►
From YouTube: DevoWorm (2021, Meeting 45): Soft Active Materials, Liquid Crystal Bio III, Biological Intelligence
Description
Presentation on Soft Active Materials and discussion of Liquid Crystal Biology. Discussion of Biological Intelligence and Biological Simulations in 3-D, 4-D. Recap of selected readings from NeurIPS 2021. Attendees: Susan Crawford-Young, Richard Gordon, Karan Lohaan, Valentina Perricone, and Bradly Alicea.
B
B
B
A
A
C
A
C
B
B
Yeah
welcome
to
the
meeting
hello
quran.
It's
the
last
meeting
of
the
year,
we'll
take
a
break
for
a
couple
weeks
and
come
back
in
january,
and
so
I
didn't.
I
don't
have
a
end
of
the
year
report
because
we
did
an
end
sort
of
a
year
summary
in
september.
I
did
it
for
the
open
world
manual
meeting
and
so
that's
sort
of
our
recap.
For
the
year,
hello,
dick.
D
E
F
G
F
Biologist
and
I
worked
on
the
mechanical
design
of
searching
the
skeletal
part,
and
I
did
a
research
on
the
on
the
structure
of
the
echinoid
test.
That
is
the
skeleton
to
identify
new
biological
insight
regarding
the
structural
organization
and
also
identify
a
functional
strategy
to
be
transferred
in
an
engineer
in
engineering,
a
civil
engineer,
the
industrial
design
product.
So.
F
E
B
Welcome
to
the
meetings
and
yeah,
we
do
a
lot
of
this
sort
of
stuff
in
the
group
here
we
talk
about
like
different
approaches.
We
look
at
you,
know,
developmental
biology,
but
also
computer
science,
and
also
some
physics
and
engineering
and
susan's
going
to
be
presenting
today
on
something
like
that
she's
going
to
be
presenting
on
soft
materials.
A
A
And
this
can
even
expand
to
small
autonomous
robots
I
get.
You
can
actually
make
a
system
with
small
autonomous
robots.
I
saw
youtube.
Videos
are
very
good
for
this
and
I
saw
a
youtube
video
with
the
the
vacuum
cleaners.
They
got
a
whole
bunch
of
them
and
they
pushed
around
the
boxes
that
they
came
in
and
they
they
made
patterns
so
they're
they're,
they're
kind
of
an
active
man
that
was
kind
of
an
active
matter
system.
A
A
B
A
Reynolds
number
and
the
movement
of
small
creatures
through
well
through
water-
mostly,
I
think
it
was
water
next
slide
yeah
here
it
is,
and
they
have
a
there's,
the
fluid
that
they
were
in.
They
shook
that
up
and
down,
and
they
found
that
this
combination
of
balls
initially
moved
backwards
and
then
it
moved
forwards.
A
So
there
was
a
crossover
point
and
that's
what
the
critical
onset
transition
was
and
the
plus
signs
are
simulations,
so
they
were
able
to
get
pretty
good
data
and
really
nice
simulations
of
what
was
going
on,
and
so
next
slide,
I'm
just
gonna.
This
is
just
a
fast,
so
you
can
see
that
the
double
ball,
which
I'm
calling
a
swimmer,
initially
moves
down
with
the
reynolds
number
of
five
and
then
as
it
gets
as
reynolds
number
increases
it
moves
forward
in
the
graph.
A
You
can
see
that
reynolds
number
of
20
is
kind
of
the
transition
point,
so
this
is
with
a
very
simple
system.
They
were
able
to
study,
reynolds
number
and
and
how
things
move
at
low
reynolds
numbers
compared
to
higher
analyst
number
just.
I
thought
it
was
really
neat
because
it's
such
a
simple
system.
A
A
A
A
B
A
A
C
A
A
A
A
E
A
A
C
A
C
A
A
C
A
F
B
A
Okay-
and
this
is
a
bit
busy
but
depending
on
the
actual
shape
of
the
cell,
will
have
it
be
a
kind
of
form
of
solid
tissue
and
if
the
cells
change
shape,
they
become
fluid
and
they've
actually
pinpointed
this.
This
number
as
q.
A
cue
point
where
it's
p,
divided
by
the
square
root
of
the
area
and
p,
is
what
was
that.
A
E
B
A
A
A
A
B
B
B
A
A
Anyway,
but
the
next
slide
is
this:
is
your
robots
and
you
can
make
small
robots
and
they
can
you
make
them
into
a
swarm
and
they
say
that
this
is
not
intelligent
matter.
Exactly
it's
adaptive
matter,
because
someone
externals
controlling
it,
but
you
could
probably
make
something
that
was
sort
of
an
autonomous
system
that
was
made
up
of
small
robots
that
moved,
moved.
D
B
Well,
no,
I
think
that's
pretty
good
yeah.
I
thought
there
was
a
lot
of
good
stuff
in
there
and
I
mean
some
of
it.
You
know
I
mean
it's
what
to
follow
up
on
a
lot
of
good
references
in
there
I
saw
at
least
six
or
seven
good
references
in
there
yeah,
so
this
is
dick's
paper
that
he
put
in
here.
Oh,
this
is
the
kinematics
of
explosively
jerky
diatom
motility.
B
H
B
A
F
A
So
this
is
says
this
instantaneous
tissue
snap
shots,
but
you
look
at
the
shapes.
This
is
the
fluid
and
this
is
the
solid
and
you
can
see
the
difference
in
the
shapes
in
the
cells
and
then
this
indicates
their
movement
like
they.
They
move
a
lot
more
and
then
each
individual
cell
moves
a
lot
more
that
that's
the
motion.
F
A
E
A
A
E
A
B
B
And
of
course
they
have
these
jamming
phase
transitions
where
they
go
from
like
moving
moving
in
a
fast,
coordinated
manner
to
getting
stuck
in
place,
and
so
you
know
there's
a
lot
of
interesting
stuff
in
this
area
that
isn't
really
like
tissue
oriented,
but
it
does
describe
sort
of
these
how
these
collectives
of
cells
move
around
and
shape.
You
know
shape
like.
E
Is
I
call
an
ising
lattice
membrane?
In
other
words,
it
was
a.
It
was
a
lattice
system
in
which
diffusion
occurs
across
a
narrow,
glass,
okay
and
what
happened
as
a
result
of
this,
there
was
a
phase
transition
inside
the
lattice,
and
when
the
phase
transition
occurred,
the
flux
across
the
membrane
suddenly
went
out
exponentially.
B
C
E
B
B
So
this
book
sort
of
you
know
goes
through
a
lot
of
like
a
lot
of
the
soft
active
matter
stuff
at
a
very
early
state,
and
this
physics,
of
course,
is
useful
in
technology
in
building
liquid
crystal
displays-
and
this
is
where
you're
using
you
know
a
certain
approach
to
this.
But
then
this
is
also
very
useful
in
biology
and
especially
like
they've
applied
this
to
bacterial
colonies,
and
things
like
that.
B
So
this
just
kind
of
goes
through
all
that
was
known
at
the
time
about
liquid
crystals
and
kind
of
he's
a
semi
theorist,
so
he
kind
of
develops
some
theoretical
techniques
for
this,
which
you'll
find
in
a
lot
of
the
modern
papers
on
this.
So
you
know
this
kind
of
goes
through
a
lot
of
different
things.
This
is,
you
know,
relevant,
of
course,
to
chemistry
or
to
other
types
of
systems
where
you
have
cells
of
different
shapes.
B
Let's
see,
there's
you
know
from
those
sorts
of
orientations
you
can
find
long
and
short
range
order,
so
he's
kind
of
hitting
on
this
theme
of,
like
you,
know,
local
versus
global
organization,
and
how
that
plays
into
sort
of
the
order
of
the
of
the
whatever
you're
looking
at
it
could
be
a
tissue.
It
could
be
a
an
array
of
something,
and
so
there's
there's
that
theme
and
then
you
can
play
around
with
yeah
what
was.
B
E
B
I'm
just
taking
a
note
here
I'll
have
to
look.
You
have
to
look
at
it.
I
I
I
should
probably
more
effectively
taking
a
short
presentation
on
this
myself
and
pulled
some
stuff
out
of
the
book,
but
I
mean
this.
I
can
send
people
this
book
if
they're
interested
in
going
through
it
it's
more
of
a
reference
yeah.
B
I
E
B
Right
and
then
pneumatodynamics,
so
he
works
out
the
equations
of
pneumatodynamics.
So
this
is
the
dynamical
state,
rather
than
just
looking
at
it
as
a
static
system
and
then
cholesterics,
which
is
where
you
get
into
a
more
complex
geometry
and
then
spectix,
which
are
liquid
and
solid
layers
with
different
orderings
and
all
that.
C
D
B
Well,
thanks
susan,
thank
you
susan
for
that
presentation.
That
was
very
good
and
you
know
in
the
new
year
we'll
probably
try
to
follow
up
on
that
a
bit,
and
if
people
want
to
do
something,
it'd
be
a
little,
you
know
we
can
maybe
do
some
sort
of
simulation
study.
I
know
there
are
some
other
things
that
dick
mentioned
about
the
archaea,
so
that
that
might
be
something
that
we
also
tie
into
it
or
we
can
just
you
know,
I'm
interested
in
seeing
the
archaea
presentation.
B
H
B
B
H
B
So
I've
been
to
transition
a
little
bit.
I
wanted
to
talk
about
a
couple
things
that
are
just
news
items
that
I
found
or
things
I
wanted
to
bring
up.
So
the
first
of
these
is
this
simularium,
which
is
a
nice
tool
from
the
allen
institute
of
cell
science,
so
the
allen
institute
they've
been
working.
This
is
one
of
the
microsoft
billionaires
and
they've.
B
Given
all
this
money
to
this
institute,
the
allen
institute
and
they're
trying
to
do
a
lot
of
stuff
with
like
using
really
high-end
technology
to
find
out
more
about
the
brain
and
about
cells.
You
know,
sequencing
the
dna
and
looking
at
the
microstructure
of
the
cells
and
putting
this
on
to
a
digital
format.
B
So
they,
you
know
they
have
these
different
types
of
things
like
actin
and
clathrin-mediated
into
cytosis,
and
you
have
this
sort
of
simulation
model
where
it's
you
know
like
a
15-second
shot
of
all
the
different
components
of
this
process.
So
this
is
written
in
cytosim,
which
is
a
simulation
language
they
have
cyto
escape.
I
don't
think
they're
related,
but
there's
this
cytosim
yeah.
So
this
is
a
software
that
they've
developed,
and
this
is
the
github
repository.
So
they
have
this
cytosim
software
that
they're
using
to
build
this
model,
and
then
it's
in
this
manuscript.
B
B
You
can
load
your
own
data
here.
They
have
here's.
Here's
spatio-temporal
oscillations
in
this
e
coli
min
system,
so
this
is
where
they
have.
I
think
this
is
written
in
smolden,
I'm
not
sure
what
that
is,
but
this
yeah
smolden
2.1.
So
this
is
from
2010.
This
is
a
platform
that
they
have
the
detailed
simulations
of
cell
biology
with
smolden.
So
this
is
a
very
specific
tool.
B
B
So
this
is
where
this
is
the
min
system,
which
is
used
to
find
the
cell
center
during
cell
division,
so
they
use
this
process
and
they
demonstrate
it
as
the
bacterial
cells.
Turning
in
the
space,
so
I
mean
you
know
you
can
do
a
lot
of
things.
There
are
a
lot
of
different
software
platforms
here
that
they're
archiving
and
it's
all
open.
B
So
you
know
you
can
add
your
own
model
or
run
from
one
of
these
models
and
I'm
not
sure
the
you
know
what
the
how
hard
it
is
to
get
some
of
these
models
to
run.
Even
if
they're
open
it
doesn't
mean
that
they'll
run
well,
so
yeah,
here's
some
more
here
on
on
cellular
biophysics,
so
this
is
this
is
also
represented
here.
B
B
E
B
B
Yeah
anyways
all
right,
so
then
the
rest
of
the
yeah.
So
this
kind
of
shows
you
kind
of
this
works.
You
have
a
json
file,
you
load
it
in.
You.
Have
this
engine
specific
trajectory
file,
you
use
it
with
a
python
converter
or
you
can
use
a
digital
notebook.
The
simulation
engine
writes
to
the
file
format,
and
then
this
is
so.
This
is
based
on
this
simulation
engine.
B
B
B
It's
called
new
clues
about
the
origins
of
biological
intelligence,
so
common
solution
is
emerging
in
two
different
fields:
developmental
biology
and
neuroscience.
So
this
is,
you
know
mike
levin.
Does
this
work
with
sort
of
developmental
systems
and
non-neuronal
cognition,
and
also
that
you
know
just
regular
cells?
Not
just
neurons?
Have
electrical
potentials,
maybe
not
action
potentials,
but
they
have
this
electrical
activity
that
contributes
to
regeneration
and
developmental
organization,
and
things
like
that,
so
you
know
they
kind
of
write
up.
This
sort
of
this
is
a
popular
science
article.
B
So
there
they
kind
of
propose
that
intelligence
is
a
purposeful
response
to
available
information,
often
anticipating
the
future.
This
is
distributed
throughout
biology
at
many
different
spatial
and
temporal
scales.
So
this
means
that
you
know
they're
kind
of
expanding
this
definition
of
intelligence.
B
But
what
they're
arguing
is
that
intelligence
is
much
more
basic
and
it's
this
intelligent,
purposeful
problem
solving.
So
you
see
this
in
cell
tissue,
single
cells
and
tissues
in
neurons
and
networks
of
neurons
in
viruses
and
ribosomes
and
rna
fragments
and
so
forth.
So
you
see
this
at
different
scales,
and
so
they
propose
that
the
origin
of
intelligence
is
the
central
problem
in
biology.
B
So
this
is
probable.
This
is
probably
a
pretty
controversial
view,
but
you
know
that,
like
the
way
they're
casting
the
net
here
is
maybe
a
bit
provocative,
but
there
are
probably
a
lot
of
interesting
questions
surrounding.
You
know
this.
This
idea
that
you
know
there's
this
sort
of
behavior,
that's
coordinated,
and
then
how
do
you
propose
this?
B
B
Do
they
use
that
autonomy
and
coordinate
their
behavior
to
create
tissues,
and
then
those
tissues
are
coordinated
into
an
organism?
And
things
like
that,
so
you
know
it's
really
kind
of
opens
up
a
number
of
interesting
questions.
So
you
know
these
are
things
that
again,
like
yes,
gets
pretty
philosophical,
but
there
are
a
lot
of
interesting
experiments.
One
could
do,
but
I
think
it's
interesting
that
it's
kind
of
like
focusing
on
this,
this
common
solution
from
two
different
fields,
developmental
biology
and
neuroscience.
B
B
It's
fit
these
sorts
of
things,
and
this
is
all
modularity,
because
these
modules
will
form
an
evolution
and
they'll
tend
to
be
conserved.
We
talked
about
hox
genes
and
how
hox
genes
are
conserved
in
their
co-linearity
and
their
function.
So
you
know
you
get
these
different
segments
in
an
organism
that
get
conserved.
You
also
have.
B
B
I'm
not
sure
they
get
to
the
third
well
they're
talking
about
hierarchy
a
lot,
but
I
don't
know
what
the
third
thing
is
in
in
this
list
about
pattern:
completion,
so
pattern.
Completion
is
where
activation
of
part
of
the
system
turns
on
the
entire
system
in
our
apartment.
Building
the
family
would
have
one
central
figure.
Let's
say
one
of
the
parents
would
represent
the
families
in
meetings
and
engaged
when
it's
needed.
E
B
Aspect
of
it
in
in
vision
where
you
have
okay
yeah,
so
you
usually,
you
have
yeah
in
a
visual
system.
You
know
you'll
get
like
these
lower
level
process,
they'll
get
lower
level,
processing
of
images
and
then
that'll
lay
out
the
pattern.
But
then,
if,
if
there's
something
that's
missing,
if
it's
an
ambiguous
pattern
or
something,
then
you
know
they're
higher
up
processes.
That'll
fill
in
the
gaps
in
that,
and
so
they've
shown
this
in
a
lot
of
psychology.
Experiments
where
you
know
they'll
give
people
a.
E
B
B
Okay,
yeah,
okay
and
I
put
the
link
to
this
article
in
the
chat
as
well.
Finally,
thank
you.
Oh
yeah,
you're
welcome.
Finally,
this
is
for
people
who
are
interested
in
machine
learning,
maybe
more
but
there's
some
interesting
papers.
So
this
new
ips
conference
is
going
on
right
or
it
went
on
last
week.
B
So
there
are
a
lot
of
different.
This
is
a
huge
conference
on
machine
learning
and
deep
learning,
and
so
I
know
that
quran
is
probably
pretty
interested
in
this,
but
this
is
the
top
10
papers
and
then
the
best
papers
of
the
conference.
So
I'll
put
this
in
the
chat-
and
this
is
so
they
in
this
article
they
give
their
opinion
about
some
of
the
maybe
the
best
papers
at
the
conference.
B
This
is
the
number
of
papers
that
are
published
at
this
conference
over
time,
so
it's
really
growing
over
the
last
10
years,
quite
a
bit,
and
so
a
lot
of
these
papers
are
on
archive.
There
are
a
lot
of
different
topics
that
are
being
explored
in
the
community
right
now.
You
know
there
are
things
like
gradient
starvation,
which
is
a
way
to
look
at
like
different
ways:
neural
networks
use
gradient
descent
and
how
to
improve
upon
that
process.
B
B
They
have
variational
base,
reinforcement,
which
is
you
know,
bayesian
techniques
for
predicting
you
know,
patterns
or
predicting
categories,
contrastive
losses,
mlp
architectures,
which
are
special
new
approach.
I
don't
remember,
remember
what
mlp
means,
but
this
is
like
these
are
some
of
the
newer
methods
revisiting
resnets,
so
people
are
innovating
in
old
models
and
then
there's
some
good
papers
in
here
where
you
have
the.
Why
of
why?
They
did
this
paper
and
the
key
insights.
B
B
B
B
So
these
a
combination
of
single
frame,
reconstruction
loss,
which
is
a
segmentation
network
and
then
a
motion
network
that
sort
of
takes
video
and
you
know,
segments
the
images
and
then
looks
at
the
motion,
cues
and
then
kind
of
integrates
that
so
they
use
a
reconstruction
and
a
motion
map
and
then
they're
able
to
do
these
self-supervised
segmentations
in
the
paper.
So
and
then
so
there
are
a
number
of
other
ones:
language
models,
retrieval
models,
you
know
again
more
stuff
with
video,
actually
multimodal
processing
multimedia
processing,
the
vatt
model.
B
B
Taking
apart
videos
and
finding
features
in
them,
so
we
usually
talk
about
microscopy
images,
but
you
also
have
these
other
types
of
data
that
might
be
useful
in
other
types
of
videos,
and
so
I
don't
know
if
there's
we
have
an
equivalent
in
in
looking
at
like
biological
images,
I'm
sure
there's
a
lot
of
auxiliary
information
that
we
could
extract
using
sort
of
a
multimodal
approach,
but
so
yeah
there's
a
lot
more
here.
I
could
go
through
all
day.
Probably
here's
koran
hello,
quran,
yeah
yeah,
thanks
for
attending
nick
okay.
J
B
B
J
J
J
Yeah,
so
the
thing
is:
there's
still
a
lot
of
noise.
You
know
within
the
image,
and
I
think
I
need
to
I
know.
Maybe
I
think
if
we
have
playing,
I
was
thinking.
Maybe
if
you
know,
if
we
had
more
like
this,
is
this
is
what
I'm
trying
to
do
right
right
now.
I'm
just
trying
to
you
know
zero
in
on
the
actual
object
here.
Yeah
and
I'm
trying
to
you
know,
reduce
the
background
noise.
Is
there,
so
this
process
will
take
some
more
time
because
there's
still
a
lot
of
noise
that.
B
A
A
Anyway,
yeah,
like
I
said
I,
I
really
would
like
to
get
some
more
images
this
this
christmas.
Usually
the
salamanders
lay
this
time
of
year
and
I
have
that
new
microscope.
So
hopefully
I'll
get
some
some
nice
images
because
I
certainly
will
yeah
and
I
can
try
to
look
up
some
more
that
are
earlier
stages
where
you
can
actually
see
the
cells.
That's
an
idea
as
well.
C
A
B
J
J
So
I
think
this
this
thing
is
still
going,
but
the
thing
is
because
there's
a
lot
of
averages
involved
right
by
because
I'm
getting
rid
of
a
lot
of
data
within
the
processing.
Instead,
I've
had
this
doubt
related
to
the
you
know
how
to
get
that
like
what
are
the
valuable
influences
that
have
been
getting
from
that.
B
Okay,
well,
susan
has
a
couple
papers
on
the
on
the
technique
that
are
kind
of
you
know
kind
of
talk
about
the
you
know
how
the
data
is
captured,
so
I
mean
what's
going
on,
is
you're
you're
getting
this
cell
and
it's
rotating
and
so
you're
able
to
see
the
surface
of
the
cell?
And
if
you
had
you
know
another,
you
know
multiple
cells,
then
you
would
be
able
to
see
sort
of
like
how
that
process
of
cell
division
happens
at
very
early
stage
of
the
embryo.
A
J
A
A
See
it
develop
but
then
it
well
then
it's
some
opaque
material
does
roll
over
and
cover
the
whole
thing
it
does.
But
that's
why
I
wanted
to
use
infrared
light
to
image
it,
so
I
could
see
more
of
it
because
you
can
see
down
into
the
tissue
with
the
infrared
light,
but
that's
on
hold,
I
think,
but
I'll
I'll
try
to
get.
B
B
B
A
J
J
J
B
J
B
B
J
A
I'm
working
with
optical
clearance
tomography
and
we're
trying
to
do
elastography
with
it
like
the
viscoelasticity
of
tissue,
and
to
do
that
you
have
to
perturb
the
tissue
with
by
pushing
it
in
some
way
or
pulling
it
and
then
measure
its
change
and
that's
the
optical
clearance
tomography
would
measure
the
change
in
the
tissue
when
when
that
occurred,
so
that's
actually
what
I'm
trying
to
work
on,
but
I'm
not
working
on
the
axolotl
leg,
partly
because
they're
very
difficult
to
image.
You
can
see
the
blastocyle
develop
like
that's
the
hollow
portion
in
the
middle.