beyond that is to ask about the space

00:28:24.276 --> 00:28:27.576
of all possible gene interaction networks.

00:28:27.576 --> 00:28:30.596
Works you know which you know okay in

00:28:30.596 --> 00:28:33.996
in in uh the examples of animals

00:28:33.996 --> 00:28:36.796
with cortical patterning that we have on planet earth at the moment

00:28:36.796 --> 00:28:39.656
uh the you know the network looks like this

00:28:39.656 --> 00:28:43.336
but but does it have to look like this does it matter that it's specifically

00:28:43.336 --> 00:28:47.716
those genes that are talking to each other or is it just that you know you need

00:28:47.716 --> 00:28:54.136
to have um uh you need to have a minimal animal network that comprises X genes

00:28:54.136 --> 00:28:58.056
of which there are N interactions between them?

00:28:58.096 --> 00:29:01.676
What are the general principles of constructing pattern-forming,

00:29:01.856 --> 00:29:05.916
pattern-constraining gene interaction networks?

00:29:06.896 --> 00:29:13.536
I think that's a separate question from what happens in the mouse brain on planet Earth in 2018.

00:29:14.396 --> 00:29:21.216
So that was extremely long-winded. I like the approach, but to be devil's advocate,

00:29:21.556 --> 00:29:29.156
I think that taking the constraint of cortical area and the shape of the cortical

00:29:29.156 --> 00:29:31.076
area in the adult animal,

00:29:31.636 --> 00:29:37.856
there are obvious problems in that.

00:29:37.856 --> 00:29:42.496
That one is you know it assumes aerialization is a late process that you know

00:29:42.496 --> 00:29:48.696
that shape happens first yeah it also assumes that um aerialization is certainly

00:29:48.696 --> 00:29:53.176
complete in in junior in young animals sort of maybe uh.

00:29:54.209 --> 00:29:56.949
Maybe they're already born, but, you know, a few weeks old at most.

00:29:57.669 --> 00:30:00.689
And they still have a lot of growing to do, including growing brain.

00:30:01.109 --> 00:30:08.669
So, I mean, are those assumptions borne out? So that's a really important observation.

00:30:09.109 --> 00:30:14.809
But I'd say it's an observation of the model that I presented at the stage of

00:30:14.809 --> 00:30:16.389
development, which I'm at, right?

00:30:16.649 --> 00:30:22.469
Yes. I do not believe that… So my assumption,

00:30:22.729 --> 00:30:27.169
the assumption that's represented in the model I presented is that only the

00:30:27.169 --> 00:30:33.049
boundary shape and the diffusion constants specify the process by which.

00:30:34.149 --> 00:30:35.429
Arialization occurs.

00:30:35.789 --> 00:30:39.829
And I showed examples of, okay, if you have different boundary shapes,

00:30:39.909 --> 00:30:43.689
as you see in different species, then how do they constrain these processes

00:30:43.689 --> 00:30:45.149
and generate generate different patterns.

00:30:45.369 --> 00:30:53.189
I personally agree with you that the model is not complete as I presented it

00:30:53.189 --> 00:31:00.449
today in the sense that the boundary itself changes over time.

00:31:00.569 --> 00:31:04.709
It expands both in size and it changes in shape,

00:31:04.929 --> 00:31:08.009
presumably under the influence of genetic processes

00:31:08.009 --> 00:31:11.449
which are not in my model and and and

00:31:11.449 --> 00:31:14.449
i haven't yet posed that question of

00:31:14.449 --> 00:31:18.469
the system that i've i've developed so um so

00:31:18.469 --> 00:31:21.609
i don't think that the adult boundary shape on its own is

00:31:21.609 --> 00:31:27.289
enough but i also what we're basically working on now is is a is a is a way

00:31:27.289 --> 00:31:32.749
of thinking about the how the shape morphs from some initial probably quite

00:31:32.749 --> 00:31:37.989
regular shape into the um the adult shape and we want to build that into the

00:31:37.989 --> 00:31:40.149
process of development.

00:31:41.669 --> 00:31:46.949
And this is where the collaborators in UC Davis and UC Riverside come in because

00:31:46.949 --> 00:31:50.089
they are currently measuring the...

00:31:51.076 --> 00:31:56.536
Boundary shapes and the efferent expression and gene expression patterns across

00:31:56.536 --> 00:32:01.436
those shapes at different points in development and in different species.

00:32:02.016 --> 00:32:10.236
So we don't have data, or I don't have knowledge of data, on the changing boundary

00:32:10.236 --> 00:32:12.216
shapes during embryonic development,

00:32:12.536 --> 00:32:16.536
but we're collecting that as part of the project and as that data comes in,

00:32:16.596 --> 00:32:20.376
that can then be used as an additional set of constraints in the model.

00:32:20.536 --> 00:32:26.316
Okay. Could you imagine a version of the model where a boundary shape is an

00:32:26.316 --> 00:32:27.496
emergent property of the model?

00:32:27.996 --> 00:32:31.636
Once it's appropriately defined, you take it out altogether as a constraint.

00:32:32.156 --> 00:32:36.576
I mean, obviously, the constraints is the size of the skull,

00:32:36.696 --> 00:32:38.616
but I mean, all of these things are changing in time.

00:32:40.696 --> 00:32:44.636
It's interesting if you look at the the literature uh

00:32:44.636 --> 00:32:48.596
on uh for example experiments with

00:32:48.596 --> 00:32:51.596
mutant mice in their ability to flip

00:32:51.596 --> 00:32:55.116
one gene and see some cascade of

00:32:55.116 --> 00:33:00.996
interactions with the developmental process that you know that compensate for

00:33:00.996 --> 00:33:05.096
the effect of flipping that gene so that rather than creating an animal that

00:33:05.096 --> 00:33:09.716
dies you know there are the the developmental process adjusts to incorporate

00:33:09.716 --> 00:33:12.596
that actually flipping. Yeah, I can imagine that.

00:33:12.876 --> 00:33:19.056
I can't imagine to the level at which I would type code into a computer and recreate that process.

00:33:19.156 --> 00:33:25.196
We haven't done that yet, but it's something that we want to do.

00:33:25.516 --> 00:33:27.176
So just to give you a kind of.

00:33:28.309 --> 00:33:33.889
Idea. So a lot of the work we've done as well as the mathematics and trying

00:33:33.889 --> 00:33:39.209
to understand the biology through the modelling is the software development,

00:33:39.449 --> 00:33:42.209
which has been done by Seb James in Sheffield.

00:33:42.449 --> 00:33:48.829
We have this kind of method for simulating these self-organising processes on

00:33:48.829 --> 00:33:51.709
a hexagonal lattice with an arbitrary boundary shape.

00:33:51.949 --> 00:33:59.349
So what we do is we take a drawing that's been made by a biologist in vector graphics.

00:33:59.589 --> 00:34:06.629
And we take that boundary shape, we cut out of a hexagonal lattice a simulation

00:34:06.629 --> 00:34:11.229
domain, and then we run the evolution of the equations on that domain.

00:34:12.029 --> 00:34:18.789
So we have a set of software tools that allow us to simulate self-organizing

00:34:18.789 --> 00:34:21.589
processes on any boundary shape that you could imagine,

00:34:21.589 --> 00:34:24.469
in filling in the details of what that boundary

00:34:24.469 --> 00:34:28.309
shape should be to run a particular simulated experiment or

00:34:28.309 --> 00:34:31.249
to you know somehow we need to define what

00:34:31.249 --> 00:34:36.029
those boundaries are um and at the moment we're doing it by drawings but i'd

00:34:36.029 --> 00:34:40.829
certainly be interested in perhaps talking to you a little bit more about uh

00:34:40.829 --> 00:34:48.569
to get your thoughts on how um on how the biology shapes the boundary at the

00:34:48.589 --> 00:34:53.369
moment we're saying that a biologist gives us a drawing of a boundary and that's

00:34:53.369 --> 00:34:59.829
our constraints on the model i think you're right the biology also plays around

00:34:59.829 --> 00:35:04.449
with boundary shape in a way that that i don't have a set of formal i don't

00:35:04.449 --> 00:35:07.709
have a formal model of that process so the um.

00:35:08.629 --> 00:35:12.569
You showed us these pictures of different mammals that you are looking to fit

00:35:12.569 --> 00:35:17.949
the brain to and you There's a huge diversity of different animals,

00:35:18.289 --> 00:35:23.529
and particularly if we're looking at the sensory motor areas of the brain,

00:35:24.249 --> 00:35:27.589
very much the morphology of the animal, its lifestyle.

00:35:29.809 --> 00:35:37.349
Largely predicts the size of some of these areas. If you get a blind mole rat, huge tactile area.

00:35:37.769 --> 00:35:40.829
Or a squirrel, highly visual animal.

00:35:41.129 --> 00:35:47.429
So those are other constraints on your model. So how rich do you want to go

00:35:47.429 --> 00:35:50.049
in terms of incorporating those kinds of constraints?

00:35:52.349 --> 00:36:00.229
What can you do to take the model towards being more realistic in terms of reflecting

00:36:00.229 --> 00:36:04.429
what's happening in terms of the evolving lifestyle of the animal?

00:36:09.171 --> 00:36:12.531
So I'll answer that on two fronts, if I can.

00:36:12.631 --> 00:36:16.851
So one is that there is nothing about the model which, as I've described it today,

00:36:17.111 --> 00:36:28.171
that excludes the possibility of thinking about stimulus-driven processes.

00:36:31.011 --> 00:36:37.211
So, for example, I showed patterns of orientation maps forming within the boundaries

00:36:37.211 --> 00:36:40.691
that I showed, And that model was based on a model by Fred Wolff.

00:36:41.451 --> 00:36:47.411
It's self-organizing, but it is self-organizing under the influence of sensory input.

00:36:47.711 --> 00:36:53.471
So that model was on each time step you have the self-organizing dynamics are

00:36:53.471 --> 00:36:58.871
adjusting the receptive fields towards the direction of the current input pattern.

00:36:59.071 --> 00:37:04.771
And you make a choice as a modeler about how you think where those input patterns

00:37:04.771 --> 00:37:08.331
come from. Do you take them from natural image statistics, for example,

00:37:08.331 --> 00:37:09.591
is one thing that you can do.

00:37:09.971 --> 00:37:16.091
And if you bias the input statistics to the model, you'll get a different organization,

00:37:16.751 --> 00:37:20.811
at the other end, which represents those biases in the input statistics.

00:37:21.391 --> 00:37:29.251
So there's nothing about the modeling at the moment which closes the description

00:37:29.251 --> 00:37:33.151
of how the functional organization emerges from the outside world.

00:37:35.511 --> 00:37:40.631
The other kind of aspect of that is some of my other work,

00:37:40.651 --> 00:37:50.311
which is where we're trying to imagine where the behavioral constraints on the

00:37:50.311 --> 00:37:52.011
sensory inputs that come into the brain,

00:37:52.111 --> 00:37:56.971
how does the behavior of the animal constrain the nature of its developmental

00:37:56.971 --> 00:38:03.611
experiences which are then driving these self-organizing processes in the brain?

00:38:03.611 --> 00:38:06.391
And for that, my approach has

00:38:06.391 --> 00:38:12.211
been to think about collective behaviour in animal groups, specifically,

00:38:12.451 --> 00:38:19.071
as you're aware of, work on rodent huddling, which I've described as a self-organising

00:38:19.071 --> 00:38:24.091
process where the individual animals are competing for nice warm locations in

00:38:24.091 --> 00:38:24.811
the centre of the huddle.

00:38:24.811 --> 00:38:32.751
But in doing so are crashing into each other and having contingencies between their visual input,

00:38:33.011 --> 00:38:37.071
their somatosensory input, the noises they hear from inside the huddle being

00:38:37.071 --> 00:38:42.131
determined by, in that case, one environmental parameter, which is the temperature of the environment.

00:38:42.971 --> 00:38:47.911
And I think it's kind of interesting to...

00:38:49.348 --> 00:38:55.408
To think about the organization of the brain and self-organization of the brain during development,

00:38:56.068 --> 00:39:02.208
as being something which is coupled to the body morphology, which determines

00:39:02.208 --> 00:39:07.508
the interaction that the body has with its environment,

00:39:07.868 --> 00:39:11.008
which includes the physical environment, and in the case of huddling,

00:39:11.048 --> 00:39:16.148
also the kind of social context in which the animal's developing.

00:39:16.568 --> 00:39:19.568
But then you reason to move it also more to sort of epigenetic view

00:39:19.568 --> 00:39:23.508
on this right yeah of course we have to specify what then

00:39:23.508 --> 00:39:26.368
these feedback mechanisms might be yeah to your

00:39:26.368 --> 00:39:29.988
to your map formation yeah so I think so the

00:39:29.988 --> 00:39:34.328
reason why I want to go as far as connecting things to the the sort of huddling

00:39:34.328 --> 00:39:41.048
work is because in that context it is very easy very clear to define what natural

00:39:41.048 --> 00:39:44.328
selection should be caring about and I think maybe that's at the crux of some

00:39:44.328 --> 00:39:47.068
of these some of some of this this discussion is,

00:39:47.168 --> 00:39:54.448
you know, if we're going to, we haven't yet said why evolution should try and

00:39:54.448 --> 00:39:55.988
make one pattern versus another.

00:39:56.148 --> 00:40:00.848
What is it that's good about some patterns in the brain versus others?

00:40:00.968 --> 00:40:04.528
And that's a question I've been sort of struggling with personally for a long

00:40:04.528 --> 00:40:09.628
time is, you know, if you want to get an evolutionary algorithm to constrain.

00:40:10.747 --> 00:40:14.627
The initial conditions of your self-organising processes, what's the fitness function?

00:40:15.067 --> 00:40:21.887
And I think in the context of the huddling, the fitness function is actually reasonably clear.

00:40:21.967 --> 00:40:29.647
The animal that exploits self-organising interactions from contact with the

00:40:29.647 --> 00:40:32.427
slitomates such that it uses less energy,

00:40:32.667 --> 00:40:41.347
such that it minimises metabolic costs in doing so, It should be one that is

00:40:41.347 --> 00:40:43.967
more favoured by natural selection.

00:40:45.987 --> 00:40:49.367
So that's, I think, ultimately everything needs to be… Yeah,

00:40:49.367 --> 00:40:50.447
well that's circular almost.

00:40:50.707 --> 00:40:55.487
That is circular, right? Because in the end that would mean that we don't huddle

00:40:55.487 --> 00:40:58.867
whenever we stand on top of each other or something, because all the huddlers,

00:40:59.047 --> 00:41:00.947
the guys who go to the centre win.

00:41:01.527 --> 00:41:04.967
So no one stands on the periphery. But there are two, so yeah,

00:41:05.067 --> 00:41:08.107
that's really interesting, but there are two ways that you can win.

00:41:08.107 --> 00:41:16.427
So you can either win and be at the center because you generate lots of heat

00:41:16.427 --> 00:41:18.967
and you're trying to cluster around you,

00:41:19.027 --> 00:41:23.327
which is individually expensive to you, but is beneficial to the group.

00:41:23.607 --> 00:41:31.467
Or you can win because you can exploit the heat that's generated by those that are at the center.

00:41:31.467 --> 00:41:36.567
So what you have in the huddle is this kind of balance of cooperation and competition,

00:41:36.907 --> 00:41:44.447
which I think selection should care about, because there is a good balance.

00:41:44.447 --> 00:41:48.767
There is, there is, um, Peter Van Doren Yeah, but I find that through problematics

00:41:48.767 --> 00:41:54.127
because the problem we've tried to solve, I think, in the end is where are the

00:41:54.127 --> 00:41:57.187
boundary conditions, where do the boundaries come from in your model? Yeah.

00:41:57.467 --> 00:42:00.587
Peter Van Doren Because that's now the assumption. You don't assume the whole

00:42:00.587 --> 00:42:02.227
gradient, but you do assume boundaries.

00:42:03.923 --> 00:42:10.123
And I don't really see how your huddling analogy helps us to solve that problem.

00:42:10.403 --> 00:42:13.243
Yeah, I don't think it's far too far away from what we were talking about.

00:42:13.603 --> 00:42:16.143
But also in your lecture, the huddling came up.

00:42:16.523 --> 00:42:20.063
Yeah. And also then I didn't really see the link. And of course,

00:42:20.063 --> 00:42:24.003
you can say, well, there are self-organizing processes, and they might be driven

00:42:24.003 --> 00:42:27.483
by simple rules and leading to complex results.

00:42:27.803 --> 00:42:30.443
But still now, we discussed this

00:42:30.443 --> 00:42:35.603
earlier, right? So the old models made assumptions about whole gradients.

00:42:35.843 --> 00:42:39.363
Then you made the next step and said, no, no, gradients are part of the self-organizing

00:42:39.363 --> 00:42:41.383
process and they can get away.

00:42:42.063 --> 00:42:47.083
I can agree to them as long as I still define their boundaries, right?

00:42:47.163 --> 00:42:53.163
So now we'd say, well, you know, you got from the panel to the fire because

00:42:53.163 --> 00:42:55.583
you still explained now where the boundaries come from. Yeah.

00:42:56.143 --> 00:43:00.283
Huddling or no huddling. Yeah, yeah, yeah. Yeah, let's ignore huddling. so

00:43:00.283 --> 00:43:03.143
so okay where

00:43:03.143 --> 00:43:06.283
so so how are we going to solve this uh i think

00:43:06.283 --> 00:43:13.743
i have to accept that that is that's where we've got to in in in what i've been

00:43:13.743 --> 00:43:20.063
considering so if correctly your the boundary is the is the final the stable

00:43:20.063 --> 00:43:24.283
point of the map right that's what the boundary defines finds?

00:43:24.423 --> 00:43:29.323
Sorry, there are two levels of boundary. There's the overall boundary,

00:43:29.463 --> 00:43:34.143
which is the edge of the cortex, if you like, and then within that we have the

00:43:34.143 --> 00:43:35.763
boundaries of the different domains.

00:43:35.963 --> 00:43:38.183
Right, exactly. So, at.

00:43:39.599 --> 00:43:45.739
So my sort of claim in the talk today was that what we have in this model is

00:43:45.739 --> 00:43:50.919
a description where there are a bunch of constants that go into the parameters of the model,

00:43:51.019 --> 00:43:55.539
and diffusion constants mostly in coupling strengths between genes that are

00:43:55.539 --> 00:43:56.879
interacting in the network and so on.

00:43:56.879 --> 00:44:04.679
And that in addition to that choice of parameter values, all kind of scalar

00:44:04.679 --> 00:44:06.659
numbers, which I think could be,

00:44:08.059 --> 00:44:12.239
sort of specified by the genetic code, if you like, in addition to that,

00:44:12.299 --> 00:44:19.319
there is the shape of the cortex which imposes a boundary condition on all of

00:44:19.319 --> 00:44:21.559
those self-organising processes.

00:44:23.559 --> 00:44:31.299
And that is as far as I've got. I agree with Tony that there's a step further

00:44:31.299 --> 00:44:36.759
if we really want to pin everything down to the level at which natural selection

00:44:36.759 --> 00:44:38.719
operates on this, which is tinkering with the DNA,

00:44:38.919 --> 00:44:46.979
then we need to think about how the DNA is defining that boundary, hand-cut boundary,

00:44:47.199 --> 00:44:49.959
which changes in shape and size over the developmental period.

00:44:49.959 --> 00:44:56.079
So that's, there's another level of kind of, if you like, reducing the system

00:44:56.079 --> 00:45:00.679
onto a genetic code, which I haven't done yet.

00:45:01.159 --> 00:45:03.459
And for which I don't have clear

00:45:03.459 --> 00:45:08.019
ideas right now today in this room about exactly how I would do that.

00:45:08.019 --> 00:45:13.319
Are you considering a recapitulation hypothesis because if you look at this

00:45:13.319 --> 00:45:18.059
hierarchy of mammalian brains, you would argue, well, from a developmental perspective,

00:45:18.179 --> 00:45:22.719
the more advanced brains go to a stage that they look somewhat like the symbol brain.

00:45:23.701 --> 00:45:27.001
But that would mean that at that stage, they have the boundary conditions of

00:45:27.001 --> 00:45:28.161
that simple brain, right?

00:45:28.241 --> 00:45:35.581
So you could, from that heuristic here, then predict that as long as you make

00:45:35.581 --> 00:45:40.581
sure that your developmental trajectory follows roughly the shape of the whole

00:45:40.581 --> 00:45:41.941
series of mammalian brains,

00:45:42.681 --> 00:45:46.381
if I want, then I just need the parameter settings of this intermediate stage.

00:45:46.501 --> 00:45:49.961
So the core process at that developmental stage is the same.

00:45:49.961 --> 00:45:56.721
All brains have that shape all of us have that volume I see where you're going

00:45:56.721 --> 00:45:58.801
with that I think so my answer to that is no,

00:45:59.461 --> 00:46:04.501
I'm not thinking about recapitulation so I'm not thinking about recapitulation

00:46:04.501 --> 00:46:09.801
at this point because I think that what discriminates in the model as it's described at the moment.

00:46:10.941 --> 00:46:17.061
What discriminates between you know this species X and species Y is not the

00:46:17.061 --> 00:46:23.861
species Y that emerged later in evolution had to have gone through the boundary

00:46:23.861 --> 00:46:28.541
conditions of species X in order to get there during its development.

00:46:29.041 --> 00:46:32.401
I mean, that might be true. That's an open question.

00:46:33.121 --> 00:46:37.301
But at the moment, the thinking is that species X and species Y have different

00:46:37.301 --> 00:46:45.081
parameters for the same set of self-organising processes, but they're parameterised

00:46:45.081 --> 00:46:47.921
slightly differently, constrained by different boundary conditions,

00:46:48.461 --> 00:46:54.901
which I acknowledge I don't have an explanation for genetically how the difference there is specified.

00:46:55.761 --> 00:47:00.501
But we're talking about different initial conditions on the same developmental mechanism,

00:47:00.701 --> 00:47:12.341
not having a developmental scaffold which is reflecting the evolutionary branching.

00:47:14.736 --> 00:47:18.056
Yeah, I mean, I think we're talking about factors which aren't included in the

00:47:18.056 --> 00:47:22.896
model, which is a bit mean considering the model's only existed for a few months.

00:47:23.196 --> 00:47:26.156
And you're starting from this point.

00:47:26.316 --> 00:47:34.656
But just continuing to speculate about that, one of the big differences in mammals is brain size.

00:47:34.976 --> 00:47:39.636
And obviously some of the factors in the cannabinoids you're talking about are

00:47:39.636 --> 00:47:44.296
going to be, size is going to be an important variable. So it's not simply going

00:47:44.296 --> 00:47:45.436
to scale with a bigger brain.

00:47:45.776 --> 00:47:48.996
But you could take species which are otherwise similar.

00:47:49.136 --> 00:47:53.136
I mean, for example, you've got the Brazilian short-tailed opossum.

00:47:53.196 --> 00:47:56.736
You've got the Virginia opossum. One's bigger.

00:47:56.816 --> 00:47:59.296
I'm not sure if the brain's a lot bigger, but I think it's a bit bigger.

00:47:59.396 --> 00:48:04.796
And then you've got all sorts of other animals where you can get big and small versions.

00:48:05.576 --> 00:48:08.536
So, I mean, that might be another useful parameter to look at.

00:48:08.616 --> 00:48:12.836
How brain size scaling impacts on all of this.

00:48:13.156 --> 00:48:20.896
So that is a really important component of this, and that is one of the kind

00:48:20.896 --> 00:48:24.836
of questions that we want to ask in the bigger picture of this project that we started.

00:48:26.716 --> 00:48:33.156
My understanding inherited from talking with Leah is that the differences between

00:48:33.156 --> 00:48:35.736
some pairs of mammals, models,

00:48:35.776 --> 00:48:45.556
some relationships between brain size and brain area size scale linearly,

00:48:45.816 --> 00:48:49.636
and some do not. Some scale non-linearly.

00:48:50.156 --> 00:48:56.056
And I think, as we discussed a little bit earlier today, I think currently in

00:48:56.056 --> 00:48:57.536
the formulation of the modelling,

00:48:57.876 --> 00:49:02.336
there is no parameter that I can

00:49:02.376 --> 00:49:05.816
point to in the model which should in some situations

00:49:05.816 --> 00:49:08.956
give you the linear scaling of brain area size

00:49:08.956 --> 00:49:11.876
with brain volume size versus a non-linear scaling

00:49:11.876 --> 00:49:15.096
i haven't done the experiments with the model to to

00:49:15.096 --> 00:49:19.796
test that's true but i also in constructing the model i don't have the model

00:49:19.796 --> 00:49:25.356
as it stands doesn't have um a natural kind of parameter that's that can be

00:49:25.356 --> 00:49:29.976
tweaked in order to make on the one hand it's scaled in every on the other hand

00:49:29.976 --> 00:49:33.936
non-linear and so i think but But, you know, this is where we need to go to the biology.

00:49:34.056 --> 00:49:37.776
This is where we need to go to the data and ask the question,

00:49:37.996 --> 00:49:41.756
you know, where might those differences come from?

00:49:41.816 --> 00:49:46.816
What is different about the relationship between the genes of two species for

00:49:46.816 --> 00:49:51.116
which the scaling is linear and two species for which the relationship is nonlinear?

00:49:51.456 --> 00:49:55.696
You know, what are those differences and can we kind of recreate some of those differences?

00:49:55.836 --> 00:49:58.236
That would be an example of what we talked about earlier where we said,

00:49:58.356 --> 00:50:03.316
you know, We're looking for the minimal model that can account for all of these things.

00:50:03.516 --> 00:50:07.536
And I think that at the moment, the model that I presented is,

00:50:08.488 --> 00:50:13.668
probably at present cannot account for those differences, and so it needs to be refined.

00:50:16.628 --> 00:50:24.188
But I think that's part of the usefulness of modelling here, is that we can take these,

00:50:24.968 --> 00:50:27.768
reasonably informal descriptions about how the biology works,

00:50:27.908 --> 00:50:34.828
and we can translate those those informal assumptions into a sort of mathematically

00:50:34.828 --> 00:50:38.268
well-defined representation of those assumptions.

00:50:38.408 --> 00:50:42.548
Then we can play around with them and we can find out what we don't know and

00:50:42.548 --> 00:50:46.288
what can't be found. Well, we can also do experiments that haven't happened in biology.

00:50:46.528 --> 00:50:53.268
So we can play with impossible cortical sizes or we can also just experiment

00:50:53.268 --> 00:50:56.048
with inventing parameters,

00:50:56.248 --> 00:50:58.948
if you like, and seeing what influence that has and then saying,

00:50:59.048 --> 00:51:04.508
well, this parameter might exist in the biology because it would be useful to have a gene that did X.

00:51:04.928 --> 00:51:10.748
So one of the other things, obviously, that changes with mammals is the number of cortical areas.

00:51:10.868 --> 00:51:14.428
You go from about 15 to 200 in people.

00:51:14.848 --> 00:51:19.408
And size is going to be a factor there, but not just size.

00:51:19.508 --> 00:51:23.768
I think dolphins have big brains, but they don't have as many cortical areas

00:51:23.768 --> 00:51:26.708
for the brain size as you would expect if you looked at humans.

00:51:27.308 --> 00:51:29.048
So what are these factors that.

00:51:30.022 --> 00:51:32.782
Impact on arealization in that

00:51:32.782 --> 00:51:36.142
sense of the number of quarter full areas yeah i i

00:51:36.142 --> 00:51:38.882
think um i mean i have i'm running

00:51:38.882 --> 00:51:43.522
a simulation in my head as we as we speak but the the kind of natural way to

00:51:43.522 --> 00:51:48.042
approach that from modeling perspective i think is to look at the the relationship

00:51:48.042 --> 00:51:54.222
between the um diffusion constants representing how how local are local interactions

00:51:54.222 --> 00:51:57.302
versus the size of the domain.

00:51:57.522 --> 00:52:04.842
So essentially, if you shrink your diffusion constants down such that cells

00:52:04.842 --> 00:52:08.662
are communicating very locally, then they will naturally form smaller boundaries.

00:52:09.022 --> 00:52:16.602
And if you form smaller boundaries on a big domain, you will have more cortical areas by definition.

00:52:17.662 --> 00:52:24.862
So that is actually the number of cortical areas that you're going to have is

00:52:24.862 --> 00:52:31.542
actually a very natural question to ask with this level of modeling because small diffusion,

00:52:31.882 --> 00:52:34.802
big boundary, you'll get lots of individual areas.

00:52:35.182 --> 00:52:38.542
So it's exciting because you potentially have a model where you tweak a parameter

00:52:38.542 --> 00:52:39.662
and turn out another paper.

00:52:43.062 --> 00:52:47.782
Well, even more so, the next thing we're going to do is think about evolutionary

00:52:47.782 --> 00:52:51.882
algorithms that can tweak those parameters and evolve a series of papers.

00:52:52.382 --> 00:52:59.222
But think about it like this, right? So what you brought into these models of

00:52:59.222 --> 00:53:03.302
coordinate development are the reaction diffusion models from mandatory.

00:53:03.602 --> 00:53:07.262
Yeah. Well, when you're speculating on very comparable lines,

00:53:07.482 --> 00:53:09.902
right, about pattern formation by natural systems.

00:53:12.161 --> 00:53:15.281
But that's a long time ago. It doesn't mean he was wrong.

00:53:15.641 --> 00:53:19.761
No, no, that's exactly what I wanted to get to. So how much progress have we

00:53:19.761 --> 00:53:23.221
really made? If you would tell the story to Alan Turing, maybe we could revive him.

00:53:23.481 --> 00:53:25.881
Yeah. Would he be surprised? Would he be shocked?

00:53:26.741 --> 00:53:31.641
No, no, I don't think so at all. And I think what has changed.

00:53:33.401 --> 00:53:37.081
I mean, what I'm doing

00:53:37.081 --> 00:53:39.841
at the moment is stitching together what I think are brilliant

00:53:39.841 --> 00:53:42.621
ideas that have been for which the foundations have

00:53:42.621 --> 00:53:46.241
been laid by very smart people um

00:53:46.241 --> 00:53:49.401
uh with with incredible work

00:53:49.401 --> 00:53:52.701
and that includes ermine by derman trout as well actually whose

00:53:52.701 --> 00:54:01.541
ideas i've been directly stitching together um the i think that the that there's

00:54:01.541 --> 00:54:05.761
nothing that there isn't anything new here i think probably if you if you pinned

00:54:05.761 --> 00:54:12.081
if you had a very if you had this level of conversation with many developmental neurobiologists,

00:54:12.381 --> 00:54:15.041
I think they probably tell you that this is how the thing works.

00:54:15.421 --> 00:54:20.241
All I'm trying to do is kind of formalise that so that we can ask you questions of that knowledge.

00:54:21.141 --> 00:54:24.381
I think what has changed though is computing power.

00:54:24.701 --> 00:54:31.901
So I can run the kind of simulation that I showed today in a few minutes.

00:54:32.181 --> 00:54:37.701
And Alan Turing would would have had to, on the Enigma code-breaking machine,

00:54:37.841 --> 00:54:39.961
would have had to have waited quite a long time.

00:54:40.061 --> 00:54:42.401
The Manchester Mark II. Yeah.

00:54:43.301 --> 00:54:47.841
And I think the hard work is in doing the analysis, really.

00:54:48.927 --> 00:54:52.287
But there's now a new era of hardware to do,

00:54:52.427 --> 00:55:00.167
which is when you start connecting these kind of quite computationally intensive

00:55:00.167 --> 00:55:06.087
simulations to an evolutionary context, because what evolution has been doing,

00:55:06.227 --> 00:55:11.767
has been trying out, has been experimenting with these systems en masse for generations.

00:55:12.367 --> 00:55:15.107
Generations and so you know it might take me only

00:55:15.107 --> 00:55:17.807
a couple of minutes to simulate on my computer now but if i

00:55:17.807 --> 00:55:20.907
want to run a sim an evolutionary algorithm

00:55:20.907 --> 00:55:23.667
which considers uh you know

00:55:23.667 --> 00:55:26.827
generations each uh comprise thousands of

00:55:26.827 --> 00:55:30.387
generations comprising thousands of populations of

00:55:30.387 --> 00:55:33.647
these these these things to find uh you

00:55:33.647 --> 00:55:37.047
know what are the what are the most likely candidates for evolution to come

00:55:37.047 --> 00:55:42.147
up with um that's the kind of thing that is you know that it's hard work even

00:55:42.147 --> 00:55:48.487
on on sort of in the current context of computational power is that a little

00:55:48.487 --> 00:55:52.467
bit of the curse set off of the new millennial generation of scientists.

00:55:53.167 --> 00:55:55.967
By not that young but i mean the future we're looking

00:55:55.967 --> 00:55:58.867
at there's some sort of complacency right because we

00:55:58.867 --> 00:56:01.887
have this enormous compute power you can just do dumb things

00:56:01.887 --> 00:56:04.807
which you can look at millions of permutations of dumb things and

00:56:04.807 --> 00:56:07.687
something will happen well i'm ensuring 80 years ago

00:56:07.687 --> 00:56:10.607
or 70 years ago had to

00:56:10.607 --> 00:56:13.387
think very carefully about the problem he had

00:56:13.387 --> 00:56:16.247
to get it right because he couldn't afford it wasn't even a

00:56:16.247 --> 00:56:19.107
possibility to look at millions of permutations of

00:56:19.107 --> 00:56:22.087
a dumb thing yeah right so so in that sense don't you

00:56:22.087 --> 00:56:25.227
feel there's a bit of a risk here that an even increasing computer power

00:56:25.227 --> 00:56:28.487
we sort of sacrifice intellectual power i

00:56:28.487 --> 00:56:35.107
i completely agree with you I like to think that I'm part of a special generation

00:56:35.107 --> 00:56:42.487
that has seen the evolution of computers over my lifetime from when I was interested

00:56:42.487 --> 00:56:47.127
in computers to now where I'm able to develop ideas using computers.

00:56:47.127 --> 00:56:54.207
Computers um i you know i've been at a really good time where i've seen crap computers.

00:56:55.487 --> 00:56:59.587
Start off as being crap and you have to talk to them on a very basic level um

00:56:59.587 --> 00:57:04.647
to to then becoming these things that you can wave your hands at um and that can you know,

00:57:05.387 --> 00:57:13.847
think on orders of magnitudes uh greater than than we can um and so uh so yeah

00:57:13.847 --> 00:57:17.427
i think there And don't forget the excellent mentor that you've had.

00:57:17.587 --> 00:57:20.127
Of course. I forgot to mention that to you.

00:57:20.487 --> 00:57:23.247
So one bit, can I...

00:57:24.522 --> 00:57:33.762
You know advice um code in c++ or code in c because it's hard work but it forces

00:57:33.762 --> 00:57:39.242
you to turn the mathematics of the system that you're studying into if statements and for loops and

00:57:39.442 --> 00:57:45.822
you know you then you understand every component of what you're simulating um

00:57:45.822 --> 00:57:50.962
you know python is wonderful for visualization and these other these other things

00:57:50.962 --> 00:57:54.822
that allow you to talk to your computer in more and more abstract ways are great,

00:57:55.082 --> 00:57:59.542
but I think you know, sometimes when you're stitching together.

00:58:00.822 --> 00:58:03.402
Computational components that have been run by the people,

00:58:04.322 --> 00:58:11.042
you're removing yourself away from the underlying assumptions of the mathematics of the system.

00:58:11.202 --> 00:58:15.842
People take apps and they go in together and they don't really know what happens

00:58:15.842 --> 00:58:17.902
inside. They don't have these patches that they link together.

00:58:18.562 --> 00:58:21.922
That's where we're at now. Yeah, and this is indeed a problem.

00:58:22.202 --> 00:58:25.902
People will be running black boxes. Yeah. I've tried not to.

00:58:26.502 --> 00:58:32.662
I've been grinding away with C++, and it's done me well. Very good.

00:58:33.822 --> 00:58:40.642
So you've taken Turing's reaction to fusion systems, and you've applied it to

00:58:40.642 --> 00:58:45.482
this new data set that Leia's collecting, and you're also taking Stuart Kaufman's

00:58:45.482 --> 00:58:48.122
Boolean nets, and you're trying to plug them in as well.

00:58:48.122 --> 00:58:52.222
So two of these foundational approaches, if you like, in A-life.

00:58:53.442 --> 00:58:59.882
And, you know, Boolean nets are interesting because they are extremely abstract

00:58:59.882 --> 00:59:06.982
compared to the sort of cellular machinery which builds animals.

00:59:07.302 --> 00:59:11.942
But, you know, Kauffman and others have argued that it's an appropriate level of abstraction.

00:59:12.282 --> 00:59:17.702
But you are kind of leaping across different levels of abstraction in building

00:59:17.702 --> 00:59:23.962
these models, picking things which you think might work and are mathematically elegant and so on.

00:59:24.582 --> 00:59:31.222
What's the sort of bigger picture here in terms of a theory of what makes a good theory?

00:59:31.502 --> 00:59:34.022
You know, sort of, you want to make a minimal theory?

00:59:34.682 --> 00:59:38.602
Is the building network the next step in having a minimal theory?

00:59:38.842 --> 00:59:43.822
Or was it that it was an elegant abstraction that you can plug in and you can see how to plug it in?

00:59:43.822 --> 00:59:48.742
I mean, there's a certain amount to which you're, you know, driven by the history

00:59:48.742 --> 00:59:52.282
of your field to use these tools, but maybe you should have,

00:59:52.302 --> 00:59:54.602
we should develop a new set of tools.

00:59:56.162 --> 00:59:56.962
Yeah, I think.

00:59:59.037 --> 01:00:02.277
I see your point um i think that

01:00:02.277 --> 01:00:08.857
the the overarching idea i mean i'm not kind of ready to declare my theory of

01:00:08.857 --> 01:00:16.617
everything but the but the um but that's why we're here but i think but i think

01:00:16.617 --> 01:00:22.177
that it that that it might touch base with something called the baldwin effect effect,

01:00:22.257 --> 01:00:29.557
which is the idea that things that emerge during the interaction between an

01:00:29.557 --> 01:00:31.357
organism and its environment,

01:00:31.637 --> 01:00:33.237
and that can include other agents,

01:00:33.637 --> 01:00:41.857
adaptations to your environment that happen during a lifetime can create a sort

01:00:41.857 --> 01:00:45.817
of scaffold that the genes then climb upon.

01:00:46.197 --> 01:00:52.337
So the genes end up specifying initial conditions for interactions with your

01:00:52.337 --> 01:00:58.337
environment that allow you to get to that within your lifetime of the next generation,

01:00:58.577 --> 01:01:04.417
to get to that good adaptation more quickly than the previous generation.

01:01:04.837 --> 01:01:13.237
And that kind of hints of Lamarckism, but it's which is where the phenotypic.

01:01:13.999 --> 01:01:16.719
Uh information is communicated back to the

01:01:16.719 --> 01:01:19.959
genome breaking the viceman barrier and so on um

01:01:19.959 --> 01:01:23.099
but it's a kind of loophole um for for for

01:01:23.099 --> 01:01:26.039
that if you like which is the organisms that that

01:01:26.039 --> 01:01:33.139
that are that are born with initial conditions that are closer to uh to to fully

01:01:33.139 --> 01:01:39.119
specifying the the the the phenotypic form are going to be favored um uh by

01:01:39.119 --> 01:01:43.539
natural selection um and and not but But isn't this captured in the current

01:01:43.539 --> 01:01:44.859
discussion on epigenetics,

01:01:45.019 --> 01:01:46.719
basically, and so on?

01:01:46.839 --> 01:01:51.199
Yeah, yeah, yeah. But in effect, to me, it's a very sort of caricature of these processes.

01:01:51.579 --> 01:02:01.819
Yeah, very much so. But if there's going to be a type of theory which is going

01:02:01.819 --> 01:02:05.399
to connect those different levels of modelling which Tony was asking about,

01:02:05.559 --> 01:02:13.219
the kind of abstract Boolean network description of gene network evolution and

01:02:13.219 --> 01:02:20.619
reaction diffusion controlled developmental processes for pattern formation in the brain,

01:02:20.619 --> 01:02:26.859
I think you're correct to point out that I've so far been thinking of the sort of polar ends,

01:02:26.979 --> 01:02:29.159
one more about the details of individual brains.

01:02:29.399 --> 01:02:32.339
One about asking about the sort

01:02:32.339 --> 01:02:35.359
of design space in which evolution and development has been occurring.

01:02:35.359 --> 01:02:40.939
And I think the point at which those two things come together will be in the

01:02:40.939 --> 01:02:45.059
context of cortical map formation, because that's always interested me.

01:02:45.059 --> 01:02:48.039
But I think I will be happy at

01:02:48.039 --> 01:02:56.299
the point where I can run a simulation where evolution of development of specific

01:02:56.299 --> 01:03:05.679
cortical plans is accelerated by the constraints that are imposed,

01:03:05.799 --> 01:03:07.979
that are inherent from self-organisation.

01:03:07.979 --> 01:03:10.899
And I think that that will end up looking like a

01:03:10.899 --> 01:03:15.859
demonstration of the Baldwin effect where self-organisation will

01:03:15.859 --> 01:03:25.219
generate useful patterns in the brain and the genes on evolutionary timescales

01:03:25.219 --> 01:03:31.279
will catch up and will become more efficient at guiding the process of self-organisation.

01:03:31.419 --> 01:03:35.459
That to me will be a Baldwin effect and that's where I think the two levels

01:03:35.459 --> 01:03:37.039
of modelling might come together.

01:03:37.979 --> 01:03:41.439
But thanks for asking an extremely difficult question. So Stuart,

01:03:41.519 --> 01:03:45.759
listen, so you're doing all the right things, right? So you made it to BCMET.

01:03:45.919 --> 01:03:49.999
I mean, it's really an early pinnacle in your career. Thank you very much.

01:03:50.379 --> 01:03:55.099
It's just fantastic. You do great work. You really work with the biologists.

01:03:55.119 --> 01:03:57.039
You get the data, the model data.

01:03:57.239 --> 01:03:59.859
So you're making all the right moves. Thank you.

01:04:00.519 --> 01:04:06.999
So if we would like now to take this as a paradigm to understand mind,

01:04:07.079 --> 01:04:11.379
brain and behavior, what is Stuart's law that we should follow?

01:04:13.477 --> 01:04:24.797
Read kaufman everyone should go back to kaufman and turing and and should think,

01:04:25.757 --> 01:04:31.897
should should try to see the value in the models which are defined at the most

01:04:31.897 --> 01:04:40.117
abstract levels and to and to not dismiss them as being abstract and therefore

01:04:40.117 --> 01:04:44.377
of no consequence to understanding real things that are out there in nature.

01:04:44.577 --> 01:04:52.237
I think Stewart's law should be to work hard to think about the details of this

01:04:52.237 --> 01:04:56.117
natural world that we live in in the context of those abstract models,

01:04:56.377 --> 01:05:02.197
because I think some of those original thinkers and people before Kauffman and

01:05:02.197 --> 01:05:04.137
Turing, I think they had it right.

01:05:04.597 --> 01:05:08.817
And I think it's fine

01:05:08.817 --> 01:05:11.717
for the rest of us in their wake to be working to

01:05:11.717 --> 01:05:14.517
catch up and filling in some of the some of the

01:05:14.517 --> 01:05:17.777
the the gaps between the data and and

01:05:17.777 --> 01:05:20.837
those elegant theories no programming in python

01:05:20.837 --> 01:05:26.377
no no no actually somebody

01:05:26.377 --> 01:05:29.937
you don't know about your your ex-mentor or your mentor uh

01:05:29.937 --> 01:05:32.897
don't have a secret project what's this

01:05:32.897 --> 01:05:36.017
he's the monitor of all predictions yeah

01:05:36.017 --> 01:05:38.857
okay so basically everybody we talked

01:05:38.857 --> 01:05:42.197
to in in our podcast um you'll

01:05:42.197 --> 01:05:45.697
have to make a prediction and tony will go check it and think of it as it

01:05:45.697 --> 01:05:49.517
is a british project so we can

01:05:49.517 --> 01:05:54.417
only deal right now with labs that are within a 10 mile radius of tony's office

01:05:54.417 --> 01:06:00.577
um so four years from now tony will go and come to your lab yeah which will

01:06:00.577 --> 01:06:05.017
then still be sheffield i'm short yeah and he will come and check when the prediction

01:06:05.017 --> 01:06:06.857
you make today will be falsified and verified.

01:06:08.292 --> 01:06:11.772
So what's the most central prediction in your research program that you want

01:06:11.772 --> 01:06:16.932
to see tested in that time frame and that you can communicate to Tony in terms

01:06:16.932 --> 01:06:20.632
of it was false or it was true?

01:06:21.312 --> 01:06:24.212
That is a great question. Can I have a moment?

01:06:28.212 --> 01:06:32.732
The central prediction, so the level of the modeling, how would I know that

01:06:32.732 --> 01:06:35.252
my model was wrong? Okay.

01:06:41.732 --> 01:06:44.952
The prediction that I have about the course,

01:06:45.152 --> 01:06:53.392
the potential of this work is that we will be able to account at the resolution

01:06:53.392 --> 01:06:57.312
of the functional organization of the entire cortex,

01:06:57.652 --> 01:07:02.892
we will be able to account for all of the variation that we see between the

01:07:02.892 --> 01:07:05.112
species for which the data has been collected,

01:07:07.452 --> 01:07:12.052
via different parameterizations of essentially the same model that I presented today.

01:07:15.692 --> 01:07:20.892
And that I won't have to add any new equations.

01:07:21.252 --> 01:07:25.472
I might have to tinker a couple of them, but I won't have to add any new equations

01:07:25.472 --> 01:07:31.272
to the modeling in order to account for that variation to the level of detail

01:07:31.272 --> 01:07:33.172
at which it's currently described. All right.

01:07:33.232 --> 01:07:36.852
Stuart Wilson, thank you very much for this conversation. Thank you.

01:07:39.572 --> 01:07:45.392
The CSN Podcast was produced by the Convergent Science Network of Biometrics

01:07:45.392 --> 01:07:51.792
and Biohybrid Systems, a project funded by the European 7th Research Framework Program.

01:07:53.392 --> 01:07:58.652
For more interviews, recorded lectures, or upcoming conferences in the field

01:07:58.652 --> 01:08:04.892
of biometrics and biohybrid systems, go to csnnetwork.eu.

01:08:05.200 --> 01:08:12.720
Music.