►
Description
Tom Portegys demonstrates how to understand and use the Morphozoic software package. Bradly Alicea moderates.
A
Tutorial
and
we're
going
to
post
this,
we
couldn't
get
it
live,
but
we're
going
to
post
it
later
so
right
now
we
have
tom
portage,
ease
and,
and
I
mentioned
about
more
fizzling,
the
flash
talk
and
I
didn't
really
do
it.
Justice,
so
he's
going
to
show
us
a
little
bit
about
morfa,
Zoe,
so
Tom
take
it
away.
B
So
what
is
morphogenesis
well
and
also
I
was
this
related
to
D,
but
where
else
related
to
deworm,
because
it's
basically
devo
worms,
rationale
for
being
I,
won't
try
and
say
that
in
french
is
how
C
elegans
develops
from
a
single
egg
to
an
organism,
fully
fledged
organism
so
morphogenic.
This
is
basically
scraped
out
of
wikipedia
it's
a
two-part
word.
It
means
the
greek
word.
Morris
means
shape
and
Genesis
needs
creation.
B
B
B
Influence
so
we
want
to
have
some
kind
of
we
want
to
be
able
to
incorporate
both
these
things,
local
and
global,
and
also
we
want
to
buy.
We
want
a
model,
that's
actually
not
going
to
go
off
the
rails
computationally.
So
this
one
does
that
by
basically
aggregating
cell
values
and
what
it
does
is.
It
allows
the
thing
to
actually
become
more
complex
in
a
linearly
growing
faster,
not
geometrically.
Okay,.
B
So
that
time
time,
so
this
is
a
very
fundamental
computer
science,
computing
structure
and
it's
very
simple,
simple:
it's
a
two-dimensional
grid,
there's
variations
that
could
be
3d,
but
the
most
common
one
has
a
2
t's
grid
of
cells.
So
each
cell
has
a
state
which
is
just
a
number
represent
anything
we're
also
going
to
call
these
states
types
later
on,
as
a
state
taught
as
a
cell
type.
B
So
each
cell,
and
not
only
knows
what
its
own
state
is,
but
it
knows
what
the
states
of
its
neighbors
are
well,
what
is
it,
what
are
neighbors?
Well,
a
very
typical
neighborhood,
as
it's
called
defining.
What
it
sells,
neighbors
are,
is
three
by
three
a
set
of
cells
surrounding
a
cell
in
a
grid
and
that's
called
a
more
neighborhood
a
nacelle
states,
so
a
cell
can
change
states
and
it
will
change
states
depending
on
what
its
own
state
is,
what
the
states
of
its
neighbors.
A
B
Okay,
probably
the
most
common
instance
of
sonar
tomica
is
Conway's
Game
of
Life,
so
you
can
see,
there's
a
grid,
it
has.
It
has
black
and
white
cells
on
it.
The
black
ones
are
called
a
live
cells,
living
cells,
the
white
ones,
are
dead
cells
and
Conway's.
Game
of
Life
has
four
rules.
So
if
you're
alive,
meaning
black
and
you
have
fewer
than
to
live
neighborhoods,
you
die
because
you're
lonely,
it
doesn't
matter
which
ones
of
your
neighborhood
neighbors
they
are,
but
it
you
die
if
you
don't
have
more
than
two
or
fewer
than
two.
B
If
you
do
have
to
sir
rule
number
two,
if
your
live
cell
and
you
have
two
or
three
live
neighbors,
you
just
get
to
stay
alive,
you're
kind
of
happy
with
neighbors.
This
is
the
optimal
it's.
The
second
acceptable
number
of
neighbors
any
live
cell
with
more
than
three
live.
Neighborhoods
is
overcrowded
and
you
die
so
the
first
three
rules
have
to
do
is
live
cell
either
staying
alive
or
dying.
So
how
do
we
get
more
live
cells?
B
B
B
What
you
see
on
the
screen
here
is
just
a
bit
of
some
random
stuff.
You
can
have
things
that
reproduce
different
patterns,
things
that
fly
across
the
grid,
I
can't
begin
to
even
tell
you
go.
Look
it
up
sports
worthwhile,
so
we're
going
to
have
also
not
to
game
of
life.
It's
something
a
different
sort
of
rule
set
in
a
cellular,
automaton,
and
so
again
our
cells
will
possess
States,
also
called
types.
A
cell
will
admit,
sense
and
react
to
signals
so
that
some
kind
of
exchange
of
information
between
cells.
B
And
the
signal
is
actually,
the
purpose
of
signal
is
to
actually
transfer
transfer
the
state
of
neighborhood
cells.
So,
basically,
that's
how
themselves
find
out
what
other
cells
are.
What
state
they're
in
and
a
field
is
a
confluence
of
signal.
Sense
beat
cell,
so
it's
the
idea
is
that
I'm,
more
of
a
genetic
field
is
called,
is
a
super
positioning
of
neighborhood
sickles
about
their
states.
B
Compact
has
to
do
with
the
linear
growth
of
complexity,
noise
tower
it
has
to
do
with.
We
don't
want
one
little
little
glitches
here
and
there
for
the
fuzzy
bits
of
noise
to
throw
the
whole
thing
off.
The
touring
example
are
going
to
look
at
in
detail.
Actually
that
really
is
important
and
we
want
to
be
able
to
have
some
kind
of
generalization.
B
B
B
B
It's
basically
the
input
to
the
process,
so,
basically
in
us
in
a
cellular
automata,
it's
a,
but
it's
the
neighborhood
states
and
we're
going
to
do
this
thing
called
nests
at
neighborhoods.
Well,
nest
of
neighborhoods
me
is
like
Russian
dolls.
Little
neighborhoods
are
part
of
bigger
neighborhoods,
which
refer
to
pick
an
average
which
are
bigger
neighborhoods.
You
can
grow
indefinitely
linearly
as
I
set,
which
is
nice,
so
we
have
to
start
with
some
neighborhood.
Of
course,
you
know,
and
in
this
case
a
new,
the
lowest
smallest
neighborhood
is
a
single
cell.
B
B
Okay,
what
we're
going
to
do
in
each
sector?
It
could
contain
multiple
cells,
a
little
sub-grid
of
multiple
cells.
We
don't
want
to
actually
incorporate
all
that
information
into
the
value
of
the
sector,
because
that's
going
to
blow
up
and
get
too
complicated
as
neighborhoods
grow.
So
what
we're
going
to
do
is
just
go
through
the
sector.
This
little
imp
is
up
and
by
end
grid
kind
of
all.
The
cell
type
densities
count
up
all
the
cell
types
and
make
a
histogram
that's
the
value
of
the
sector.
B
Here's
a
picture-
hopefully
this
makes
more
sense.
So,
on
the
Left
we
have
a
grid.
We
have
a
bunch
of
cells,
some
of
them
are
white.
Some
of
them
are
black,
so
that
we're
great
I
think
there's
four
different
flavors
are
so
that
one
cell,
that
the
orange
line
points
do
is
the
most
elementary
neighborhood
of
that
cell.
B
So
that's
its
that's!
That's
neighborhood
sub
0!
Now,
if
we
look,
if
we
add
a
couple
of
layers
of
neighborhoods,
basically
we're
nesting
this
thing
we
can
put
that
cell
in
its
neighborhood
next
higher
neighborhood,
which
is
neighborhood
sablan,
that's
three
by
three
in
this
case
now
more
physics,
more
flexible,
it
doesn't
have
to
be
three
by
three,
but
it's
three
by
three
in
this
case.
B
So
that's
neighborhoods.
That
one
is
a
is
a
three
by
three
and
each
one
of
the
sectors
in
that
nine
cell
and
that
nine
sector
neighborhood
happens
to
be
a
single
cell,
because
it's
the
smallest
name,
it's
the
smallest,
not
elementary
neighborhood
and
right
below.
Here
you
see,
we
have
a
little
histogram
of
we
caught
up
all
the
different
types
of
cells.
That
would
make
it
a
little
histogram.
B
B
B
Same
histogram
trip,
so
we
go
ahead
and
so
here's
nine
sectors
in
neighborhoods
of
two
for
each
one
of
these
sectors
are
going
to
head
count
the
different
cell
types,
and
you
can
see
a
little
histogram
here
and
that
trick.
It's
not
the
only
trick
for
irrigating.
These
things
allows
us
to
kind
of
collapse.
The
amount
of
information
in
here
to
some
kind
of
manageable
amount.
B
So
now
we
have,
we
have
got
this
thing
called
a
morphogen
which
is
the
input
to
the
process.
So
morphogen
describes
what
we've
got
so
a
single
cell
can
sit
there
and
go.
Oh
I've
got
this
morphogen
sitting
here,
it's
a
bunch
of
mess
of
neighborhoods
surrounding
me
and
it
tells
me
in
a
in
a
nested
sort
of
local
to
global
way.
B
A
B
B
So
it's
a
state
action
role.
Ok
now
we
found
it
very
bit
of
beneficial
and
variance
of
interesting
and
useful
effects
by
using
an
artificial
neural
network
to
Train
min
immersed.
So
if
we
we're
going
to
look
at
a
bunch,
so
we're
going
to
have
runs
different
configurations
of
see
the
cellular
automata
they're,
going
to
generate
a
lot
of
metamorphs
which
contain
morphogens,
and
we
want
to
learn
which
morphogens
should
should
cause
certain
transitions
to
cell
types
and
a
useful
mechanism
for
compacting.
B
B
Oh,
here's,
my
here's,
my
origin,
plug
it
in
to
the
neural
network
and
outcomes,
it's
cell
transition,
so
we
train
this
patch
they
taking
the
morphogen,
which
is
this
nested
n
by
n
thing
and
flattening
all
that
stuff
out
and
stuffing
it
into
the
inputs
on
the
left
hand,
side
here
of
a
neural
network
and
then
through
the
through
the
magic
of
training
of
a
neural
network
which
I'm
not
going
to
discuss
here,
we
can
have
the
neural
network
figure
out
which
which
cell
transition
should
occur.
B
So
now,
once
we
do
that,
we
got
this
really
nice
little
box
here
that
we
get
quickly
at
this.
The
beauty
of
this
thing
is
its
moist,
tolerant
to,
and
that's
going
to
be
super
important
later
on.
For
the
example
we're
going
to
show
also
general
relatives,
which
is
also
super
important,
okay.
So
in
the
Pinta
chapter
we
have
we
decided
to
do
this
and
that
is
going
to
demonstrate
different
features
and,
of
course
we
should
baseline
it
so
to
speak
with
Conway's
Game
of
Life,
which
we
did
do
so.
What
does
that
mean?
B
Well,
we
we
played
some
games
of
life.
We
started
out
with
certain
input
country,
starting
configurations,
granite,
more
visit,
watched
a
generated
metamorphs
containing
morphogens
and
we
trained
it
to
be
able
to
play
the
game
of
life,
so
it
can
learn
essentially
specific
scenarios
of
game
of
life
by
watching
game
of
life
scenarios.
B
We
also
taught
it
how
to
do
cell
regeneration,
so
in
the
if
you,
in
the
chapter
or
you'll,
see
like
images
of
this
famous
image
of
alina
think,
it's
called
Lena
where
we
started
out
with
certain
pretty
sparse
patterns
of
Lena.
That's
a
woman's
face
and
it
can
regenerate
the
image
by
simply
watching
how
it's
done.
B
B
So
what
we
can
do
is
try
and
model
this
with
Martha's
owing
and
it
can
find
basically
neurons
can
seek
out
other
neurons
that
are
nearby
from
signals
in
a
very
generalized
manner,
so
they
don't
have
to
be
taught
specifically
this
neurons
here
and
that
neurons
there
send
an
axon
over
to
that
neuron.
It
can
use
these
signaling
cues
to
do
that.
It
can
learn
to
do
that.
The
last
thing
is
the
actual
demo
we're
going
to
do,
which
is
a
simulation
of
Turing's
reaction-diffusion
morphogenesis.
B
Here
it
is
so
Turing
came
up
with
a
scheme
for
take
a
mathematical
scheme
using
differential
equations
to
take
what
looks
like
kind
of
randomness
and
produce
patterns
that
are
credit
recognizable
as
some
kind
of
biologically
plausible
patterns
of
something.
So
here
is
a
cellular
automaton
version
of
this.
So
on
the
left,
this
is
a
random
pattern
on
the
right.
It
looks
like
something
like
a
cheetah
coat.
This
is
actually
known
as
the
cheetah
coat
pattern,
so
his
process
will
take
a
random
left
hand,
change
it
into
the
orderly
right
hand.
How
does.
B
Well:
here's
the
paper,
1952
and
sort
of
historically
for
historical
significance.
This
is
his
last
paper
before
his
tragic
demise
and
I
also
included
a
little
bit
of
background
about
it
from
Wikipedia.
So
you
can.
His
process
will
create
striped
spot
spirals
things
that
you
seemed
is
that
people
intuitively
see
in
biological
organisms,
not
just
the
cheetah
pattern,
and
it's
called
reaction-diffusion
is
the
general
term
for
this.
B
Here
is
the
equations
for
a
two
chemicals
system.
This
is
the
one
we're
actually
going
to
be
using.
This
is
the
one
that
creates
the
cheetah
banner.
So
if
you
have
a
chemical
you,
chemical
ve,
you
can.
The
chemical
you
and
the
chemical
V
will
change
any
cell
depending
upon
the
levels
of
concentration
of
these
two
chemicals
in
the
cell
and
there's
a
couple
parameters
here
that
if
you
set
these
differently,
if
you
set
these
one
way,
you'll
get
a
cheetah
pattern.
If
you
set
these
differently,
you
might
get
a
spiral
by
the
same
equation.
B
Now,
what
we're
going
to
do
is
expose
more
facility
to
this
and
see
if
it
can
learn.
What's
going
on
so
buff,
it's
always
going
to
take
a
random
pattern
on
the
same
way
on
the
pattern,
a
different,
random
pattern.
This
is
where
this
neural
network
comes
in
really
handy
because
it
can
actually
sort
of
settle
into
certain
patterns
that
it
learns.
B
Ok,
let's
get
into
the
nuts
and
bolts
here
so
to
do
more.
If
is
it
more
of
a
Zurich
is
written
in
Java,
so
you
need
Java
and
you
can
get
it.
You
need
not
just
Java,
but
you
need
the
JDK.
So
jdk
has
a
compiler
it.
So
that's
what
you're
interested
in
so
Chavez
aren't
by
Oracle.
Go
to
that
web
site
and
download
it
and
install
it.
B
A
A
B
B
Eclipse
is
really
good
for
java
development.
You
don't
have
to
use
java
clips
clips,
but
you
can
so
we're
going
to
do
it
manually,
though
so
we're
going
to
take
the
zip
file
unzip
it
we're
going
to
go
to
the
directory
and
we're
going
to
run,
build
up
at
because
I'm
out
windows
here.
But
if
you're
on
Linux,
you
can
even
run
the
build
up,
sh
and
I'm
also
going
to
run
a
couple
of
SH
commands
even
on
my
windows.
B
I
have
a
bash
cygwin
shell
2
to
the
first
one
I'm
going
to
run
is
actually
the
code
for
the
actual
original
touring.
Morphogenesis
and
morph
is
always
going
to
be
watching
while
it
does
that
and
learning
and
putting
it
in
a
save
file.
And
after
that
I'm
going
to
run
more
fuzz
Oh
train
a
neural
network
and
see
if
it
can
actually
reproduce
what
it
saw.
The
Turing
orogenesis
do
so.
A
B
B
B
B
B
B
B
B
B
B
B
A
B
B
B
Ok,
so
that's
how
you
can
actually
run
the
thing
e
and
obviously
the
codes
all
there.
Ok,
so
I've
got
a
couple
extra
minutes.
I
want
to
plug
another
project.
Very
briefly.
If
we
take
this
concept
of
global
local,
global
meaning
distant
in
space
and
time,
local,
meaning
like
recent
and
very
less
distant,
can
we
actually
use
that
to
compute
things,
not
just
morphogenesis,
so
basically
canna
brain
kind
of
do
this
so.
B
B
B
So
this
more
of
a
gnosis
is
the
project
day.
It's
called
once
again
morpho
for
shape
and
noses
for
knowledge.
Basically,
it's
the
shape
of
knowledge.
So
it's
it's.
It's
a
pyramid
and
it's
going
to
look
like
a
pyramid
of
space-time
chunks
which
we're
going
to
see
in
the
next
picture.
So
at
the
apex
are
the
most
recent
and
nearby
events
in
space
farther
away
on
the
pyramid
down.
The
pyramid
are
things
that
are
less
reason:
I
mean
more
distant
in
the
past
and
more
dispersed
in
space.
Here's
the
picture.
B
B
A
B
B
So
it's
the
neighborhood's,
the
neighborhood
is
3
by
3
and
we
have
this
little
histogram
bit
deal
so
basically,
this
is
a
chunk
of
space
time
with
a
histogram
of
the
information
that
was
in
that
chunk
of
space
tight.
Now
we're
going
to
use
that
to
learn
what
to
do
who
to
find
food-
and
here
is
it's
not
going
to
actually
use
this,
but
it
this
is
the
source
of
the
sister
grams.
So
again
it's
a
three
by
three
set
of
cells,
so
this
represents.
B
And
so
that
is
actually
an
ongoing
project
which
is
working
on,
and
that
is
my
presentation
they
might
want
to
give
us
since
I've
got
a
couple
of
minutes
here,
give
a
sneak
peek,
I'm
also
on
the
open
worm
education
committee.
So
I'm
working
on
getting
open
or
a
presence
in
the
mobile
arena
and
what
we're
doing
is
a
building
an
educational
have
to
allow
students
or
interested
parties
to
train
a
C
elegans
to
find
food,
and
this
is
based
on
a
couple
of
technical
papers.
B
B
B
B
Just
its
muscles
doesn't
have
any
other
energy,
and
so
we
want
the
word
to
actually
try
to
find
and
eat
this
day.
It
can't
do
it
right
now,
but
we're
going
to
be
able
to
these
are
neurons.
These
are
the
steering
neurons
within
the
worm,
the
actual
work
and
the
names
of
these
neurons
that
control,
how
it
senses,
salt
in
its
environment
and
how
it
turns
itself
left
right.
How
can
steer
itself
undulating
in
its
environment
to
get
to
that
food
source
defined
by
the
salt?
B
So
the
trick
is
you're
going
to
be
able
to
take
these
little
bubbles
here,
which
is
which
are
the
I
know.
This
is
all
very
quick
but
you're
going
to
be
able
to
wait.
These
connections
between
the
derides
such
that
they'll
signal
each
other
to
become
more
active
or
not
so,
for
example,
I
can
do
this.
One
and
I
will
be
able
to
change
the
waiting
on
that
synapse.
The
synapse
is
a
connection
between
two
neurons
and
I'm,
going
to
be
able
to.
B
B
So
that's
going
to
be
a
mobile
app
for
students
and
interested
parties
to
have
little
bit
of
fun
figuring
out
how
to
actually
modify
the
neural
network
of
this
C
elegans
in
a
very
sort
of
biologically
plausible
way.
It's
all
very
scientific
here
with
the
these
reputable
neuro
scientists
doing
this
and
that
will
be
forthcoming
before
the
end
of
the
year.
I
think
so.
That
concludes
my
talk.
Thank
you
very
much
and
back
to
you,
Bradley.
A
B
B
A
So
the
first
observation
is
that
if
you
look
at
the
game
of
life-
and
you
look
at
what's
going
on-
some
people
may
be
like
biological
cyberneticists
or
someone
like
that
would
argue
that
there
is
a
complex,
auto
regulatory
system
at
work
here,
but
the
cells
are
regulating
each
other
in
communities
and
furthermore,
some
people
might
argue
that
that's
a
you
know,
there's
something
called
biological
decision-making
going
on.
That's
you
know
one
view,
but
it's
interesting.
A
B
B
You
know-
are
the
basis
of
sub,
really
complex,
really
intuitively
interest
seeing
patterns
here.
So
so
you
can
get
things
like
a
little
flying
things
and
reproducing
things
and
so
on
with
this.
So
so
in
nature
you
know
so
that
I
mean
this
is
always
the
fascinating
questions
like
how
do
the?
How
did
nature
involve
certain
rules?
It
sells
a
bye-bye
to
sort
of
regulate
themselves
in
a
sensible
wastage
of
them,
survive
and
reproduce
you
know.
So.
B
A
B
B
A
A
Too
I
wouldn't
talk
about
that
later,
but
I
next
go
down
to
the
metamorph
or
the
like
them.
Where
you
have
the
no
down
a
little
bit
more
okay,
yeah
there
we
go
so
yeah.
So
I
guess
I'd
like
to
point
out
that
this
is
a
nonlin.
You
said
it
was
a
non
linear
mapping
between
inputs
and
some
sort
of
cell
type.
That's.
A
B
If,
if
it
turns
out
that
there's
some
kind
of
nonlinear
I
mean
some,
some
inputs
are
more
impor
than
other
influence,
and
maybe
they
don't.
You
know
like
there's,
maybe
some
kind
of
Gaussian.
You
know
bell-shaped
function
involved
with
the
inputs
and
the
outputs
it'll
earn
that
yeah.
So
it's
quite
happy
doing
it
also.
A
Important
to
point
out
that
C
elegans
were
dealing
with
like
a
deterministic
form
of
differentiation,
but
many
organisms.
You
have
many
signals
that
have
to
be
sorted
through
and
then
there's
some
decision
made.
That's
within
a
context.
So
you
know
it's
not
deterministic,
but
this
sort
of
model,
actually
you
know,
allows
us
to
really
kind
of
get
a
grip
on
it
without
really
being
too
hard
to
define,
or
you
know,
combinatorially
explosive.
So
yeah.
B
This
is
quite
compact,
also,
like
speaking
to
the
non-deterministic
thing.
I
mean
a
neural
network,
will
match,
do
a
best
match,
and
so
in
the
outputs
here
you
can
see
on
the
right
hand,
side
it
will
produce
essentially
a
probability
or
a
strength,
very
confident
swagger
for
each
one
of
these
cell
states
that
it
should
transition
to.
You
can
then
make
a
probabilistic
decision
as
to
what
the
social
change
too.
Okay.
A
B
A
So
a
final
question
I
had
if
you
go
up
to
the
where
we
talked
about
the
nested
neighborhoods
and
the
cell
densities,
yeah
I
hear
ya,
so
wanted
to
point
out
that
this
is
a
sort
of
a
strictly
hierarchical
system
where
you
have
nesting
of
complete
sets
within
complete
sets.
But
in
some
cases
you
might
not
have
that
in
a
biological
system
where
you
might
have
a
non,
you
know
strict
hierarchy
where
you
have
things
that
are
hierarchical,
but
they're,
not
nested,
necessarily
ya.