►
From YouTube: Frontiers of neural arts - GitHub Satellite 2020
Description
Presented by Gene Kogan
Over the past several years, two trends in machine learning have
converged to pique the curiosity of artists working with code: the
growing popularity of powerful open source deep learning frameworks like
Torch and TensorFlow, and the emergence of data-intensive generative
models for hallucinating images, sounds, and text as though they came
from the oeuvre of Shakespeare, Picasso, or just a gigantic database of
digitized cats. This talk will review these developments, present
various artworks, and offer a set of interdisciplinary tools and
learning resources for artists and data scientists alike, if ever there
was a difference to begin with.
Gene Kogan is an artist and a programmer who is interested in autonomous systems, collective intelligence, generative art, and computer science. He is a collaborator within numerous open-source software projects, and gives workshops and lectures on topics at the intersection of code and art. Gene initiated ml4a, a free book about machine learning for creative practice, and regularly publishes video lectures, writings, and tutorials to facilitate a greater public understanding of the subject.
GitHub Satellite: A community connected by code
On May 6th, we threw a free virtual event featuring developers working together on the world’s software, announcements from the GitHub team, and inspiring performances by artists who code.
More information: https://githubsatellite.com
Schedule: https://githubsatellite.com/schedule/
A
A
It
was
a
machine
learning
problem
how
to
recommend
music
to
people
and
in
that
ecosystem,
I
kind
of
learned
that
a
lot
of
people
interested
in
the
intersection
between
music
and
machine
learning
are
also
interested
in
problems,
how
they
could
do
with
creativity
and
composition
and
production
of
music
I
started
working
with
a
good
friend
of
mine
named
Jeff
Snyder,
who
built
these
really
awesome
musical
instruments
like
what
you
see
here.
This
is
called
a
burl.
It's
an
electronic
flute
instrument
and
it's
just
the
controller
has
a
whole
bunch
of
breath.
A
Brush
it
has
a
breath
pressure
sensor
and
an
embouchure
sensor
on
the
mouthpiece
and
then
the
whole
bunch
of
capacitive
touch
sensors
and
the
trick
that
we
employed
was
to
well.
It
doesn't
make
any
sound
by
itself,
and
so
we
would
try
to
hook
it
up
with
a
digital
synthesizer,
which
we
would
try
to
then
parameterize,
using
the
controllers
on
the
on
the
Conda
controller
that
the
musician
was
playing,
so
it's
kind
of
inverting
the
normal
process
for
for
music.
A
So
this
kind
of
got
me
interested
in
just
in
general,
into
interactivity
and
new
media
arts.
I
began
to
take
interest
in
things
like
computer
vision,
data
visualization
and
all
of
these
really
cool
things.
One
thing
that
I
made
that
that
gained
some
traction
a
few
years
ago
was
Kinect,
projector
toolkit
what
you
see
in
the
right
side.
This
enables
somebody
to
calibrate
a
projector
video
projector
with
a
Kinect
depth
camera,
and
this
would
allow
you
to
create
sort
of
spatially,
aware
body,
reactive
projections.
A
That
will
be
relevant
for
the
rest
of
the
projects
that
I
show.
You
is
that
neural
networks
are
really
great
at
learning
patterns
and
sort
of
salient
patterns
that
define
the
things
that
we're
interested
in.
So,
for
example,
this
is
a
neural
network
looking
at
an
image
of
a
cat,
and
all
of
these
different
cells
represent
the
amount,
the
the
activation
of
a
particular
feature
or
pattern.
A
In
the
image
at
every
possible
spatial
location-
and
you
can
see
that
some
of
the
cells
appear
to
be
looking
for
things
on
the
foreground,
some
on
the
background-
some
are
looking
for
simple
things
like
edges
and
ingredients,
but
all
of
them
are
sort
of
finding
features
and
in
the
features
that
are
sort
of
relevant
to
a
cap,
and
and
now
this
kind
of
what,
when
deep
learning
began
to
make
inroads
in
computer
vision.
Many
people
wanted
to
know
what
do
neural
networks
actually
see,
because
it's
a
little
bit
of
a
black
box.
A
We
don't
really
know
exactly
what
the
neurons
are
trained
to
recognize
and
so
some
of
the
experiments
that
began
maybe
around
ten
years
ago.
Looking
at
research
like
like
what
you
see
in
the
screen
above
tempted
to
answer
this
question
by
trying
to
you
know
it
make
interesting
procedures
for
generating
visuals.
That
would
help
us
understand
what
neural
networks
were
seeing
and
what
you're
looking
at
here
are
what
the
process
would
be,
for
example,
for
this
image.
A
It
was
taken
to
the
next
level
by
a
few
researchers
at
Google
who
developed
the
deep
tree
technique
and
before
deep
dream.
They've
they've
made
a
whole
bunch
of
innovations
to
this
visualization
technique,
which
made
much
more
vivid
pictures
that
were
much
more
realistic.
You
know
you
could
see
starfish
and
bananas
and
parachutes,
and
so
on
and
they're
very,
very
realistic,
and
this
was
kind
of
when
neural
networks
first
caught
the
eye
of
you,
know.
New
media
and
interactivity
and
artists,
new
media
artists
and
interactivity,
designers,
and
so
on.
A
So
I
was
super
interested
in
this
technique.
I
began
to
work
with
it
myself,
and
the
first
thing
that
I
really
wanted
to
do
was
to
try
to
gain
back
some
of
the
control
that
an
artist
is
used
to
having,
with
with
techniques
that
they
used
to
make
make
their
art
one
of
the
things
about
deep
dream.
Is
it
felt
a
little
bit
like
a
magic
trick
like
it
would
just
make
these?
A
That
was
using
masks
and
the
idea
of
masks
would
be
to
try
to
allocate
different
parts
of
the
canvas
to
different
neurons
that
I
wanted
to
visualize,
and
so
here
there's
a
mass
which
goes
from
left
to
right
kind
of
you
know
zero
to
one
and
then
from
left
to
right,
one
zero
and
each
one
was
devoted
to
a
particular
neuron.
And
then,
if
you
apply
this
mask
during
this
iterative
process
that
optimizes
the
pixels,
what
you
get
is
all
on
one
side,
you
know
one
particular
feature
is
visualized.
A
On
the
other
side,
another
feature
and
in
the
middle
it
kind
of
sees
the
attempts.
This
sort
of
kind
of
get
features
of
both
of
these
neurons
made
it
sort
of
satisfies
both
of
them
a
little
bit,
and
so
this
is
kind
of
an
example.
A
full-size
example
of
what
that
looks
like
you
could
see
that
there's
two
neurons
here
connected
by
this
interpolated,
horizontal
mask
and
the
mask
can
be
composed.
A
However,
you
want
so
in
this
example
that
you're
looking
at
the
mask
is
circular,
so
there's
a
sort
of
one
neuron
being
visualized
in
the
center
of
this
of
this
of
this
rectangle
and
then
on
the
outside.
You
have
a
another
neuron
which
appears
more
like
flowers,
something
like
that.
The
technique
could
also
be
used
to
generate
video
because
it
always
begins
with
an
input
image.
A
Here's
another
example
here
and
a
picture
of
an
eye
was
segmented
into,
and
then
each
of
the
segments
became
masks
and
then
these
masks
were
combined
with
a
bunch
of
different
neurons
that
I
chose
and
I
had
to
compose
this
final
result,
and
you
can
see
that
there,
it's
really
great
I,
think
kind
of
how
the
masks
fit
in
and
if
there's
no,
you
know
sort
of
there's
no
white
space.
There's
no
blending.
A
Just
kind
of
it
works
for
it
just
fits
in
very
nicely,
and
so
I
was
always
really
really
really
enthused
about
that
here.
This
is
kind
of
doing
the
same
feedback
loop,
but
but
also
creating
a
completely
never-ending
loop.
So
the
trick
about
these
two
videos
that
you,
you
probably
won't
notice
for
a
little
while,
is
that
there
are
perfect
loops
that
are
just
three
seconds
long.
A
So
if
I
hover
over
the
video
you'll
see
that
they
keep,
you
know
that
it
just
keeps
on
rotating
about
if
they're
every
three
seconds,
and
so
this
was
something
that
I
figured
out
how
to
do,
and
it's
a
little
bit
of
a
cool
illusion,
because
it
looks
kind
of
like
it's
always
expanding
outward,
but
it
never
really
goes
anywhere.
So
it's
kind
of
I
think
of
it
as
the
visual
equivalent
of
a
Shepard
tone.
So
this
endless
loop
can
be
combined
with
other
neural
neural
Arts
technique.
A
And
you
know
you
can
do
this
pretty
pretty
effectively
with
almost
anything
that
you
put
into
it.
Even
things
like
Google
Maps
end
up
working-
and
this
is
probably
my
favorite-
why?
Why
I
have
it
last?
It
looks
kind
of
like
a
nightmare.
You
know
you're
just
kind
of
on
your
phone
inside
of
the
nightmare
zooming
in
and
the
map
just
never
really
really
resolves
into
anything
readable,
more
examples.
This
is
the
loops
in
the
canvas
and
the
masking,
combined
with
with
the
d-train
technique
and
again
really.
A
The
idea
is
always
to
try
to
take
these
very
powerful
techniques,
but
give
back
more
compositional
ability
too
asks
or
through
canvas
distortion
or
some
other
means
another
example
of
using
source
images
as
masks
on
the
left.
You
see
the
Mona
Lisa,
pretty
recognizable.
All
of
the
segments
of
the
Mona
Lisa
have
been
devoted
into
masks,
and
this
is
also
a
perfect
loop
and
then
the
fellow
on
the
right,
which
maybe
some
of
you
recognize.
A
If
you
keep
track
of
the
field
of
deep
learning,
this
is
Yong
Laocoon
who's,
one
of
the
pioneers
of
deep
learning.
He
first
got
convolutional
neural
networks
to
work
on
some
kind
of
a
meaningful
applied
problem,
so
I
mentioned
earlier
style
transfer.
This
is
another
technique
that
I've
had
quite
a
lot
of
fun
with
over
the
last
over
the
last
few
years.
This
is
basically
the
recomposition
regeneration
of
one
image
in
the
style
of
another
image.
A
So
what
you're
looking
at
is
the
Mona
Lisa
in
a
style
of
an
NGO
in
the
style
of
Hokusai
and
style
of
of
Google
Maps
I
really
never
get
tired
of
using
this
Google
Maps,
texture
and
style
transfer
has
been
one
of
my
most
popular
techniques.
I've,
you
I've
installed
it
in
the
form
of
a
mirror.
So
what
you're
looking
at
here
is
an
installation
that
I
call
cubist
mirror.
A
It's
called
that
because
the
first
version
of
it
was
using
a
cubist
painting,
but
now
it's
it's
using
a
variety
of
different
paintings
and
it's
basically,
a
style
transfer
mirror.
So
you
get
in
front
of
it,
and
you
see
yourself
as
though
you're
a
Picasso
painting
or
in
the
style
of
hocus.
I,
kids
really
love
it,
as
you
can
tell
so.
I'm
gonna
switch
gears
and
talk
about
another
area
of
machine
learning
that
really
excites
me
and
that's.
A
Generative
models
and
generative
models
are
neural
networks
which
are
capable
of
actually
synthesizing
images
or
text
or
sound.
They
are
they're
extreme
of
extreme
interest
to
scientists
and
engineers,
and
they
have
many
many
applications
in
content
generation,
so
content
generation
for
things
like
chatbots
question,
answering
services,
video
games,
you
know
perhaps
even
completely
synthetic
bands
and
music
and
all
of
this
stuff
is
in
the
early
pipeline.
But
it's
but
it's
you
know
rapidly
and
and
also
very
interesting.
A
Whatever
it
is
that
you
trained
on
whether
it
be
faces
or
cats
or
dogs
or
trucks.
Or
you
know,
or
airplanes,
and
so
on,
and
they
output
images
which
do
not
come
from
the
original
data,
but
look
as
though
they
could
have,
and
there's
kind
of
a
few
dominant
paradigms
for
this
autoencoders,
where
the
the
sort
of
granddaddy
of
generative
models
they've
been
chugging
along
for
the
last
ten
to
fifteen
years,
generative
adversarial
networks
or
gans
have
sort
of
up
all
the
press.
A
In
the
last
few
years,
they've
been
responsible
for
most
of
the
most
of
the
sort
of
like
well
the
most
visually,
the
images
that
had
the
highest
visual
fidelity,
let's
say
and
of
the
last
few
years,
and
so
lots
of
people
are
pretty
excited
about
games
and
the
first
implementation
of
a
of
again,
which
worked
on
images,
was
a
DC
gun
which
was
made
by
these
people
in
2015,
and
they
demonstrated
really
really
wild.
You
know
characteristics
of
this
of
this
technique.
A
They
showed
how
you
can
do
arithmetic
on
the
feature
space,
so
you
can
take
an
image
of
a
man
with
glasses
and
then
find
the
latent
vector,
for
just
man,
add
the
latent
vector
for
a
woman,
and
you
get
man
with
glasses.
Man
plus
woman
equals
woman
with
glasses.
So
you
have
this
sort
of
ability
to
add
and
delete
and
modify
features.
A
generative
images
as
though
they're
kind
of
like
DNA,
so
I
use
their
implementation
of
dcen
to
make
a
a
project
called
a
book
from
the
sky.
A
This
was
my
first
project
with
Ganz
and
it's
basically
trained
on
a
dataset
of
handwritten
Chinese
characters
and
which
I,
which
was
being
collected
at
the
University
of
Harbin,
basically
for
the
purposes
of
optical
character.
Recognition
I
was
really
interested
in
using
them
for
for
Ganz,
and
so
I
got
permission
to
use
this
data
set,
and
it's
really
interesting
to
look
at
these
interpolations
and
see
the
different
ways
that
people
write
in
Chinese
I
learned
quite
a
lot
a
bit
about
Chinese
handwriting.
You
can
read
more
in
this
blog
post
that
I
wrote
about
it.
A
This
is
another
cut
from
there.
These
are
interpolations
between
different
characters
and
you
can
see
that
the
characters
that
lay
in
between
are
really
quite
striking
because
it's
almost
as
though
they
stand
for
concepts
that
that
exists
in
this
sort
of
you
know
this
conceptual
fabric
that
we
were
given
from
language,
but
but
that
the
Bernal
characters
devoted
to
and
so
I
was
I
was
really
quite
yeah.
This
was
kind
of
a
really
satisfying
project.
For
me,
my
most
recent
use
of
these
I
trained
again
on
around
100,000
paintings,
collected
from
a
website
called
wiki.
A
Our
wiki
art
is
this
open
source
collection
of
you
know
hundreds
of
years
worth
of
art,
and
you
can
see
that
it's
capable
of
synthesizing
what
looks
like
medieval,
portraits
to
document
archives
to
landscapes
to
photographs.
To
you
know
all
sorts
of
genres
are
kind
of
inside
this
inside
this
again
again.
A
In
the
same
spirit,
this
is
usually
again
called
big
again
here:
I'm
combining
different
classes
that
are
inside
again
and
finding
so,
for
example,
on
the
left
side
you
have
an
owl
and
in
the
middle
you
have
a
dog
and
then
on
the
right
side.
You
have
basically
an
owl
dog.
This
is
kind
of
in
the
same
idea
that
I
showed
in
the
DCN
paper,
in
which
they're
able
to
add
and
subtract
features
that
we'll
another
example
of
this
alga.
A
It
looks
a
little
bit
like
a
Harry
Potter,
Harry,
Potter
character,
I've
always
kind
of
really
enjoyed
this
this
one-
and
this
is
a
owl
mixed
with
a
cap
like
a
Snow
Leopard.
Something
like
that.
This
is
a
this
is
my
favorite.
This
is
a
whale
or
an
orca
and
then
a
tram
in
the
middle,
and
then
this
is
the
whale
plus.
The
tram
and
the
whale
plus
a
tram
ends
up
looking
like
a
ferry,
so
this
is
kind
of
like
a
really
unexpected
sender.
A
For
this
result,
the
last
thing
I'll
show
you
with
big,
and
this
is
a
on
the
left
side.
What
you
see
is
a
video
taken
from
YouTube.
This
is
planet
earth,
no
BBC
documentary
about
how
beautiful
our
earth
is
and
then
there's
a
pipeline
in
which
each
frame
is
analyzed
and
predicted
for
all
the
classes.
A
All
of
the
different
objects
that
appear
in
the
image-
and
that's
here
that
becomes
the
label
vector,
which
you
then
feed
into
begin
in
order
to
generate
an
image
with
those
labels,
and
so
in
some
sense
it's
it's
began,
attempting
to
to
regenerate
the
video
on
the
left,
and
so
whenever
it
has,
you
know,
mushrooms,
you'll,
see
mushrooms
whenever
it
has
different
kinds
of
birds
and
monkeys.
You'll,
see
those
see
the
monkeys,
kind
of
being
generated,
they're
foxes.
A
From
undersea
creatures,
sometimes
it's
a
little
unsure,
so
it
can
be.
It
can
be
pretty
wild.
Another
category
of
generative
models
that
I'm
really
interested
in
are
these
image
to
image
translation
networks.
The
first
one
of
these
that
gained
a
lot
of
the
attention
of
the
art
world
was
pics
the
pics,
and
so
they
demonstrated
how
you
can
train
a
neural
network
to
basically
create
the
most
advanced
image
filters.
A
So
one
image
filter
which
takes
a
label
map
and
then
converts
it
into
a
facade
like
a
photograph
of
a
building
or
a
label
map
into
a
street
view
scene
black-and-white
photo
to
a
color
photograph,
taking
satellite
imagery
and
converting
it
into
maps
or
possibly,
maps
and
satellite
imagery
day
photos
into
photos,
edges
no
edge
drawing
since
your
photographs
and
so
on
lots
of
different
examples
of
this.
My
first
project
with
this,
which
was
with
some
collaborators
and
in
Milan,
an
open
dot
lab
for
a
project.
We
called
invisible
cities.
A
This
is
named
after
the
book
by
Italo
Calvino,
and
here
what
we
did
was
we
treat.
We
downloaded
lots
of
map
tiles
from
OpenStreetMaps
and
and
corresponding
satellite
imagery
from
a
whole
bunch
of
different
cities
and
train
picks
the
pics
in
order
to
convert
the
map
tiles
into
the
satellite
imagery
tiles,
and
then
what
you
could
do
is
once
you
have
these
models,
you
can
take
the
map
map
tiles
from
from
one
city
and
run
it
through
the
generative
model
of
another
city,
and
so
you
get
this
sort
of
city
style
transfer.
A
So,
for
example,
on
the
left
here
you
have
overhead
of
Milan
and
then
in
the
middle
you
have
Milan
that
part
of
Milan
in
the
style
of
Los
Angeles
and
then
Milan
in
the
style
of
Venice.
So
on
and
I
also
use
pics
the
pics
to
create
sort
of
I
guess
deep
fakes
before
they
were
called
that
this
is
from
and
I
think
2016,
which
I
basically
made
like
a
president
Trump
or
me
puppet
as
I
called
it.
A
You
know
steal
somebody's
face,
essentially
in
their
likeness,
but
if
you-
but
of
course
you
know
now
it's
becoming
hyper
realistic
and
so
that
you're
looking
at
work
by
by
Nvidia,
in
which
they
are
demonstrating
this
technique,
to
the
point
that
it's
almost
impossible
to
distinguish
the
real
faces
from
the
from
the
fake
ones
and
so
I
kind
of
think.
Sometimes
my
job
as
an
artist
is
to
warn
people
about
the
future.
You
know
so
something
like
being
able
to
steal
somebody's
face
there.
A
Does
you
two
like
this
and
didn't
seem
so
compelling
in
2016
when
it
looked
like
this,
but
you
know
if
you're,
if
you're
looking
into
it,
you
may
realize
that
it's
actually
quite
a
bit
closer
than
it
first
seems
so,
I'm
also
interested
in
interactive
installations.
So
here
you
can
see
an
installation
made
at
a
museum
in
Berlin
called
future
diem.
A
It's
not
completely
realistic,
but
what
looks
like
a
photograph
of
what
you
just
drew
and
along
the
same
lines,
this
is
invisible
cities
except
reimagined
as
an
interactive
installation.
So
here
you
have
these
plastic
pieces
which
correspond
to
two
grass
tips
water
into
buildings
and
then,
as
you
move
them
around
this
little
playpen
and
there's
a
camera
which
is
picking
them
up
and
then
converting
the
image
into
what
looks
like
a
satellite
image
of
Berlin.
A
And
so
all
of
these
are
installations
have
been
I've,
been
pleased
to
use
for
the
last
little
while
another
interactive
installation
shows,
with
the
collaborator
of
mine
named
andreas
Jeff's
card,
and
the
idea
here
is
that
you
can
draw
musical
instruments
and
then
have
the
music
music
automatically
generated
using
those
instruments.
So,
let's
see.
A
You
know
versions
of
myself
that
might
exist
on
there
in
some
parros's
and
you
could
also
use
it
to
make
interpolation
between
yourself
and
other
people.
This
is
me
being
converted
into
a
certain
Canadian
pop
star
that
you
might
all
recognise.
This
is
me
being
converted
into
some
heads
of
state.
That's
well,
I'm
sure
you
recognize
all
of
them.
Yeah
well,
I'm.
Just
saying
have
you
ever
seen
me
in
the
room
with
them
with
any
of
them?
At
the
same
time?
That's
really
all
there's
another
app
that
I
had
little
fun
with.
A
Maybe
some
of
you
remember
face
app.
It
was
pretty
popular
in
the
app
store
for
a
little
while
and
iPhone,
and
it
allows
you
to
take
a
picture
of
a
face.
Then
put
her
a
smile
on
it,
and
so
this
is
not
my
actual
smile.
This
is
just
a
saps.
You
know
version
of
my
smile
and
then
what
you
could
do
with
I
thought.
A
It
would
be
interesting
to
take
the
output
of
that
image
and
then
run
it
again
through
the
smile
filter
and
then
take
the
output
of
that
image
and
running
again
through
the
smile
filter,
and
so,
if
you've
ever
been
interested.
What
that
might
look
like
it
looks
basically
something
like
this
and
then
right
around
now
it
stops
picking
up
faces,
and
so
the
algorithm
basically
breaks
at
this
point.
A
So
there's
someone
told
me
that
this
looks
a
little
bit
like
a
text
win
so
I
think
that
was
kind
of
what
I'm
going
for
and
then
you
know
just
having
fun
lots
of
ways.
I
think,
there's
a
lot
of
fun
to
be
had
in
these
kinds
of
neural
arts.
Here
I'm
projecting
Einsteins
face
into
the
Crab
Nebula.
This
is
really
closely
related,
can
style,
transfer
and
then
so
I
started
inserting
myself
into
various
paintings.
A
This
idea
of
what
really
true
autonomy
looks
like
in
in
a
computer
and
I
thought
that
we
can
I
looked
in
nature
for
some
inspiration,
so
I
find
inspiration
from
things
like
beehives
and
flocks
of
bees
and
termite
mounds,
and
all
of
these
are
examples
of
super
organisms
and
the
idea
that
I'm
kind
of
after
is
this
metaphor.
We
have
in
which
we
associate
some
kind
of
intelligence,
no
collective
intelligence
from
a
super
organism.
A
After
this
hive
mind
metaphor
to
try
to
to
you,
know
to
demonstrate
something
that
really
truly
has
its
own
agency
and
the
construction
that
I
that
I
think
might
be
able
to
achieve
this.
Is
they
decentralized
the
centralized
machine
learning
process
in
which
a
generative
model
is
trained
and
using
a
set
of
technologies
which
which
allow
a
group
of
people
to
co-own
that
model
together
as
a
shared
secret?
A
You
know
medical
applications,
financial
applications,
social
applications
and
so
on,
and
maybe
it
can
also
make
art,
and
so
this
is
kind
of
the
the
goal
of
this
project
that
it's
a
really
slow
work-in-progress.
It's
very
multi,
dear
multi-dimensional
kind
of
draws
from
from
these
entropy
centralization
technologies,
from
AI
and
computer
art
and
and
from
philosophy
of
mind,
and
and
so
it's
it's
really
I
know
my
new
obsession.
A
Okay,
so
I
just
have
a
few
seconds
left,
I'm,
just
gonna,
really
quickly
mention
I.
Do
a
lot
of
workshops.
I've
been
teaching
workshops
as
a
full-time
job
for
the
last
couple
of
years,
and
all
of
this
has
led
into
a
book
called
machine
learning
for
artists.
This
is
a
website
online.
At
ml.
Freight
I
give
her
but
IO,
where
I
create
a
home
where
I
and
along
with
collaborators,
including
add
Reyes,
who
I
mentioned
earlier.
A
We
work
on
lots
of
demonstrations,
tutorials
applications
that
are
able
to
show
you
how
to
use
machine
learning
for
your
own
work.
I
also
teach
classes
online
and
we
can
post
those
on
ml
for
a
so.
If
you're
interested
in
all
of
the
30
hours
of
this,
then
you
can
go
learn
for
yourself
how
this
stuff
works,
I'm,
also
an
advisor
to
run
the
way
you
probably
just
saw
Chris
and
honest
Asus,
I'm,
really
big
fan
of
Runway
and
very
excited
about
the
ability
to
kind
of
make
this
stuff
super
easy
for
people
to
use.