Our guest spot

Our guest spot

is Professor David Krakauer. David is a

theoretical biologist and he directs the Wisconsin Institute for

Discovery. He’s also professor of genetics at the University of Wisconsin

and external professor at the Santa Fe Institute.

Before moving to Wisconsin, he was on the resident faculty of SFI and

served as chair of the faculty. So, welcome David.

I’m very pleased to be with you. Great. Earlier

in the class we talked about emergence -- the concept of emergence and the notion

that you could have these emergent properties that have to be

explained at the system level instead of the individual level and

some people have said that emergence is kind of a subjective notion

that maybe it’s only emergent because we’re not smart enough

to understand it at the individual level. Do you think

that’s true? Or is there something fundamental about emergence?

I have to confess I struggle with this concept, and I

am sort of on the fence. Part of me thinks

that remark is true, that you could

always, if you had infinite computational capacity, describe

a system in terms of its most elementary constituents

and that the reason why we look at what we call effective variables

or aggregations of the data is because it’s easier.

So thermodynamics, for example, heat, temperature,

entropy, and so forth. But another part of me says

that that’s insufficient. That if you think

about sophisticated intellectual capabilities

like reasoning about music

it’s not obvious to me that a microscopic description of the dynamics

of the proteins in our neurons will ever

sufficiently capture what we mean by an appreciation of music, so

I have to say I struggle. I don’t have a good answer to that.

Let’s see -- so I asked

you to define complex system, a different

version of that question is to talk about definitions of complexity

itself. Do you think that’s been

hard for people when you ask what is complexity, people often

give very different answers. Do you think there’s a notion that

there’s a single definition of complexity

that we might find that might be useful across systems, or is it really dependent

on the particular system itself?

So I’m quite optimistic about this. I think it that has to

do with the level of predictive accuracy you’re going to demand.

So I think there is a sense in which there could be very general definitions

along the lines that people like Murray Gell-Mann

and Seth Lloyd, and many others, have suggested having to do

with description length, and so forth, in other words

how many pages of algorithms or mathematical

equations are required to capture the variability you care about?

The problem with those measures is they don’t tell you a great deal

about the mechanisms that in some sense are the ones you want to

intervene into or control or understand.

So I think this is where the difficulties arise, because at that point,

these complexity measures start bifurcation and

proliferating and I think that’s okay, actually.

I think that as long as we’re careful and

any one of us states our operational definition

we state clearly our assumptions, our axioms,

I think it’s acceptable for the definitions

to not necessarily be unified. But

I suspect, by the way, if you were to look carefully

all of those slightly different definitions would all have a little bit of the same

flavor, that they would be capturing

something about the

number of components

required to effectively predict and control the environment

I think that’s probably a common denominator, even if on the surface they

appear very different. Okay, maybe

you could tell us a little bit about what you’re working on these days.

What you’re excited about. Yeah, so my very modest

research program is to

in one sense understand the evolution of

you can put it two ways -- the grand way of saying it is

the evolution of intelligence on our planet, in the solar system

and the universe. Which sounds

at one level as ambiguous as the evolution of complexity, right? It’s substituting one ambiguous

term for another. Another way to put it, slightly more modestly,

but not much more, is the evolution of mechanisms

of information processing.

A slight slight of hand, because it’s not really that much clearer

than the first, but people seem to find it more acceptable.

And so, for example,

simple organisms, single-celled organisms,

really can respond to the environments in which they live. They can

pursue food sources, they can travel along gradients

they can aggregate into collectives, and then in the multicellular

state, those cells take on differentiated roles.

Cells that just do metabolism, or cells

that just sense their environment, and so forth, or structural.

And so what I’m interested in is why that

happens, I mean, it’s a funny thing, we know more about

the universe that the smallest scales and the largest scales

than we do than at the scale at which we live. So we understand

extraordinary depth, essentially the origin of our solar system, the origin

of all the elements, the periodic table, you know we understand

nuclear fusion, but we don’t really understand

an ant, and that’s because

it’s a complex system, in other words it doesn’t yield to these

elegant mathematical formalisms. And the question is, how

did a complex system emerge out of a simple system? So that's

in some sense the big question. How did life, some people would say,

emerge from non-life? And

that’s what I do, and I do it in the context, I guess, of

natural science, I’m not a mathematician, I’m not a computer scientist,

I use tools from those fields, and I apply them to natural phenomena, for which we

have reasonably good data, and increasingly better data.

Can you give an example?

Yes, of course. So for example, over the course of the last few decades,

the extraordinary increase in the number of complete genomes

we have from a very large range of organisms. It used to be that we

only had that data for one or two, because it was so expensive to

collect. And now it’s not so if you’re asking questions about

evolution you clearly need data from very diverse

species, not all from one group, not all from humans or mice

or yeast, or flies, and

so that’s made a big difference, so we can now ask about patterns

of change, infer patterns of change from

very very diverse tips of the tree of life, and I think that’s made

a big difference, and along with that data is data

about things like, all of the proteins we find on nerve cells

so we can ask questions about changes in the way nerve cells

express genes across diverse lineages, and when

so that’s the kind of data, I think largely genetic,

that’s been made available via these high frequent mechanisms

that I think really complements this sort of more basic theoretical

research program. So just a simple question

when you look at these genomes and

look at changes in the genomes and so on, how

does information processing come in? You mentioned that as fundamental -- That’s a very good question.

In fact, that is the great challenge. So

what you have to do, yes, is commit to

a model of computation and then ask to what extent

the elements that you observe fulfill

the requirements of that model, so let me give you an example, a simple example.

After the discovery of

the structure of DNA, there was a lot of interest in how genes are turned on

and off, and this is often, this work is often associated with Jacob and Monod

who are interested in what is now called gene regulation

how genes turn on and off at different times. We can take one organism that is living

like an E. coli bacterium, and we can study in great detail

how these genes are turned on and off, so we know

what the structure to function mapping is, because we have a living species.

And now we can go back and look at a whole range of

genomes, for which we do not have laboratory experiments

generalize and say, since they share those structures,

it’s probably fair to assume that they have similar mechanisms, so that’s what we do

we try to write down simple mathematical models for systems

that we know, and if we find analogous structure

in species for which we do not have the capability to

perform experiments, ask whether the mathematical models

have all the necessary ingredients, and if not, try to make some

conjectures about what’s missing or what they've added.

Okay, interesting. Let me ask one last question, which is

you’ve talked about all these different areas, like information processing

computation, mathematics, dynamics

and so on. So if someone wanted to become a complex

systems scientist, how do they master all of these different fields?

How can you do it? Yeah, that is a good question.

I’m not sure that we do.

Well, so the first answer is that we don't.

Collaboration is key. And in fact that’s one of the

really fun things about working in this field, because it really requires

deep collaborations, and that’s

part of what makes it so much fun.

The other thing, of course, is just to

throughout your career, not to lose track of other fields. I think

that the human brain is remarkably flexible, and remarkably

capacious and the way

that our universities are structured sort of denies that fact, and says

you have to specialize because you can’t make a contribution unless you

only know about this group of proteins, and

I defy that convention. I think that

it’s not true. I think that people

when interested can learn huge amounts, and there’s really an issue of time

and structure, and I think part of the

huge benefits of your course, of something like this

making this information available online, is that people will have access to

this information. So I think

as a director of an institute, I go out of my way to

try to promote seminar series, and courses, that have this kind of

transdisciplinary quality. I think we have

to find ways of rewarding people for pursuing this kind of work. I think

we have to tenure people who work across disciplines. Yeah, that’s the hard

challenge. There are some significant challenges ahead, but

they’re just practical challenges, they’re not logical impossibilities.

Well great. Thank you so much. This has been great.

And I hope you don’t freeze too much in Wisconsin.

I won’t. I have many, many layers.

Alright. Thanks a lot. Thank you.