►
From YouTube: DevoWorm (2022, Meeting 5): Topical Follow-up, Hypergraphs for Development, Popgen of Daisyworld
Description
Follow-up on recent topics (Ising Models, Tensegrity, Molecular Modeling, Wolfram Patterns in Biological Pattern Formation). Hypergraphs for Developmental Phenotypes (cell and tissue differentiation), the Cybernetics and Population Genetics of Daisyworld and biologically-rigorous Gaia models. Attendees: Mainak Deb, Richard Gordon, Karan Lohaan, Susan Crawford-Young, Jesse Parent, and Bradly Alicea.
A
I'm
pretty
good
I'm
going
on
vacation
next
week,
so
I'm
busy
because
I'm
trying
to
sort
of
get
ready
for.
B
A
So
anyway,
so
monday,
monday
morning,
I'm
probably
well,
I
might
be
able
to
make
the
meeting
next
week.
I'm
not
sure.
C
All
right
yeah,
if
you
can
yeah
well
good
luck
with
your
vacation,
you
seem
like
you've,
been
working
very
hard
to
deserve
it.
A
Math,
like
they
did
finite,
element
analysis
and
found
the
equation
for
the
stiffness
matrix,
and
so
I
said,
find
the
one
for
10
seconds
and
apparently
it's
called
what.
D
D
Going
to
start
there
other
than
that,
even
I
think
dr
gordon
had
given
that
hydrophobicity.
So.
D
D
D
A
D
Such
a
way
that
you
know
I
can
make,
I
can
at
least
give
it
much
more
better
contrast
like
these
are
the
contrasting.
A
I
really
need
to
to
get
some
new
images.
A
A
Photogrammetry
might
help
give
a
3d
image
of
of
the
salamander
eggs
if
you've
got
all
sides
of
it
and,
like
I
said,
the
new
microscope
probably
is
what
I
need
to
use
to
get
some
new
images,
because
it
has
it's.
The
angle
doesn't
change
the
angle
of
the
images
doesn't
change,
so
I
think
it's
going
to
be
better,
but
so.
E
A
D
D
D
A
Yeah,
so
we'll
we'll
have
to
give
them
pep
pills
or
something
the
sound
matters.
I
don't
know
what
to
do
and
by
the
way
that
my
finite
element
analysis
and
the
stiffness
matrix.
I
found
what
it's
called.
It's
called
the
flexibility
matrix
when
it
when
it's
in
cantegrity.
A
A
C
C
A
A
Are
in
tension
and
that's
why
they're
called
a
tensegrity
structure?
They
are
already
in
tension
and
they're,
so
they're
they're
pulling
and
the
compression
elements
are
the
rods.
Okay,.
F
Now
the
analog
analogy
would
be
that
the
microtubules
presumably
are
much
stiffer
than
microfilaments.
The
microphones
are
contractual,
so
the
microfilaments
would
would
be
conceptually
represented
by
the
elastic
compounds.
The
problem
is
on
on
the
real
tensegrity
toy,
the
elastic
components
decompose
after
20
years.
C
A
But
there's
there's
youtube
videos
about
this
people
building
yeah.
C
F
A
A
A
C
Oh
I'm
doing
just
fine
good
okay.
So
that's
let's
see!
Oh
richard
dick
asked
karan
interested
in
mixed
chorality
problem.
I
guess
that's
something
you
discussed
before
or.
G
F
Predicting
the
hydrophobicity.
F
A
F
F
F
Yeah
and
it's
not
they're
not
from
racemic
mixtures,
because
the
the
preponderance
of
evidence
seems
to
me
that
the
amino
acids
from
meteorites,
which
are
presumed
to
be
the
ones
that
the
origin
of
life
are
biased
towards
l
compared
to
the
amino
acids.
But
it's
a
it's
an
incomplete
bias,
so
that
could
be
included
in
the
simulation.
So
you
get
more
l
d.
F
Okay
and
then
there
are
other
weird
phenomena
like
once.
You
start
an
l
alpha
helix,
it
seems
to
be
biased,
so
it
will
propagate
its
own
kind
and
there
are
some
subtle
differences
between
l
d,
which
might
also
introduce
a
bias.
F
G
F
The
the
problem,
the
problem
is
nasty
because,
even
if
you
only
confine
yourself
to
l
amino
acids,
the
prediction
compared
to
experiments
are
often
wrong.
F
Detailed
molecular
dynamics,
okay,.
F
Okay,
because
at
least
membranes
that
exist
now
are
chiral
the
the
lipids
or
the
phospholipids
that
can
be
chiral.
A
C
Sounds
like
an
interesting
problem,
so
I'm
gonna
share
my
screen
for
a
minute.
I
think
we
talked
about
a
couple
weeks
ago.
We
talked
about
ising
models
and
I
don't
know
where
we
are
on
that,
but
we're
in
a
state
of
inertia,
okay,
yeah.
C
F
C
C
So
this
is
see
if
I
can
zoom
in
here
a
bit
more.
F
G
F
G
C
Full
resolution
through
the,
but
so
this
is
the
ising
model,
india
for
physicist
ernst,
eising
and
wilhelm
lens-
is
a
mathematical
model
of
ferromagnetism
and
statistical
mechanics.
So
this
is
something
that
you
know:
people
develop
models
and
statistical
mechanics.
C
We
often
borrow
them
for
biology,
especially
if
it's
it's
you
know
biophysics,
but
also
people
use
it
for
more
ambitious.
You
know
types
of
systems
like
you
know,
traffic
systems
and
things
like
that.
You
know
people
will
use
statistical,
mechanics,
sort
of
as
the
basis
for
their
model.
So
there's
a
lot
of
good
mathematics
and
statistical
mechanics.
This
is
one
sort
of
model
in
that
area,
and
so
this
this
particular
model.
C
It
consists
of
discrete
variables
that
represent
magnetic
dipole
moments
of
atomic
spins,
so
you're
looking
at
things
that
have
these
spins
and
they
give
me
one
of
two
states,
so
this
is
plus
one
and
negative
one.
So
this
is
very
similar
to
a
binary
model
in
a
computer
science
simulator,
a
computer
science
model
of
a
bit
where
we
have
two
states
plus
one
well,
we
either
actually
have
zero
or
plus
one.
This
is
plus
one
or
negative
one,
and
it's
usually
the
polarity.
C
These
spins
are
arranged
in
a
graph,
usually
a
lattice
and
then
allowing
each
spin
to
interact
with
its
neighbors
and
then
the
neighboring
spins
that
agree
have
a
lower
energy
than
those
that
disagree,
so
they
minimize
their
energy.
If
they
agree,
the
system
tends
to
the
lowest
energy,
but
it
disturbs
its
tendency.
So
you
have,
you
can
apply
heat,
you
know
at
some
analog
of
heat
or
you
can
have
this,
this
sort
of
agreement
between
the
states-
and
this
gives
you
this
lowest
energy,
and
so
you
look
at
the
lowest
energy
states.
C
So
this
is
basically
the
approach,
and
so
they
kind
of
go
through
a
lot
of
the
mathematics
of
it.
So
this
is
yeah.
So,
let's
see,
if
there's
anything,
I
guess
there's
a
their
connections
to
graph
theory,
which
I
found
interesting
and
there
are
other
connections.
You
know
there
may
be
connections
to
machine
learning
here
as
well,
especially
with
respect
to
energy
minimization.
C
Now
there
are
some
people
who
are
doing
paper
or
I've
seen
some
papers
where
people
are
kind
of
starting
to
make
the
connection
between
like
energy,
minimization
and
statistical
mechanics
and
like
gradient
descent
in
machine
learning.
Basically,
the
gradient
descent
is
minimization
of
state
minimization
of
energy.
Well,
as
they
define
it
in,
say,
a
machine
learning
model
of
error
and
then
there's
this
minimization
of
energy
as
well.
On
statistical
mechanics
and,
like
I
said
last
week,
we
talked
about
fitness
landscapes,
that's
a
very
similar
model
where
you
have
this.
C
In
this
case
you
might
be
maximizing
fitness,
and
so
it's
just
flipping
that
over
and
saying
what's
the
maximal
fitness,
so
you
know
you're,
looking
at
a
bunch
of
states
which
are
these
individuals
of
the
population
and
the
population,
not
the
entire
population
is
maximizing
its
fitness
and
then
that's
your
distribution
of
of
of
states.
You
know
whether
they're
maximized
or
not-
and
so
that's
it's
a
very
similar
sort
of
thing
across
all
those
areas.
C
People
are
using
those
kind
of
similar
they're
using
you
know,
mathematical
models,
but
they're,
also
using
this
broader
set
of
analogies,
and
so,
but
I
just
wanted
to
point
that
out
for
people
just
maybe
to
get
people
interested
in
some
of
these
topics
and
then
seeing
what
the
connections
are.
I
know
we
didn't
get
have
a
lot
of
time
last
week
to
talk
about
the
connections
between.
You
know
different
areas.
I
mentioned
it
briefly,
but
then
they
have
this
thing
about
money,
carl
methods.
C
So
these
this
you
know
quenching
approach
where
they
heat
the
system
up
and
then
they
cool
it
off.
So
it's
quenching.
You
can
use
that
as
well.
People
use
that
in
simulated
annealing,
which
is
a
computer
science
technique
which
is
somewhat
similar
to
this,
where
you,
you
know,
apply
heat,
which
is
basically
some.
C
C
F
Here,
yeah
eisen
came
up
with
that
around
1930
in
a
one-dimensional,
less
okay
yeah.
He.
This
is,
of
course,
before
computers,
he
presumed
that
that
would
represent
the
phase
transition
in,
for
I
guess,
that's
paramagnetic
magnetism
and
he
was
raw
okay,
it
did
not
produce
a
clean,
sharp
phase
transition.
F
Then
later
somebody
discovered
that
if
you
make
it
two-dimensional,
you
do
get
a
sharp
face
transition
and
then
the
excitement
does.
It
became
the
first
simple
model
for
a
phase
transitional
material.
Okay.
Now
it's
also
applicable
to
what's
called
a
lattice
gas.
F
A
lattice
gas
means
simply
the
zeros
and
ones
represent
empty
and
occupied
of
some
molecule
on
some
lattice
same
kind
of
interactions
between
neighbors
okay.
So
I
got
involved
with
this
in
the
1960s
in
terms
of
glass
gases.
F
That's!
That's
where
my
interest
comes
from
okay.
Now,
if
you
go
back
and
take
the
one
dimensional
model
and
make
it
two-dimensional
by
making
the
other
dimension
the
states
versus
time,
then
you
get
something
akin
to
a
wolf
of
that,
and
this
has
raised
questions
like
what's
the
relationship
between
such
a
propagated,
one-dimensional,
hyzing,
lattice
gas
model
and
wolfram
patterns,
especially
the
ones
that
are
claimed
to
be
chaotic,
okay,
okay
and
the
other
question
which
came
up
which
was
kind
of
interesting.
F
F
F
A
F
Like
yeah
go
from
pattern
in
a
search
for
images.
A
C
So
this
this
is
well
I'm
going
to
share
my
screen.
First
yeah.
Okay,
so
I
think
like
something
like
this
is
a
wolfram.
This
is
rule
110,
and
this
is
something
that's
generated
by
this
simple
rule
and
it's
just
applied
recursively
and
you
and
it's
it's
like
basically
forming
your
triangle
and
then
you
just
keep
applying
this
and
you
get
this
at
edge
here
where
you
have
larger
triangles
and
you
go
down
to
finer
and
finer
scales.
F
There
are
256
sets
of
rules,
you
pick
one
of
them
and
you
just
propagate
that
rule
deterministically,
and
this
is
the
pattern
you
get
each
horizontal
line
being
a
time
step
as
you
go
down
the
page.
The
time
is
zero
at
the
top
and
goes
down
yeah.
Okay.
Now
the
the
biological
interest
in
this
has
mostly
been
that
some
of
the
patterns
look
like
some
of
the
snail
patterns
right
but
nobody's
ever
tested
that
quantitatively.
C
F
C
C
H
H
C
All
right,
let's
see,
we
have
something
in
the
chat.
Oh
okay,
jesse
says
hi.
I
won't
be
able
to
speak
much,
but
just
checking
in
so
hello.
We
haven't
seen
him
since
the
new
year,
so
it
looks
like
my
knock
is
back
as
well,
so
good
so
yeah.
We
were
talking
about
these
sorts
of
things.
So
if
if
people
are
interested
in
this
topic,
you
know
by
all
means
you
know
we
can
discuss
this
further.
Maybe
someone
can
present.
F
To
do
a
little
similar
question
is
the
the
literature
you
did:
a
survey
of
one
dimensionalizing
models
and
came
up
with
800
papers.
G
F
So
clearly,
we
need
to
find
reviews.
We
need
to
do
a
sampling,
not
all
800.,
yeah,
okay,
and
we
might
want
to
make
up
some
rules
nobody's
looked
at.
F
F
Okay,
even
though
the
originalizing
one-dimensional
icing
hollow
was
not
did
not
give
a
clean
fight,
you
know
what
phase
transition,
that's
overlapping,
so.
F
And
the
regularizing
model,
if
you
make
it
two-dimensional,
it
does
give
you
a
first-order
basis.
F
Sublimation
would
be
analogous
to
absorption
desorption
onto
a
surface
and
I've
simulated
that
long
time
ago,
but
I
think
that
was
first
shorter.
A
Oh
okay,
and
by
the
way,
what
is
rule
30?
Oh,
what
is
rule
30.
C
I
actually
had
110.
I
think
I
can
go
back
to
30
here
so
rule
30.
Is
this
go
back
to
the
search
here?
Oh,
why
is
it
doing
that.
C
C
C
And
how
is
it
applied
so
this?
I
think
this
is
the
rule
set
here
in
olive
or
firm's
elementary
cellular
or
automata,
an
infinite,
one-dimensional
array.
Only
two
states
is
considered,
then
this
is
the
sort
of
the
truth
table
for
the
different
patterns.
So
for
current
pattern,
one
one
one,
you
get
a
new
state
zero.
C
F
Let's
do
the
two
to
the
fourth
states
which
are
applied,
and
how
do
you?
How
do
you
pick
what's
the
rule,
whether
you
go.
C
Zero
one
right,
yeah
yeah,
so
each
of
these,
each
of
these
string
of
three
binary
numbers
generates
a
new
state.
So
if
I
guess,
if
you
have
a
one
and
a
well,
I
guess
it
would
be
like
two
ones
in
in
the
state:
it
generates
a
zero.
C
If
you
have
three
zeros,
it
generates
a
zero.
I
don't
know
what,
if
that's
a
rule,
but
does
basically
it's
generating
these
different,
so
this
is
basically
like
zero
through
seven,
if
you
think
about
this
in
in
decimal.
So
this
is
zero.
This
is
one
two
three
four
five,
six
seven
and
so,
and
these
are
just
for
the
different
cells
surrounding
the
center
cell,
so
this
could
be
in
time.
C
I
think
this
is
like
a
time
step
previous
and
then
it
generates
a
new
state
in
this
center
cell,
so
you'd
have
like
in
the
neighborhood
here.
You
have,
I
think,
basically,
how
many
of
these
either
it's
like
the
state
of
the
different
cells
around
it
or
how
many
of
them
are
of
state
one.
I'm
not
really
sure
how
they
count
this,
but
so
the
corresponding
formula
is
left
cell
exclusive
or
central
cell
or
right
cell.
C
So
that's
the
rule
in
binary.
You
have
this
0
0
0,
1,
1,
1,
1,
0,
2
and
then
30.,
so
it
equals
30..
So
it's
a
basically
they're
using
binary
to
calculate
the
this.
A
C
What
30
is
supposed
to
mean?
So
I
don't
know
I
might
do
a
longer
a
deeper
dive
into
this.
It
might
be
useful
to
find,
like
you
know,
interest,
because
there
are
a
lot
of
different
rules
and
just
to
kind
of
say
like
what
is
this.
You
know
it
would
be
interesting.
Here
is
if
we
could
test
it
out
on
different
things,
not
just
these
simple
patterns
but
think
about
like
can
we
create
like
a
case
where
you
have
to
be
like
a
three-dimensional
cellular
automata
or
what
are
the
consequences
of
these
rules?
C
I
don't
I
don't
know
if
people
are
into
doing
like
some
sort
of
mathematical
analysis
of
this,
but
that
might
be
something
interesting
if
we
could
just
explore
this
and
pull
it
apart
and
see
what
it's
doing,
because
I'm
not
entirely
clear
on
what
it's
doing,
but
it
might
be
useful
because
we
had
this,
so
we
had
this
paper
a
couple
weeks
ago
where
we
had
where
they
talked
about
seashell
patterns,
which
are
these
things
up
here,
and
they
talked
about
how
they're
generated
by
some
neural
mechanism.
C
So
there's
this
secretion
system-
that's
based
on
some
set
of
neurons
in
the
snail
and
they
generate
this
pattern
with
this
with
this
secretion
system,
but
of
course
it
has
to
be
put
there
in
a
way.
That's
you
know
based
on
some
sort
of
pattern
formation,
and
so
that's
maybe
where
rule
30
comes
in
or
as
dick
says,
maybe
it
rule
30
doesn't
have
a
lot
to
do
with
it.
Maybe
it's
just
that
it
looks
very
similar
to
it.
F
C
C
C
You
have
something
in
the
chat.
Okay,
my
next
says:
roll
30
is
super
interesting,
interesting
video,
where
stephen
wolfram
talks
about
rule
30
and
what
makes
it
special.
So
this
youtube
video
on
the
chat
is
something
that
you
might
want
to
check
out
so
yeah.
If
people
want
to
follow
up
on
that,
please
let
me
know,
but
I
might
follow
up
on
myself
actually
we'll
see
it
might
lead
to
something
interesting.
C
So
I'm
gonna
switch
gears,
and
maybe
it
isn't
switching
gears
so
much
but
and
talk
a
little
bit
about
an
abstract
that
I
want
to
submit
to
netsize.
So
we
have
this
conference
every
year
called
nutsai
and
we
had
it
last
year.
I
don't
think
I
have
a
link
to
it,
but
this
is
something
that's
being
held
virtually
again
and
I've
been
working
on
this
idea
for
a
while.
So
this
is
c
elegans.
C
C
It's
a
little
too
fast
to
really
follow
it,
but
you
can
see
that
it
starts
from
a
couple
cells
they
divide
and
then
they
form
this
comma,
it's
kind
of
shape,
it's
changing
its
shape
and
then
it's
twisting
around
like
a
pretzel
and
then
it's
forming
this
worm
here
which
hatches
out
of
the
egg
and
then
it
becomes
a
larval
work.
So
this
is.
This
is
what's
happening
in
the
c
elegans
embryo.
Okay,
this
all
unfolds
according
to
this
lineage
tree.
So
we
have
this
lineage
tree,
that's
what
they
call
deterministic.
C
A
G
A
G
C
C
Times
because
the
cell,
the
cells
are
moving
around
with
respect
to
the
anatomy
and
then
they
shift
around.
So
you
know
this:
a
b
is
like
at
the
front
of
the
arm
at
the
anterior
around
the
p1
is
at
the
posterior
end,
but
these
a
b
cells
actually
form
all
sorts
of
tissue
type
cells
that
represent
tissue
types
and
they
tend
to
be
in
in
different
parts
of
the
worm,
so
they
have
to
move
around.
So
you
see
there's
a
lot
of
switching
in
this
tree
from
the
straight
lineage
to
these
different
places.
C
So
there's
this
process
of
sorting
going
on
in
the
anatomy.
So
one
way
to
deal
with
that
is
to
look
at
like-
and
I
mentioned
this
before.
We
have
the
embryo
networks,
where
we
have
all
the
cells
that
we
can
look
at
the
you
know
use
like
the
centroids
as
markers
or
something
else,
and
we
can
build
a
network,
a
proximity
network
that
mimics,
perhaps
cell
signaling
or
other
types
of
substructure
within
this
embryo.
So
you
have
like
different
cells
that
are
forming
and
as
they're
moving
around
and
as
they're,
you
know
forming
tissues.
C
They
have
different,
maybe
exhibit
different
motifs
in
that
network
structure.
So
this
is
the
a
like
a
a
tree
of
heredity,
but
what
I'm
talking
about
is
a
network
of
proximity
more
than
heredity.
So
if
you
combine
those
two
what's
what's
that
going
to
look
like,
and
so
this
example
here
is
just
from
like
it's
just
a
drawing
that
people
have
made
based
on
like
the
relative
physician
and
the
adult
worm,
and
then
this
hereditary
unfolding
here.
C
So
there's
no
real,
it's
not
quantitative
at
all.
Really
I
mean
it's
just
kind
of
a
drawing,
so
I've
been
working
on
this
idea
of
using
a
hyper
graph
to
characterize
this
or
what
I'd
call
sometimes
a
hyperlineage
tree,
and
so
I
think
I
presented
this
before.
I
don't
know
if
I
presented
I
presented,
I
think
a
drawing
of
it,
but
I
worked
it
out
a
little
bit
here.
So.
C
Basically,
a
computational
model
of
this
process,
so
you
have
this
tree
where
you
start
from
a
single
cell.
You
move
to
two
cells,
the
four
cells
and
then
at
the
four
cell.
You
start
to
get
differentiation
by
function.
You
get
seven
cells
that
are
in
this
eight
cell
embryo
that
are
somatic
cells,
and
then
you
get
one
cell,
which
is
a
germ
cell,
and
then
you
follow
this
out
down
this
germ
line
and
you
start
to
get
more
and
more
germ
cells.
C
A
C
C
Four
cell
is
kind
of
trivial
as
well,
and
then
you
start
to
get
to
like
12
and
24
cells.
When
you
get
maybe
like
networks
that
are
interesting,
but
in
the
germ
line
we
don't
really
think
that
there's
any
sort
of
reason
to
have
well.
Maybe
there
is
a
reason
to
think
that
proximity
plays
a
role
in
in
the
germline
we
don't
really
know,
but
the
idea
here
is
that
you
have
these
colored
circles
that
represent
different,
like
sort
of
parts
of
our
sub
graphs
of
this
main
graph.
C
So
this
embryo
is
going
as
it
differentiates
it's
going
to
have
like
a
full
graph
of
proximity
and
then
subgraphs.
So
when
it
gets
to
this
comma
stage,
for
example,
you
know
this
tail
end
is
going
to
have
a
sub
graph.
This
other
part
is
going
to
have
a
sub
graph
or,
as
you
start
to
get
tissues
forming.
You
know
there
are
areas
of
the
anatomy
that
might
have
subgraphs,
like
muscle
area,
where
you
have
muscle
you
have
different.
C
In
this
case
you
have
these
subgraphs.
So
this
these
white
circles
are
the
connectome.
The
neural
connectom,
these
gr
or
orange
patches
are
embryo
networks.
So
this
is
like
the
all
the
cells
that
are
somatic
cells,
that
aren't
neuroconnectome
and
then
these
yellow
circles
are
the
germline
and
they
all
have
their
own
subgraphs
and
they
all
connect
to
one
another.
C
So,
for
example,
here
you
have
an
embryo
network
of
194
cell
or
you
have
194
cells
in
this
subgraph
50.
In
the
subgraph
for
the
next
step
forward,
you
have
cells
that
join
this
subgraph
because
they
differentiate
into
neurons
or
this
subgraph,
which
is
26,
which
is
at
a
slightly
different
time
step.
These
are
actually
where
the
neural
connectome
starts
to
differentiate
actually
into
different
sub
to
different
modules,
and
so
the
undifferentiated
cells
will
join
these
subgraphs
over
time.
C
You
can
have
exchange
from
like
one
generalized
subgraph
to
a
specialized
subgraph
as
the
cells
differentiate
and
then
they're
all
essentially
connected
in
some
way,
but
we
try
to
break
them
out
into
these
lineages
to
show
kind
of
functional
differences,
and
this
difference.
This
is
what
I
call
the
density
bifurcation
model,
which
is
basically
a
model,
that's
similar
to
what
they
consider
in
graph
theory
as
preferential
attachment,
but
it's
a
little
bit
different
than
that,
because
you
can't
use
that
model
for
a
number
of
reasons.
C
People
do
use
it
in
looking
at
connectomes,
but
this
is
a
different
way
of
looking
at
it
and
so
and
then,
ultimately,
you
know
we'd
like
to
be
able
to
model
these
types
of
embryos
where
you
have
processes
going
on
in
development,
not
c
elegans,
but,
like
you
know,
you
have
drosophila
where
you
have
some
of
these
folding
movements
or
other
types
of
movements,
spatially,
localized
differentiation,
that's
more
distinct.
They
can
really
generate
some
of
these
subgraphs
that
really
have
different.
C
You
know
they
really
have
some
interesting
properties,
so
this
one
is
based
loosely
on
c
elegans.
It's
not!
You
know
it's
not
exactly
mapped
to
c
elegans,
but
it
basically
describes
that
that
process
and
drosophila
would
look
far
different
than
this.
It
would
have
a
different
sort
of
topology
and-
and
you
know
how
these
trees
branch,
how
the
subgraphs
look
and
all
of
that
so.
G
F
Okay,
so
the
question
is
given
a
hyper
lattice
like
yours:
can
you
interfere
with
it
and
predict
what
that.
C
C
C
A
C
Have
and
if
you
have
to
leave
at
10,
that's
five
or
at
the
top
of
the
hour,
that's
fine,
but
I'm
gonna,
I'm
gonna
get
into
the
let's
see
what
we
have
here
today.
So
dick
sent
me
this
paper
this
a
couple
days
ago.
I
guess-
and
so
this
is
called
the
time-
complexity
of
self-assembly
and
zoom
in
here
to
the
so.
This
is
self-assembly.
C
A
C
The
idea
of
self-assembly
is
where
things
kind
of
come
together
and
form
something
on
their
own,
and
the
idea
is
there's
a
physics
behind
that,
and
you
know
we
don't
really
know
how
we
don't
really
know
how
it
works,
but
people
have
tried
to
figure
it
out,
so
this
paper
talks
about
the
time
complexity
of
this
process,
so
this
is
a
physical
process,
so
we're
talking
about
you,
know
energy
landscapes
and
there's
stat
mech,
that's
involved
in
this
just
so
we
know
where
we
are
so
time.
C
C
Moreover,
with
the
advances
of
nanotechnology
time,
efficiency
and
artificial
self-assembly
becomes
even
more
important
so
being
able
to
do
this.
This
is
now
an
artificial
system
and
in
nanotechnology
they're
very
interested
in
how
do
you
assemble
pieces
of
things
into
one
sort
of
machine
or
something
like
that?
So
people
are
building
machines
at
a
very
small
scale.
Sometimes
it's
at
the
nanometer
scale,
sometimes
that's
a
little
bit
bigger,
but
the
idea
is:
how
do
you
direct
this
process
because
it's
very
hard
to
manufacture
things
at
that
process
according
to
specifications?
C
C
Kinetic
aspects
concerning
the
time
efficiency
remain
much
more
elusive,
so
this
is
dealing
with
kinetics,
which
is
this
energy,
this
energy
and
movement
thing
so
in
computer
science.
The
concept
of
time
complexity
is
used
to
characterize
the
efficiency
of
an
algorithm
and
describes
how
the
algorithm's
runtime
depends
on
the
size
of
the
input
data.
C
So
this
is
something
that
I've
actually
got
done
with
a
paper
with
a
collaborator
where
we
talked
about
time
complexity
and
there's
this
oh
big,
o
notation
that
they
use.
Basically,
it's
this
idea
that
your
algorithm
runs
at
a
certain.
Has
a
certain
run
time
or
some
problem
is
a
certain
requires,
a
certain
amount
of
search
time
to
find
a
reasonable
answer.
C
So,
like
there's
some
problems,
if
you've
ever
heard
of
np
hard
or
mp
complete
those
terms
refer
to
the
time,
complexity
and
it's
you
know
usually
non-polynomial,
but
there
are
other
complexities
that
are
much
larger
than
non-polynomial
and
they're
problems
that
are
much
much
bigger
than
that.
So
we,
our
computers,
can
solve
things
like,
maybe
getting
into
np
hard
problems
but
they're.
C
You
know
np
complete
problems
which
are
hard
to
solve
and
then
other
problems
which
are
even
harder
to
solve
with
our
computers
and
if
you
think
about
like
a
self-assembly
process,
you
know
how
do
you
overcome
that
time,
complexity
with
a
natural
process?
So
that's
what
they're
trying
to
characterize
here
here.
They
characterize
the
time
complexity
of
non-equilibrium
self-assembly
processes
by
exploring
how
the
time
required
to
realize
a
certain
substantial
yield
of
a
given
target
structure
scales
with
its
size.
So
this
is
their
kind
of
where
they're
going
with
this.
C
C
What
is
the
process
to
build
this
thing
and
then
what's
the
time,
complexity
or
how
long
does
it
take
to
unfold,
and
then
they
show
that
this
exhibits
drastically
different
degrees
of
complexity.
So,
as
with
algorithms,
you
see
this
complexity
of
different
problem
domains.
Our
analysis
enables
us
to
identify
optimal
control
strategies
for
non-equilibrium
self-assembly
processes.
C
Furthermore,
we
suggest
an
efficient,
irreversible
scheme
for
the
artificial
self-assembly
of
nanostructures,
which
complements
a
state-of-the-art
approach
using
reversible,
binding
reactions
and
requires
no
fine-tuning
of
binding
energies,
so
they're
interested
in
this
process
of
sort
of
trying
to
find
ways
to
to
initiate
self-assembly,
but
also
how
long
it
takes
for
different
formulas
to
play
out.
So
in
nature,
you
see
virus
assembly
which
is
self-assembly.
C
It
must
be
fast
to
produce
the
many
virus
particles
before
the
infected
cell
was
eliminated
by
the
host's
immune
system.
So,
in
this
case,
this
is
something
that's
evolved
over
millions
of
years
where
viruses
will
invade
a
cell,
the
cell.
The
immune
system
will
respond
with
the
virus,
needs
to
outwit.
C
That
system
hit
the
replication
machinery
and
make
more
of
itself,
and
so
this
is,
you
know
something
that
is,
of
course,
a
process
that
has
to
be.
You
know
it's
it's
optimized,
of
course,
but
it's
also
something
that
requires
a
lot
of
self-assembly
within
the
cell.
It's
not
just
a
matter
of
you
know
two
steps
or
something
like
that.
A
C
There
are
also
larger,
complex
nanostructures
that
we
see
in
nature,
but
also
that
we
want
to
realize
for
technology
and
medical
applications,
and
so
this
time
efficiency
becomes
vital,
and
so
this
is
a
figure
figure
one
where
they
show
this
sort
of
scheme.
So
they
have
these
different
conditions
where
they
look
at.
C
They
have
and
identical
copies
of
s,
different
species
of
monomer
assembled
into
one
or
two-dimensional,
or
three-dimensional
structures
with
an
edge
length
of
l,
so
they're
measuring
the
edge
length
and
using
that
to
look
at
the
complexity
of
this
thing,
that's
assembled
then
they're
taking
these
monomers
and
they're,
putting
them
together
in
different
ways
and
doing
and
doing
so
they
time
everything
and
they
see
what
the
time
complexity
is.
So
some
of
these
take
a
while
to
assemble.
C
Some
of
these
are
very
easy
to
assemble,
and
so
then
they're
trying
to
characterize
that,
and
so
they
have
these
different
control
parameters,
diameterization
reversible,
binding,
just
in
sequence
activation.
So
there
are
all
these
different
things
you
have
to
account
for
in
this
process.
There's
an
activation
or
activation
energy
there's
attachment,
and
then
you
have
this.
You
know
you
have
a
complete
structure,
an
incomplete
structure,
so
sometimes
different
algorithms
will
result
in
these
incomplete
structures
which
aren't
really
viable.
C
C
So
we
talked
last
week
actually
about
about
like
mutational
hypercubes
and
the
idea
there
is
like
they
basically
take
something
like
an
rna
phenotype,
which
is
mostly
sequence,
there's
some
secondary
structure,
but
we're
interested
in
the
sequence,
and
the
idea
is,
if
you
mute,
you
know
there,
so
you
can
actually
look
at
mutational
steps
to
different
changes
or
different
phenotypes,
and
so
then
they
build
this
this
network
and
say
if
we
start
at
point
a
how
many
steps
does
it
take
to
get
to
point
b?
C
And
so
you,
you
measure
that
out,
and
you
can
tell
how
long,
because
you
know
how
long
what
a
mutation
rate
is,
how
long
it
takes
to
get
from
point
a
to
point
b
and
it's
a
similar
thing
here,
where
you
have
all
these
steps
that
you
have
to
follow
and
there
are
no
shortcuts.
Maybe
there
are
shortcuts,
but
you
want
to
characterize
that,
and
so
how
long
does
it
take
what's
the
time,
complexity
and
so
forth?
C
So
this
is
interesting,
looking
at
it
in
this
way
now
I
don't
see
any
notate,
the
the
conventional
computer
science
notation
for
this,
so
I
don't
know
what
they're
here's
where
they
plot
it
out.
So
this
is
like
complexity
and
time
based
on
structure
size.
So
in
computer
science
they
would
do
like
a
big
o
notation.
C
Exponential
times
are
faster,
so
in
this
case
you
have
some
of
these
scenarios,
where
they
map
out
the
time
based
on
the
structure
size,
and
they
give
you
this
idea
of
what
it
looks
like.
So
that's
that's,
basically
what
they're
trying
to
do
here
and
they
have
a
lot
of
different.
They
do
a
lot
of
different
types
of
tests.
So
I
think
that's
a
very
interesting
paper,
so
yeah
the
dimerization
scenario
turns
out
to
be
the
most
time
efficient
scenario
in
all
dimensions.
C
Reversible
binding
is
the
least
efficient
approach
to
assembling
large
linear
structures,
but
it
is
efficient
for
the
assembly
of
higher
dimensional
structures
and
then
becomes
competitive
with
the
gis
scenario,
slightly
outperforming
it
for
large
structure
sizes.
So
this
is
a
little
bit
different
than
in
computer
science.
Where
you
have
you
know,
your
algorithms
are
basically
just
like
you.
C
Maybe
abstract
you
know
the
problem
and
how
you
represent
it
in
this
case
you
have
these
things
that
are
actually
physical
things
that
you
want
to
do.
You
know
physical
parameters,
physical
objects,
that
you
need
to
optimize,
so
there
are
different
ways:
you
can
do
this
and
then
they
link
to
different
things
in
in
the
system.
So
it's
a
little
bit
different
than
the
computer
science
approach
and
as
such,
you
have
these
different
types
of
scenarios,
and
they
have.
You
know
it's
a
little
bit
different
than
just
like.
C
C
Let's
see
the
other
thing
I
want
to
talk
about
today.
Last
thing
I
want
to
talk
about
is
this
paper
on
a
fitness
based
analysis
of
daisy
world?
So
I
don't
know.
I
know.
C
We've
talked
about
daisy
world,
maybe
in
the
distant
past
in
this
group,
but
the
idea
of
daisy
world
is
where
you
have
this:
it's
a
model
of
an
ecological
set
of
ecological
regimes,
so
daisy
world
is
where
you
have
this
planet
and
the
planet
is
filled
with
daisies,
nothing
but
daisies,
and
that
represents
something
that
we
see
early
in
life
where
you
have
like
one
or
two
species
of
microbe,
that
kind
of
flourished
and
then
took
over
the
surface
of
the
planet.
So
you
know
it.
It's
like
you.
C
So
you
have
like
a
very
small
number
of
phenotypes
that
cover
this
planet,
and
so
the
idea
is
that
you
can
run
these
dynamical
simulations
that
show
what
happens
when
you
have
one
color
daisy
versus
two
colors
of
daisies.
The
population
proportion
fluctuates
over
time
and
you
can
actually
see
how
what's
growing
on
the
planet
can
regulate
the
atmosphere.
C
Atmosphere
and
see
how
the
those
different
morphologies
are
affecting
the
atmosphere
of
the
planet,
and
this
has
to
do
with
the
reflectance
of
the
different
colors
and
you
know
other
properties
of
the
daisies
that
that
are
populating
the
planet.
So
it's
a
toy
model,
it's
not
really
something
that
you
would
see
in
life,
but
it
gives
you
some
really
nice
insights.
The
problem
with
daisy
world
is
it's.
C
You
know
they
talk
about
it
in
in
cybernetics,
for
example,
there's
a
long
history
of
that,
but
in
in
biology,
especially
in
evolutionary
biology,
teoeology
is
kind
of
a
dirty
word
because
it
harkens
back
to
you,
know
creationism,
and
things
like
that,
where
you're
basically
saying
that
there
has
to
be
some
external
force
that
you
know
determines
the
flow
of
time.
Of
course
you
know
this
is
the
sort
of
thing
that
where,
when
you
come
up
with
something
you
know,
they're
they're
different
fields
and
they
have
different
sort
of
expectations
for
things.
C
That's
where
that
kind
of
comes
in
the
clashes.
So
the
idea
of
teleology
here
is
that
self-regulation
emerges
because
it
is
preordained
to
do
so.
That's
the
criticism
of
this
model-
and
this
is
you
know,
maybe
more
of
an
ecological
model,
but
the
idea
is
it's
just
kind
of
like
hitting
the
idea
of
evolution
just
right,
so
that
people
are
making
are
accusing
this
model
of
sort
of
promoting
this
idea
of
teleology.
So
that's
that's
the
problem
here.
C
So
the
daisy
world
parable
a
simple
mathematical
illustration
of
gaia,
went
some
way
to
addressing
these
critiques,
so
people
have
modeled
this
using
dynamical
systems,
theory
and
other
things
as
agent-based
models.
You
can
build
a
nice
agent-based
model
with
this,
but
despite
recent
success
in
incorporating
natural
selection,
when
people
have
done
this,
it
remains
a
widely
held
view
that
the
ideas
are
inconsistent
with
biological
principles.
H
C
C
That's
another
question,
so
that
would
be
another
problem.
Okay,
so
the
system
regulates
its
temperature
due
to
the
low-level
evolutionary
dynamics
of
competition,
so
between
the
thermally
coupled
daisies
and
no
higher
principle
is
invoked.
So
basically,
there
are
these
daisies
of
different
colors
and
in
different
states
and
they're
regulating
the
thermal
conditions
of
the
planet
and
then
that's
selecting
for
certain
daisies
over
others.
Then
they're
able
to
look
at
this
in
terms
of
so
population.
Genetics,
of
course,
is
this
idea
that
you
have
these
different
genetic
variants.
C
They
have
different
frequencies
in
the
population
and
that
does
change
with
selection
or
with
like
processes
like
random
processes
and
other
types
of
things.
C
That
we
can
see-
at
least
you
know
it
doesn't
just
say
that
there's
this
sort
of
you
know
tibiological
change
or
that
there's
some
other
thing
going
on
that
we
can't
explain
this
kind
of
roots
it
into
what
we
kind
of
know
is
biology,
so
a
reconciliation
of
gaia
with
evolutionary
theory,
male
1,
further
development
of
evolutionary
arguments
for
the
existence
of
global
self-regulatory
systems.
C
So
this
is
this
kind
of
goes
through.
The
the
idea
of
this
is
james.
Lovelock
proposed
this
idea
originally,
and
it's
interesting
because
lovelock
was
thinking
in
terms
of
life-induced
feedback
loops.
So
you
know
the
this
is
a
different
sort
of
approach
to
looking
at
life.
It's
not
necessarily
not
evolutionary.
C
It's
just
that
they're
using
a
different
way
to
look
at
it.
So
this
basically
shows
something.
It's
just
that
you
have
to
make
a
connection
with
the
you
know,
different
aspects
of
the
organism
more
and
so
only
through
stressing
the
unconscious
nature
of
this
regulation,
particularly
within
reference
to
daisy
world,
as
the
guy
hypothesis
been
able
to
refute
claims
of
teoliology.
C
So
people
have
done
work
on
this
to
refute
those
claims,
but
but
the
way
to
do
this,
of
course,
one
way
to
build
a
an
effective
model
of
daisy
world
and
to
demonstrate
it
is
to
build
a
spatial
model
where
you
have
these
daisies
distributed
in
space.
They
have
these
different
albedo
types,
which
albedo
is
the
reflectance
off
the
surface
and
light
surfaces
reflect
more
than
dark
surfaces
which
absorb
heat,
and
so
these
have
an
effect
on
the
overall
condition
of
the
atmosphere
in
this
model.
C
So
you
know
light
colored
daisies
will
reflect
heat
back
up
into
the
atmosphere,
dark,
colored
disease
will
absorb
heat
and
it
has
different
effects
on
the
on
the
model
of
the
planet,
and
so
it'll
it'll
also
affect
the
fitness
of
the
organ
of
the
daisies.
But,
of
course,
there's
no
fitness
imperative
in
the
initial
model.
C
It's
just
basically
the
change
of
the
phenotypes,
and
so
people
have
worked
on
this
they've
built
different
dynamical
systems
models
in
this
paper,
they're
building
a
model,
that's
based
in
population,
genetics,
and
so-
and
so
it's
you
know
it's
it's
very
kind
of
surprising
that
no
one's
done
this
before,
because
these
are
kind
of
easy
to
roll.
I
would
think
they'd
be
fairly
easy
to
relate,
but
this
is
something
that
people
have
kind
of
ignored,
and
so
you
basically
have
this
model
of
the
phenotypes.
C
You
have
this
model
of
population
genetics
where
you
model
a
lot
of
these
processes
going
on,
and
then
you
basically
link
them
together,
and
so
you
end
up
with
a
different
explanatory
framework.
Maybe
it's
more
acceptable
to
biologists,
maybe
not!
It
depends
on
kind
of
a
biologist.
You
are,
if
you're
an
ecologist.
You
probably
don't
care
as
much
perhaps,
but
this
is
the
so.
C
C
You
can
see
that
they've
been
able
to
build
that
they've
been
able
to
look
at
the
regulation
of
temperature
on
the
planet
and
the
albedo,
and
so
this
is
basically
yeah.
It's
a
nice
paper.
It
kind
of
goes
through
has
some
pretty
decent
mathematics
in
it.
They
do
a
lot
of
dynamical
systems
modeling
and
spatial
cellular
automata.
A
C
Different
approach
to
a
classical
problem,
as
with
our
first
paper,
so
we
have
quran
actually
pointed
out
that
they're
one
dimensional
cellular
automata
simulators.
He
put
some
links
in
the
chat
on
that,
and
so
you
might
want
to
check
those
out.
Those
are
actually
quite
interesting.
Daisy
world.
You
can
run
as
a
cellular
automata
as
well.
I
don't
I
don't
know
if
there
are
any
models
out
there
pre-made,
but
that's
another
option
for
making
these
kind
of
simple
toy
models
of
homeostatic
regulation.
C
Do
they
get
a
mix
of
phenotypes?
In
the
end,
I
I
haven't
really
look
at
the
paper
here.
Let
me
share
my
screen.
I'm
not
sure
that
they
have
like
a
table
of
what
they've
gotten.
F
C
All
right,
I
don't
know
if
they
actually
provide
a
good
map
of
that.
C
No
pictures
of
daisy
world
maybe
later
in
the
paper
here,
let's
see.
C
C
Yeah
so
yeah
they
kind
of
they
don't
really
talk
too
much
about
what
those
distributions
look
like
that's
another
thing
about
daisy
world.
Is
that
it's
a
you
know
one
of
these
models.
That's
sort
of
you
know
the
the
power
of
it
is
sort
of
the
intuition.
Like
you'll,
see
the
change
in
these
different
states.
You'll
see
okay.
I
understand
why
you
know
what
this.
What
this
means,
perhaps
or
you
know
what's
going
on-
is
that
you
have
basically
this
change
in
state.
Sometimes
you
can
maintain
multiple
states.
Sometimes
one
dot
predominates.
F
H
C
All
right
any
there
are
no
more
questions,
let's
wrap
it
up
for
today,
thanks
for
attending.
If
you
want
to
present
anything,
if
you
want
to
follow
up
on
anything,
please
let
me
know
or
bring
it
to
the
meeting
and
we'll
it
can
give
you
time
to
do
it.