►
Description
Taking on Genetically Evolving Cellular Automata with JavaScript - Irina Shestak, Small Media Foundation
"Genetically evolving cellular automata -- that’s a mouthful. Let’s break it down
1. We will be looking at cellular automata. A cellular automata is a set of units governed with very simple rules. These units make up a complex system, or model.
2. We can train cellular automata to genetically evolve, making themselves better overtime. We can make them adhere to a set of rules that would make the system reach a certain outcome at the end of N generations. We can even use these concepts in the wild.
3. We will be doing it all with JavaScript."
A
Hey
everyone.
So,
thanks
for
having
me
note
interactive,
this
is
going
to
be
awesome.
My
name
is
irina.
I
am
a
that
door.
Oh
my
god.
Okay,
I'm,
a
development
team
lead
at
small
media
foundation.
I'm
based
out
of
london.
I
used
to
be
out
in
vancouver,
which
is
why
the
canadian
accent
but
I
live
in
london.
A
So
the
title
is
a
little
bit
complicated
and
I
think
we
should
kind
of
break
it
down
a
little
bit
and
start
with
a
story
as
to
even
how
I
got
into
this
in
the
first
place
so
story.
Time,
like
I
said
I
am
from
vancouver
and
in
vancouver
it
rains
about
99
of
the
time
and
so
last
november
I
was
out
to
go
for
a
hike
and
our
library
which
is
like
this
coliseum-like
library,
not
as
cool
as
this
building,
but
still
pretty
cool.
A
I
read
the
first
couple
of
pages
and
it
was
awesome
and
she
talks
about
a
bunch
of
different
things
that
make
up
complexity,
melanie
mitchell,
who
is
the
author,
but
she
also
talks
about
cellular
automata
in
about
10
pages
of
through
the
300
pages
of
the
book,
and
that
particular
concept
kind
of
struck
me-
and
I
thought,
okay,
how
about
I
explore
it?
I
know
I
know
javascript.
A
A
He
described
two
kinds
of
automotive,
one
being
kind
of
the
artificial
and
the
other
one
being
the
natural
like
seashells
and
things
like
that,
where
certain
parts
of
automata
are
present
and
we're
obviously
going
to
be
talking
about
the
computer
ones.
Okay,
so
I
kind
of
want
to
know
more
a
little
bit.
What
is
it,
what
is
a
settler
automata,
so
I've
been,
as
I
was
reading
through
his
work.
A
I've
put
together
kind
of
the
four
I've
extracted
the
four
concepts
that
make
an
automata
the
way
he
would
would
envision
the
automata
to
be,
and
so
rule
number
one
or
like
the
trade
number.
One
is
the
fact
that
whatever
the
result
that
comes
out
of
the
automata
is
dependent,
how
these
elementary
parts
are
defined,
so
it's
dependent
on
how
the
rules
are
set
and
everything
that
functions
with
automata
is
based
on
the
rules.
A
A
Three
operations
are
not
directed
themselves,
so
whenever
something
happens,
it's
not
they're
not
changing
themselves.
I
guess
for
considerations.
Von
neumann
actually
did
or
like
he
had
like
this
interesting
relationship
when
he
wrote
about
it.
So
operations
are
not
directed
at
themselves,
they're
transfiguring,
something
else,
and
it's
a
decentralized
system
and
kind
of
these
four
different
traits
that
have
been
put
together.
A
A
The
puncher
itself
is
the
automaton,
and
the
tape
that
is
infinite
in
both
directions
is
the
one
that
either
has
a
punched
hole
through
it
or
an
unpunched
hole
through
it,
and
so
the
automaton
produces
the
action
of
actually
punching
the
hole,
and
the
tape
is
what
the
result
is
at.
So
basically,
the
action
that
it
produces
is
not
directed
at
themselves,
which
basically
means
that
entirely
the
automata
itself
and
the
tape,
and
what
comes
out
of
it.
A
So
another
definition
that
melanie
mitchell
put
together
for
this
automaton
was
the
fact
that
simple
components
with
local
information
and
communication
give
rise
to
coordinated
global
information
processing.
So
what
happens
is
that
these
simple
rules
that
have
been
defined
to
the
cellular
automata
are
basically
directing
directing
the
way
the
information
gets
processed.
A
So,
let's
go
back
to
kind
of
the
trade
number
two
that
we've
mentioned
previously.
The
fact
that
automata
is
aware
there
have
been
kind
of
I've
boiled
it
down
to
three
different
kind
of
series
of
ways
that
awareness
could
be
structured
and
they're
based
on
kind
of
three
different
people
that
worked
on
automata.
So
one
was
john
von
neumann
who's.
We
who
we
talked
about
before
so
initially
when
he
was
writing
his
or
doing
his
lectures
on
self
for
producing
automata
he.
He
started
off
with
like
this.
A
This
grid,
this
lattice
of
light
bulbs
that
he
hooked
up
and
he
put
them
together.
So
I
made
a
replica
obviously-
and
I
think
it's
pretty
close
so
basically,
what
would
happen
is
that
light
bulbs
would
be
connected
between
one
another
and
they
would
be
turning
on
and
off,
based
on
what
their
neighbors
were
doing.
So
if
their
neighbors
were
on
per
se,
eventually
the
light
bulb
would
turn
off,
and
this
was
kind
of
his
initial
replica
actual
hard
light
bulbs
actual
wires,
trying
to
kind
of
put
this
together.
A
A
So,
if
you're,
a
cup
of
coffee,
you're
dependent
on
kind
of
your
neighborhood,
which
are
the
cake,
so
there's
only
two
states
in
automata,
it's
a
binary
states,
binary
state
kind
of
concept,
so
you
can
either
be
one
or
the
other
with
a
lot
of
computational
papers.
You
will
have
like
this
concept
of
dead
or
alive,
so
they
would
just
use
black
or
white
dots.
A
I
just
describe
it
with
cake
and
coffee,
so
you're,
dependent
on
your
neighbors
and
conway's
game
of
life
works
is
the
fact
that
you
have
this
giant
grid
that
you
will
then
turn
on
certain
states.
You
would
turn
them
into
that
and
then,
if
basically,
given
this
example,
if
you're
a
cup
of
coffee-
and
you
have
three
cakes-
you
become
a
cake.
A
If
there's
two
or
three
cakes
around
in
your
still
cup
of
coffee,
you
stay
a
cake,
and
if
there
are
less
than
two
of
you,
you
become
a
coffee
kind
of
like
that,
a
little
bit
confusing.
But
basically
what
happens
is
that
you
have
this
grid
that
changes
over
time
and
the
grid
stays
the
same
and
so
different.
A
different
concept
that
came
out
of
that
was
kind
of
these
interesting
patterns
that
you
would
end
up
getting
so
over
time.
A
So,
with
conway's
game
of
life,
it's
more
of
a
two-dimensional
binary
concept
that
is
relevant,
but
it's
two-dimensional
because
you're
looking
at
the
grid
itself
as
your
canvas
and
binary,
obviously,
because
there's
only
two
states
that
the
cells
could
take
on
so
another
person
came
in
and
he
said
okay.
This
is
a
little
bit
complicated.
I
want
to
simplify
it
and
I
want
to
be
able
to
derive
certain
solutions
from
that.
A
So
again,
as
a
cell
you're,
still
aware
of
your
surroundings
and
you
still,
your
actions
are
not
to
work
toward
towards
yourself
but
towards
others,
and
you
affect
everybody
around
you,
but
in
code
perspective,
this
is
how
it's
going
to
look
like.
So
you
start
off
with
a
one-dimensional
array
and
over
time,
you're
going
to
progress
and
you're
going
to
change
as
you
go
along.
A
A
So
I
think
when
I
started
off
with
this,
I
just
tried
to
code
out
parole,
110
and
see
where
I
get
to
it,
and
I
was
able
to
kind
of
replicate
something
that
looks
like
this,
and
it
was
pretty
awesome.
So
the
grid
would
be
just
like
this
giant
grid
that
progresses
over
time,
and
then
you
basically
explore
the
patterns
that
came
out
out
of
it.
A
A
I
wanted
to
look
at
what
exactly
is
like
this
pattern
that
goes
on,
and
then
I
looked
into
cellular
automata
in
this
particular
rule
110,
and
I
realized
that
in
fact,
because
the
pattern
that
comes
out
of
it
is
so
complicated
and
so
different,
they
actually
call
this
particular
rule
touring
complete,
and
what
that
means
is
that
a
computational
system
that
can
compute
every
turing
computable
function
is
called
touring
complete
or
turing
powerful.
Alternatively,
such
a
system
is
one
that
can
simulate
a
universal
turing
machine.
A
You
come
out
with
a
quite
complicated
pattern
that
takes
on
so
we've
noticed
that
certain
part
patterns
end
up
dominating
in
the
particular
environment
and
kind
of
looking
into
the
pattern
as
a
concept
itself.
It's
basically
a
set
of
configurations
sharing
a
common
spatial
structure
and
in
particularly
with
most
of
wolfram's
256
rules.
A
So
what
was
then
said
by
a
bunch
of
people
who
were
looking
into
this
is
the
fact
that
there
is
still
a
diversity
of
these
spatial
temporal
behaviors,
but
the
system
is
organizing
itself
into
patterns
based
on
the
rules
that
it
follows,
and
that
was
really
interesting
to
me
that
it
organizes
itself.
So
I
even
had
three
slides
to
point
out
the
fact
that
it
organizes
itself
into
patterns
which
is
incredibly
fascinating.
So
that
leads
me
to
kind
of
this
entire
concept.
A
So
let's
go
back
to
kind
of
the
fact
that,
given
these
particular
patterns,
we
were
looking
into
this
incredibly
diverse
spatial
temporal
behaviors
and
the
fact
that
the
system
is
organizing
itself
into
patterns
which,
given
the
particularly
simple
examples
or
simple
rules,
that
automata
follows.
We
come
up
with
a
very
complex
systems
that
turn
out
in
the
very
end.
A
So
and
then
I
guess,
as
I
was
kind
of
researching
this
or
putting
this
together
in
a
sense,
I
was,
I
guess
fascinated
by
this
entire
con
concept
of
complex
systems
in
general
and
what
is
even
a
concept,
complexity
or
complex
system.
So
I
was
researching
this
and
I
went
on
the
internet,
and
google
straight
up
gives
me
that
it's
a
state
or
quality
of
being
intricate
or
complicated.
A
But
otherwise
I
I
think
that
still
doesn't
answer
the
question
of
complexity
of
what
it
is
in
this
sense
or
in
the
space
of
automata.
So
another
person
came
up
with
a
couple
of
questions
and
he
does
really
awesome
videos
about
complexity
and
kind
of
model.
Thinking
and
he's
put
together,
three
different
answers
or
three
different
questions
that
we
could
answer
to
try
to
define
whether
something
is
complex
and
three
of
them
are,
is
how
hard
is
it
to
describe
the
particular
system?
A
How
hard
is
it
to
create
and
what
is
the
degree
of
such
organization?
And
so
I
guess
that
still
kind
of
helps
us
out
with
identifying
what
it
is
complex
and
what's
not
complex,
but
also
not
a
very
standard
answer.
In
fact,
I
went
searching
for
quite
a
few
more
answers
and
it
seems
like
folks
that
study
this
can't
agree
on
the
definition
of
complexity,
much
like
javascript
developers
can
degree
and
what
framework
to
use.
A
So
I've
decided
to
go
back
to
like
the
initial
stage
of
this
and
look
into
melanie
mitchell's
definitions
for
complexity,
and
she
says
that
basically,
the
average
information
content
or
the
amount
of
surprise
that
you
can
come
up
with
based
on
your
particular
algorithm,
is
what
something
is
complex.
So
if
you're,
I
guess
surprised
by
your
result.
It's
a
much
more
complicated
system.
A
A
Okay,
so
this
has
been
as
I,
as
I
said
like.
This
has
fascinated
a
lot
of
people,
because
it's
a
very
simple
model
that
could
reproduce
something
or
help
people
work
with
much
more
lifelike
models.
And
so
the
next
thing
that
people
decided
to
do
is
that,
okay,
we
have
this
model
that
we
could
write.
A
What
happens
if
we
work
with
evolving
these
particular
cells
and
see
where
we
get
to
it,
and
so
why
this
kind
of
came
to
people's
minds
is
because
it's
a
lot,
if
you
look
at
it
or
if
you
think
about
it,
it's
very
organic
in
a
way
humans,
reproduce
or
things
that
are
alive
or
produce
and
kind
of
very
similar
to
genetics,
work,
and
so
people
started
looking
into
it.
But
the
first
was
a
kind
of
annoyment
and
when
he,
when
he
was
working
on
his
theory
of
self-reproducing
automata,
he
kind
of
envisioned
this
robot.
A
This
was
very
awesome
of
like
honestly.
For
me,
he
envisioned
this
robot
that
basically
would
be
kind
of
structured
in
the
middle
of
nowhere
and
or
it
would
have
structure
in
the
middle
of
nowhere.
And
what
would
happen
is
that
it
would
have
access
to
all
the
possible
different
parts
it
may
or
may
not
need
ever,
and
then
it
would
kind
of
build
itself
as
it
required
and
as
it
evolved.
A
It
would
require
more
and
more
parts,
and
if
it
were
to
reproduce
it
would
require
more
and
more
parts,
but
it
would
always
have
this
access
to
the
parts,
but
once
people
have
been
putting
it
together
they've
they
decided
to
simplify
the
task
and
see
what
they
could
do
with
like
this
one-dimensional
lattice.
So
there's
work,
kind
of
in
both
one-dimensional
and
two-dimensional
space
and
people
have
been
putting
together,
algorithms
and
genetic
evolutions
based
on
both
kind
of
both
models.
A
A
So
you
start
off
with
a
random
state
lattice
or
an
array
or
a
matrix,
random
state,
again
you're
stuck
to
the
binary
state
solution
and
you
perform
you
kind
of
run
the
generation
over
time.
So
you,
you
would
have
several
generations
to
kind
of
convince
your
neighbors
to
vote
for
the
opposite
member,
and
the
goal
of
this
is
to
make
sure
that,
if
you're
in
a
minority
at
the
beginning,
you
want
to
be
in
the
majority
at
the
end.
A
In
fact,
you
want
to
win
at
the
end
and
you
kind
of
evolve
and
evolve
and
evolve,
and
each
one
of
these
items
or
cells
in
the
lattice
start
off
with
a
different
set
of
rules.
So
normally
would
like
wolfram's
automata.
You
would
have
a
single
state,
a
set
of
rules
that
every
single
one
of
them
follows.
A
If
you
were
one
state,
you
want
to
be
the
other
state
at
the
end,
and
if
you
get
there
majority
of
the
time
in
150
years,
you're
good
and
then
you
can
kind
of
have
your
own
metrics
as
to
how
much
those
particular
variables
could
be.
So
you
start
off
kind
of
over
time.
You
pick
out
okay,
so
this
wasn't
very
successful
because
I'm
still
in
the
same
state
over
time,
but
I
want
to
be
in
a
different
state,
the
next
time
around,
but
basically
the
concept
over
with
this
density
classification
task
is
very
straightforward.