►
Description
Physical Intelligence in Development: regulation of dynamic tension and sensing as intersecting networks. Presented at Dynamics Days 2023. Presenter: Bradly Alicea
A
The
development
of
a
complex
organism
typically
involves
two
processes.
The
first
is
a
series
of
differentiation
events,
and
these
are
summarized
in
the
tree
at
the
right.
This
is
a
binary
tree
that
has
branching
events
throughout
its
topology
and
is
organized
in
terms
of
Developmental
time.
So
the
farther
we
go
in
developmental
time,
the
more
division
events
we
have.
A
The
second
process
is
the
emergence
of
intersecting
subnetworks,
So
Below.
You
can
see
cell
tracking
data
that
are
presented
in
two
dimensions
and
three
dimensions.
These
are
newly
differentiated
neuronal
cells
and
the
C
elegans
embryo
shown
in
the
red
and
the
blue,
but
we
also
have
developmental
cells
or
embryonic
cells
that
are
in
the
transparent.
A
So
these
two
networks
overlap
and
intersect.
One
way
to
think
about
this
intersection
is
this
concept
of
generative
Divergent
integration,
and
we
can
use
a
toy
model
to
understand
how
this
emerges
and
development.
So
on
the
left.
We
have
an
eight
cell
example,
which
is
something
that
shows
the
emergence
of
something
we
call
an
embryo
Network.
This
embryo
Network
emerges
from
cells
based
on
their
proximity
to
one
another
and,
as
you
can
see
the
eight
cell
stage,
you
have
these
disconnected
networks
that
only
involve
one
cell
type
at
the
right.
A
We
have
the
24
cell
example
and
the
24
cell
example
has
two
cell
types
and
it's
a
bit
more
crowded
in
terms
of
cells.
So
there's
more
more
connectivity.
The
network
is
continuous,
but
we
also
see
the
Divergence
of
neuronal
cells
in
green
and
developmental
cells
in
blue
and
some
of
these
developmental
cells
will
become
somatic
cells
and
turn
into
their
own
sub
networks.
That
will
eventually
have
less
connectivity
between
other
group
different
groups
of
cells,
but
the
connection
between
different
groups
of
cells
will
be
maintained.
A
So
we'll
consider
the
relationship
specifically
between
tensegrity
structures
that
form
the
generic
animal
body
and
the
network
of
neuronal
cells
that
exist
within
this
generic
animal
body.
So
we're
not
going
to
be
talking
about
any
specific
organism,
we'll
be
talking
about
a
generic
organism
as
an
example
of
this.
This
forms
the
connectome
and
forms
the
organism.
The
developing
organism,
as
both
of
these
networks,
emerge
from
an
embryonic
precursor.
They
have
critical
intersection
points,
as
shown
in
the
Divergent
integration
model
in
the
last
slide.
So
in
this
case
we
see
a
very
simple
example
of
this.
A
We
have
say,
for
example,
our
tensegrity
structure
was
represented
by
the
sixteagon
and
hairball
Network,
which
represents
the
connectome
and
those
overlap
in
critical
ways
shown
in
the
context
of
an
organism
on
the
right.
We
have
we're
a
generic
organism.
We
have
this
connectome
in
the
front
at
the
anterior
under
the
organism,
we
have
a
midline,
and
then
we
have
these
structures
that
are
linked
60-egons,
which
serve
as
this
tensegrity
network.
They
give
the
organism,
stability
and
shape
so
biological.
A
First,
top
thing
we
have
to
cover
is
this
idea
of
biological
tensegrity,
so
biological
tensegrity
Is
Not
A,
New
Concept.
It
was
somewhat
popular
in
the
90s
biological
tensegrity,
making
the
cover
of
Scientific
American.
We
see
at
the
left
and
on
the
right.
We
see
cells,
I,
think
these
are
fibroblasts
and
these
cells
have
the
structural
Integrity
that
has
to
be
maintained,
as
the
cell
changes
shape
it
moves
around.
So
you
can
see
that
there's
a
calculation
of
forces,
as
the
cell
adheres
itself
to
a
surface
and
moves
around
freely.
A
A
So
this
is
an
example
of
a
tensegrity
network.
We
talked
about
the
60
agon
already.
This
is
a
sphere.
This
is
a
something
that's
stable,
that's
round,
so
it's
biological
in
that
sense.
So
we
have
that
type
of
structure,
but
we
also
have
other
types
of
structures
like
the
psychosahedron
that
is
20
a
gun
on
the
right,
and
this
is
something
will
actually
work
through.
A
How
this
is
this
network
is
characterized
mathematically
in
a
little
bit,
but
we
can
see
that
we
have
these
nodes,
which
are
these
structural
pieces,
and
then
these
connections
are
these
arcs,
which
are
the
things
that
introduce
the
tensions
and
loadings
of
forces.
So
you
can
see
how
this
thing
is
held
in
in
stability
by
being
organized
in
a
way
that
allows
us
to
have
stability,
Dynamic
stability
of
dynamic
tension.
A
So
tensegrity
networks
are
defined
as
Dynamic
tension
introduced
as
a
series
of
rigid
and
discrete
compression
elements
joined
together
by
continuous
tension
forces.
So
you
can
see
that
we
have
these.
This
network
is
discrete,
but
it's
also
rigid
because
that's
held
together
through
these
forces,
so
the
discrete
components
are
the
nodes
and
the
rigidity
is
introduced
by
the
arcs
and
the
forces
along
those
arcs.
A
So
the
network
topology
then
leads
to
structures
that
are
Ultra
stable
resilient
to
shape
affirmations
and
are
highly
modular.
So
the
psychosahedron
on
the
on
the
right,
as
well
as
the
60,
a
gun
on
the
left,
are
extremely
stable
structures.
They
can
be
modular,
so
they
can
break,
can
be
decomposed
into
different
sub
components
and
they're
resilient
to
different
shape,
deformations,
which
you
see
a
lot
of
in
development,
so
in
cells,
specifically
tensegrity
networks
can
form
a
cytoskeletal
configuration
with
an
in
between
cells,
the
interface
of
cytoskeletal
elements
being
the
site
of
force
balance.
A
So
we
can
see
a
similar
process
in
this
figure.
This
is
actually
in
an
architectural
structure
where
we
have
these
components
that
are
organized
along
an
axis
they
get
put
together.
In
this
sort
of
structure,
and
then
they
can
be
bent
and
they
can
bend
and
underway
in
all
these
other
things.
So
if
you,
you
know,
if
you
want
to
have
a
structure,
that's
resilient
over
a
number
of
Transformations
and
you
want
to
have
some
sort
of
tensegrity
network.
A
So
this
is
the
icosahedron.
This
is
the
mathematical
representation.
We
usually
use
a
connectivity
Matrix,
we
denote
d
sub.
I
j,
and
this
is
a
way
to
represent
this
network
as
a
series
of
connections.
This
will
become
important
when
we're
looking
at
the
intersection
with
connectoms
a
little
bit
later
on,
but
these
are
discrete
edges.
So
the
edges
are
these
lines
and
the
edges
are
either
zero,
which
is
a
passive
force
or
one
which
is
an
active
Force.
A
The
active
force
is
keeps
these
things
in
Dynamic
tension,
but
also
sometimes
at
some
points
in
time.
There
needs
to
be
a
passive
force
on
that
connection,
so
that
these
nodes
can
change
their
position
relative
to
one
another.
So,
as
you
can
see,
we
have
a
three-dimensional
coordinate
system.
Much
like
we
had
for
the
cells
in
the
embryo
and
those
coordinates,
can
change
over
time,
and
so
do
these
connections
between
them.
These
edges,
biological
edges
are
flexible
over
time,
so
they
can
have
a
value.
A
Now
we'll
talk
about
the
connectome
and
give
different
examples
of
a
connectome.
A
lot
of
people
are
familiar
with
the
Hairball
Network
and
the
Hairball
network.
Connectome
is
a
connectome
with
a
lot
of
connections.
Some
of
them
are
ordered
hierarchically.
Some
of
them
are
ordered
somewhat
randomly,
and
this
is
the
type
of
connectivity
we
expect
out
of
a
connectome.
This
is
an
example
of
a
generic
air,
ball
Network,
so
it
doesn't
point
to
any
one
connectome
on
the
upper
right.
A
We
have
the
labeled
C
elegans
connectome,
which
has
actual
C
elegant
cells
that
are
labeled
according
to
nomenclature,
and
you
can
see
the
structure
where
it's
a
little
bit
more
mature
than
the
one
shown
in
the
in
the
first
two
slides,
but
this
one
has
some
structure
going
from
head
to
tail
and
down
the
center
of
the
body,
so
we
can
see
that
there's
some
anatomical
structure
to
it.
We
also
have
the
fly
connectome,
which
has
recently
become
available.
A
The
C
elegans
connectome
has
been
available
for
quite
a
while
at
the
cellular
level
and
actually
at
the
connection
level,
so
we
know
from
which
cell
connects
to
which
cell,
but
in
flies.
We
also
have
this
type
of
connectome,
where
we
can
trace
the
connections
between
cells,
and
this
is
a
larger
number
of
cells
in
a
larger
number
of
connections.
As
you
can
see,
it's
still
the
same
thing.
You
have
the
structure,
that's
anatomically,
specific
structured,
so
these
are
small.
A
Animal
connectives
are
actually
quite
easy
to
work
with
the
reason
I
bring
this
slide
up
is
because
it's
interesting
the
kind
of
data
that's
available,
so
in
poly
Keats,
which
are
small
worms.
There
are
these
very
small
nervous
systems
of
under
100
neurons,
and
you
can
map
those
out
and
Trace
the
connections
between
them.
We
also
have
other
connect
domes
from
other
organisms:
platy
nurses,
siona
and
c
elegans,
and
so
all
of
those
networks
are
available
in
different
ways
in
the
literature.
So
we
could
build
these
networks
from
the
literature.
A
So
the
neural
connectome
consists
of
cells,
of
course,
connected
via
electrochemical
action
and
maintains
information
processing
between
the
body,
surface,
the
periphery
and
the
central
nervous
system.
So
we
can
see
at
the
upper
left.
We
have
an
example
of
this
nervous
system.
We
have
cells
and
we
have
the
nerves
that
come
out
and
we
have
this
conforming
to
an
adult.
A
I
j,
but
this
is
C
sub.
I
j,
and
this
is
just
the
connectivity,
whether
it's
positive
or
negative,
from
one
cell
to
another.
So
this
is
the
navigation
circuit.
We
have
I,
have
I,
think
11
cells
by
11
cells
in
this
Matrix
and
I
have
positive
and
negative
marks.
So
those
two
we
have
both
of
those
types
of
data
there
are
available
as
matrices.
A
A
That's
embodied
much
as
like
our
connectome,
and
it
goes
from
some
sort
of
input,
some
sort
of
sensory
input
to
an
output
and
so
put
aside
what
we
know
about
connect
ohms
for
now
and
just
think
about
the
path
between
a
sensory
organ
or
a
sensory
cell
and
an
effector
cell,
which
could
be
an
effector
organ,
something
like
a
fin
or
or
maybe
like
muscles.
And
so
we
have
these
paths
between
the
input
and
output.
Basically,
and
we
can
embody
these
like.
A
We
did
with
the
connectome
in
the
previous
cases,
and
we
can
also
have
signals
that
pass
from
the
input
to
the
output.
Now.
This
is
a
very
simple
example
of
an
embodied
network,
but
this
is
what
we
want
to
get.
We
want
to
move
towards
a
little
bit
more
complex
model
of
this,
so
we
can
actually
put
introduce
interneurons
which
are
processing
units
in
between
the
input
and
output.
A
Sometimes
they
summarize
different
Paths
of
information,
sometimes
they
disperse
them
across
the
number
of
cells
or
interneurons,
and
so
this
is
a
way
we
can
build
a
more
complex,
Network,
that's
embodied
and
that
maybe
can
actually
do
some
sort
of
behavior.
So
these
are
informational
relays
that
can
be
go
that
go
from
input
to
output,
and
so
we
can
embody
these
networks
in
that
way
now
we
can
also
not
just
use
them
as
networks,
but
also
a
hypergraph.
A
So
in
the
case
of
hypergraphs
we
have
nodes
that
have
a
number
of
cells
within
them
and
they
have
a
diverse
set
of
properties,
and
they
also
have
a
diverse
set
of
connectivities
that
we
can
also
summarize.
But
in
the
hypergraph
case
here
we
have
the
ability
to
characterize
a
sets
of
cells
that
contribute
to
salic
and
organ
or
tissue
and
use
that
as
a
single
node.
A
So
we
can
actually
do
some
really
interesting
thing
things
with
that,
and
so
we
can
actually
Orient
these
networks
towards
Behavior,
Behavioral
or
sensory
inputs
and
behavioral
outputs,
and
we've
talked
more
about
this
in
this
pre-print
listed
here
in
each
hypergraph.
Node
is
the
structure
or
tissue
type
and
it
generates
a
spectral
graph
of
cell
type
properties.
So,
at
the
end
of
the
day,
we
can
take
each
one
of
these
nodes
and
get
cell
type
properties
out
of
them.
This
may
describe
multiple
scales
of
differentiation
and
precursors
to
phenotypic
change.
A
What
we
can
also
do
is
it
can
describe
its
relationship
to
this
other
network,
this
Integrity
Network,
and
it
can
interact
with
that
Network.
So
this
is
an
example
of
where
we
are.
We
have
the
connect
and
we
have
this
tensegrity
network-
it's
organized
in
an
embodied
manner,
from
anterior
to
posterior
and
in
a
left
to
right
symmetry
with
the
midline.
A
Now
this
is
the
interesting
part
when
these
two
networks
intersect,
so
we're
connectome,
C
sub,
I
j
and
our
tensegrity
Network
d
sub.
I
j
intersect
in
different
ways,
and
this
is
just
showing
a
very
simple
case
where
these
two
parts
of
the
network
or
the
Matrix
intersect,
and
we
can
see
this
in
in
biological
systems,
and
you
know
these
intersections
come
in
the
form
of
something
like
appropriate
scepter.
A
So
this
example
is
of
a
skin
proprioceptor
and
an
escanosusceptor,
and
so
this
shows
examples
of
these
different
sensory
organs
at
the
surface
of
the
body,
and
the
body
of
course,
is
being
held
in
rigidity
and
there's
information
processing
going
from
the
surface
of
the
body
into
this
nervous
system.
And
then
it
goes
on
to
generate
a
behavior.
So
you
can
see
that
there's
this
network
that
comes
out
of
this
out
of
the
surface
and
down
into
a
internal
connectome
and
it's
generating
Behavior,
which
in
turn
affects
the
structure.
The
stability
of
the
tensegrity
network.
A
A
This
nervous
system
Network,
and
you
can
imagine
that
there's
some
sort
of
sensory
input
at
the
side
of
the
worm
or
at
the
front
of
the
worm
and
it's
being
transduced
into
the
worm,
it's
nervous
system
and
then
that
nervous
system
causes
behavior
and
that
affects
the
stability
of
the
organism.
It
affects
the
stability
of
the
internal
Network
and
so
forth.
A
So
we
can
see,
as
this
changes
its
perspective,
that
you
get
different
views
of
things
and
you
can
see
how
that
those
two
structures
interact
physically,
but
we
can
also
look
at
this
in
terms
of
a
systems
context,
so
they're
appropriate
receptors
around
the
left.
You
have
an
input
of
sensation,
they're
transduce
forces
that
come
into
the
organism,
and
this
this
these
transduced
forces
come
through
the
tensegrity
network.
So
these
cells
here
represent
a
Loosely
organized
tensegrity
Network.
A
They
can
shift
their
position
as
there
is
growth
as
there's
regeneration
as
there's
movement
of
the
organism,
and
those
proprioceptors
are
just
inputting
a
lot
of
sensory
information
that
gets
transduced.
But
of
course,
then
we
also
have
our
connectome
and
the
connectome
is
sending
signals
out,
especially
as
it's
starting
to
develop
and
form
connections
with
the
neural
cells.
So,
as
I
told
you
before,
there
is
this
Divergent
integration.
A
Some
of
this
integration
comes
in
this
form
where
you
get
these
interactions
between
the
structure
of
the
body,
the
sensory
inputs
and
the
outputs
of
a
connectum
there's
some.
In
addition,
there's
cell
growth
and
shape
deformation
and
a
response
to
forces
which
is
action
from
this
emerging
connect
to
them.
A
What's
interesting
is
that
we
expect
that
in
development,
these
interactions
move
from
being
so
autonomous
or
reflexive
to
goal
directed
or
being
the
product
of
the
connect
Dome
entirely.
So
in
development,
the
cells
move
around.
They
have
a
lot
of
motility
in
the
embryo
and
then
later
you
have
muscles
that
sort
of
twitch
without
being
connected
to
anything
and
then
eventually
the
connectome
is
actually
managing
a
lot
of
these
interactions.
So
it's
a
very
interesting
set
of
interactions
going
on,
so
we
can
actually
go
beyond
the
linear
case.
A
A
A
Finally,
we
can
look
at
this
in
a
broader
from
a
broader
lens
as
what
they
call
4E
cognition.
So
we
can
look
at
this
in
terms
of
embodiment,
but
also
an
action
embeddedness
and
extended
cognition,
and
we
don't
think
that
embryos
exhibit
cognition
where
we
can
look
at
information
processing
using
some
of
the
things
from
this
literature.