►
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
C
A
A
A
Okay,
so
so
in
various
theories
of
object,
recognition,
including
capsules
and
including
the
things
I've
created
posters
about
what
gets
represented,
as
is
like
what
you're,
what
you're
viewing
and
like
where
it
is
or
equivalently.
Where
is
your
sensor
and
the
reference
frame
of
the
object?
So
I
mean
right
now,
it's
changing
and
and
basically
your
er
actually
representing
how
far
away
something
is
and
a
little
bit.
D
For
and
that's
that's
also,
essentially,
what
capsules
tube
capsules
are
will
be
like
a
feature
with
some
pose
and
that's
essentially,
this
is
a
feature
and
where
it
is
in
3d
space.
My
needs
consist
of
all
you
can
have
a
cap,
a
where
gatherer
is
invitation
up
where
something
is
relative,
jovani,
yeah
yeah,
it's
not
yeah.
D
A
D
B
A
So,
where
I'm
coming
from
is
like
I
was
I
was
walking
around
thinking
like
what
I
was
what
was
coming
into.
My
visual
stream
I
was
riding
the
train
and
watching
houses
in
the
distance
and
the
idea
that
I
was
when
I
see
this
house
in
the
distance
on
top
of
a
hill.
That
I
am
representing
the
distance
from
my
retina
to
that
it
was
just
insane
to
me
the
way
that
I'm
recognizing
this
house
on
a
hill
in
the
distance
of
your
neighborhood
on
the
hill
in
the
distance.
D
D
A
So
the
end,
so,
let's
think
of
objects
as
spatial
arrangements
of
components
or
child
objects
or
whatever
you
want
to
call
them.
So
I
have
a
strong
suspicion
that
we
recognize
spatial
arrangements
of
components
in
a
way
that
is
really
an
independent
of
how
far
away
those
components
are
yeah.
So
so
I'm
working
on
that
problem
of
building
up
those
types
of
objects.
D
A
So
start
reading
up
the
board
yeah
a
point
of
view.
The
brains
object,
modeling
system
stores,
like
the
relative
sizes
of
child
objects,
not
third
absolute
size
and
I'm,
going
to
talk
about
sizes
and
distances
is
all
being
had
to
one
looking
at
one
scale-invariant
coordinate
system,
so
learning
how
to
talk
about
it,
but
here
I'm,
showing
a
set
of
these.
These
child
objects,
these
components
that
are
in
a
certain
arrangement
I'm
going
with
kind
of
this
toy
of
Coach
or
just
simple
shapes
and
arrangement
just
to
get
this
off
the
ground.
A
So
likewise,
just
as
like
you,
okay,
so
your
brain
sees
this
and
actually
one
way
to
frame
what
I'm
doing
is
is
also
trying
to
explain
why
our
brain
can
just
look
at
photographs.
Why
we
can
look
at
a
picture
of
why
we
can
look
at
drawings
and
see
something
coherent,
I
think
I'm
also
sort
of
attacking
that
problem.
When
you
look
at
this,
drawing
of
these
three
shapes,
you
have
no
notion
of
how
large
these
are
and
like
am
I
showing
skies
skyscrapers
here.
Am
I
showing
like
Legos?
A
D
A
D
D
A
C
Looking
to
the
transition
object
allows
you
to
create
actions
like
a
psychotic
or
the
location.
This
cause
we're
going
to
be
size.
Invariant,
it's
the
same
like
two
degrees
in
Chicago
or
West,
is
going
to
give
you
singing
the
visual
input
and
zero
in
one
regardless
of
where
it
is.
Our
way
is
the
appearance
and
render
of
the
same
size.
So
your
locations.
C
A
Different
visual
image,
so
yeah
so
one
year
when
you're
moving
around
and
predicting
how
it's
changing.
At
that
point,
you
need
at
that
point,
you
do
know
the
actual
size
of
woods
in
front
of
you,
not
necessarily
your
you're
on
this
sort
of
sailing.
Storing
than
anywhere
you
might.
You
might
see
the
object
immediately
and
for
how
large
it
is,
and
everything
like
that
and
represent
that
in
activity,
and
that's
how
that
prediction.
A
very
narrow
version
of
this
task,
which
is
like
fixed
observer,
are
just
visual
recognition.
All
you.
A
So
what
I
think
is
when
you
see
something
like
this,
you
do
learn
the
spatial
relationships
of
the
components
but
they're
in
a
different
kind
of
unit
they're,
not
a
or
some
equivalent
evaders.
There
they're,
essentially
like
you
know
that
you
know
that
this
this
this
or
something
is
this
distance
from
this
is
that
you
don't
know
how
large
it
is
in
terms
of
meters,
but
you
know
that
it's
about
you
know
about
the
height
of
this
is
about
how
far
away
they
are,
or
it's
like.
A
It's
like
I,
don't
know
like
four
widths
of
in
a
way
that
you
can
tell
like
if
you're
looking
far
off
into
the
distance
of
the
set
of
buildings
or
if
you're,
looking
at
something
that's
nearby
you,
you
can
tell
the
relative
arrangement
of
things
in
a
way
that
it
is
sort
of
relative
to
its
own
sizes.
Yeah
I
mean.
D
Pete
the
problem
here
we
have
abstract
shapes
that
have
no
inherent
size,
so
in
this
case
I'm
gonna
remember
that
arrangement
that
I'm
doing
it
not
with
physical
objects.
But
it's
a
picture
like
this
I
can't
tell
what
it
is
then
you're
right,
I
guess
I'd
have
to
agree
in
this
case,
but
that
is
you
painted
a
case
where
this
is
you
not
you're
forcing
me
to
do
it
that
way
or
like.
D
The
car
is
an
absolute
absolute
dimensions
to
it,
so
the
model
of
the
car
is
stored
in
those
dimensions.
That's
what
I'm
arguing
and
Department
I'm
inferring
it
there's
a
separate
question.
Do
I
do
I
am
I
looking
at
you,
the
two
tires:
are
they
the
absolute
size?
Also
I,
don't
know
model
itself.
The
first
sensor
is
object.
Modeling
system
me.
That
means
the
internal
representation
for
what
the
car
is,
and
that
seems
to
be
it
habit
sizes
where
I'm
inferring
and
looking
at
something
that
those
don't
come
into
play
tomorrow.
D
A
A
Okay,
so
I
guess
I'll
just
read
this
off
and
continue
to
the
next
part
so
for
purpose
I'm
making
these
maybe
over
broad
statements,
but
for
purpose
of
inferring
an
object.
So
to
answer
that
before
the
sentence,
we
don't
represent
the
distance
from
the
retina
to
the
object.
They
give
an
image
like
cube
or
cylinder,
activates,
the
same
representation
and
say
v1
or
individual
string,
whether
it's
one
meter
or
one
mile
away,
but
we
do
detect
and
store
the
3d
arrangements
of
the
child
objects.
A
So
I'm
making
a
statement
about
2d
versus
3d
is
kind
of
its
kind
of
strange
at
the
first
stage,
if
I'm
getting
less
right
that
that,
like
you,
see
like
a
cylinder
in
your
original
string
and
you're
going
to
represent
it's
like
3d
orientation,
yeah
you're
gonna
represent
us
through
orientation,
but
you're
not
really
representing
its
distance
away
from
you.
So.
A
A
D
A
D
A
Yeah
or
the
third
possibility
is
you,
you
choose
some
some
higher
objects
that
to
symbolize
all
three
of
these,
like
I,
don't
know,
choose
some
gain
access
and
and
figure
out
where
they
are
relative.
To
that
all
these
any
of
these
could
be
right.
It
could
be
some
fluid
system
that
eventually
moves
from
here
to
here
and
goes
from
here
to
everything.
I'm
saying
here
doesn't
depend
on
which
of
these,
which
of
these
different
trade-offs,
young
each
of
those
yeah
yeah
I
mean
yeah.
A
A
A
So
my
bet
is
that
when
you
see
stuff
off
in
the
distance,
although
you
might
not
have
very
a
very
precise
notion
of
how
far
it
is
away,
you
can
use
cues
you
can
use
like
here.
I
drew
them
as
like
these
dotted
lines,
but
you
have
these
various
clues
for
figuring
out
how
far
away,
how
are
they
actually
arranged
in
3d?
So,
even
though
they're
just
a
set
of
like
cylinders
and
prisms
on
your
retina
that
are
in
various
like
retinotopic
sizes,
you.
B
A
A
When
you're,
either
psychotic
or
a
component
to
component
or
when
you
see
the
scene
of
components,
what
you're?
This
is
what
you're
detecting
is
like
the
object.
Identities
are
the
child
object
identities
and
like
their
spatial
arrangements
and
the
relative
sizes,
and
this
that's
the
information
that
you
want
to
kind
of
save
away.
A
A
B
A
C
A
C
A
A
B
A
A
Is
this
further
back
and
larger
and
in
order
to
store
that
kind
of
so
so,
as
suppose,
I
I
tell
you
like,
okay
with
it,
what
this
is
actually
is
two
cylinders
and
this
one's
actually
behind
this
one,
this
one's
actually
far
back,
then
your
model-
if
you
were
right
now,
learning
this
composite
object,
your
model
of
it
would
need
to
keep
track
of
like
okay,
so
there's
there's
a
cylinder
and
then
way
back
behind
it.
There's
a
larger
cylinder,
so
you
need
to
store
way
back
behind
it
and
it's
larger.
A
What
was
it
when
I
was
when
I
was
intersecting
with
this,
with
these
models
of
object,
recognition
that
I
know
like
capsules
and
the
stuff
I've
proposed
and
I
was
very
uncomfortable.
I
was
like
there's
no
way
that
I'm,
using
my
like,
whatever,
instead
of
grid
cell
modules,
to
figure
out
where
my
sensor
is.
D
D
A
D
So
I
am
on
that
clock.
I
know,
I
know
it's
general
sighs
I
mean
I,
it's
not
a
relative
size
that
I
know
I
shouldn't
notice.
Absolute
signs
have
been
done
thanks
to
it
and
he
showed
it.
They
mean
different,
different
positions
all
by
saliva
it's
time
for
what
it
is
that
we're
done.
I,
don't
see
why
you
have
to
abandon
that
idea.
D
D
What
I
constantly
battled
with
this
thing
issue
I'm
talking
about
Marcus
but
and
I
first,
did
it
with
the
melody
thing
right
and
don't
start
with
that,
because
it's
you've
got
you.
You
actually
remember
that
you
remember
the
app
you
remember
the
absolute
pitch
of
melody,
but
you
can
recognize
it.
You
can
infer
it
and
other
pitches
of
the
scales
other
other
keys.
If
you
go
the
same
with
tempo,
you
remember
the
absolute
tempo
and
absolute
and
you
can
infer
it
at
different
tempos
when.
D
No,
no,
no,
it
means
that
when
you've
learned
the
song
from
like
you've
heard
performer
singing
like
not
not
only
different
ones,
you've
heard,
like
don't
take
a
second
some
song.
Some
artist
comes
out
with,
and
you've
heard
it
now.
You've
heard
that
recording
it's
a
fact
you
you
actually,
although
you
can
play
back
at
any
key
and
you'll,
recognize
any
key.
You
actually
memorized
a
computer,
and
you
think
that.
D
D
Although
you
may
not
even
be
able
to
tell
me
what
it
is,
but
if
you
do
that,
if
accuse
then
he
gets
the
same
thing
was
tempo.
You,
you
always
played
back
right
tempo,
but
you
can
hear
it
the
tempos,
and
so
then
we
have
the
same
issues.
It's
something
like
a
physical
object
on
your
coffee
cup.
Alright
I
have
my
pneumatic
coffee
cup
and
I.
Remember
it
at
a
particular
size.
D
D
Think
all
the
ones
in
my
house,
tea
cups,
but
in
a
general
way.
So
but
the
point
is
we
can
infer
an
object,
I've,
never
seen
before
different
slides
that
hey
that's
similar
to
that's
the
same
basic
structure.
That's
what
you're
arguing
the
same
basic
structure,
vision,
relative
positions
and
sizes
of
things,
even
though
it's
different
so
to
me
it
seems
like
we've
got.
We've
got
these
two
representations
of
everything
it's
like
somehow
I
have,
and
this
only
bothered
economy
that
we
never
gonna
solve.
This
problem.
D
I
got
a
solution
for
this
problem,
the
tempo
problem,
and
then
we
are
also
dealing
with
a
problem
here.
We
never
this
criteria,
but
I
suggested
before
that
we
have.
We
should
just
accept
the
fact
that
we
can
do
these
things,
and
so
we
came
up
with
a
displacement
self
I
thought
that
was
that
many
times
I
think
the
spice
miss
I
was
like.
Well
that's
a
good
solution
here,
because
somehow
this
placement
cells
are
representing
the
they're
representing
the
you
could
say
it's
I
don't
want
to
face
the
relative
position.
D
It's
it's
kind
of
like
the
displacement,
is
kind
of
a
relative
position,
but
but
it's
something
I
can
scale
up
and
scale
down.
It's
independent
of
the
specific
instance
so
like
like
it
would
be
solve
the
problem
of
the
absolute
pitch
anyways.
The
question
is:
can
we
do
we
have
two
representations
right?
Do
we
have?
D
If
you
do,
if
you
have
an
object
and
then
you're
because
I'm
or
something
like
that,
then
I
think
the
idea
that
this
this
idea,
that
you
have
the
scale
factor
through
the
thalamus,
helps
that
greatly
so
I
can
take
an
absolute
object
and
recognize
it
at
different
scales
and
they're
sweet.
If
you
use
a
family
mechanism
for
that,
and
then
you
just
basically,
if
you
scale
down
your
eye
movements,
move
laughs
or
move
more,
it's
all
that
it's
basically
the
same
model.
It
says
you
know,
but
he
being
at
this
amount
to
do
something.
D
Man
look
at
that,
but
the
motor
commands
are
same
and
it's
just
gives.
The
thalamus
is
just
scaling
all
movements,
it's
getting
all
the
time.
So
so
you
can.
You
cannot
yeah
I
can
take
an
absolute
model
of
a
coffee
cup
and
manipulated
at
a
different
scale
and
recognize
it
and
different
scale.
Just
by
changing
this
whole
onyx
scale.
Back
averaging
the
brain
uses,
the
same
memories
use
the
absolute
memory
and
a
different
scale.
I,
don't
know
I
just
I'm
trying
to
I'm.
D
A
D
D
D
D
D
D
D
It
seems
like
be
careful
because
it's
a
seedling
one,
its
entire
I.
Am
this
entire
memory
of
how
the
manipulates
that
cup
that's
physical
and
am
I
kind
of
calculate
the
relative
movements
every
time
I
do
that
it
seems
like
it.
It's
sort
of
like
the
same
with
a
melody.
It's
like
you're
gonna.
Remember
you
can
you
can
do
that
like
it
would
be
a
default
scale.
They'd
be
the
first
scale
of
your
arm
to
that.
D
So
maybe
maybe
that's
what
you're
saying
let's
say:
I
learn
the
melody
and
only
in
displacement,
but
then
there
is
a
default
scale
and
people
scale
every
day.
It's
like
the
normal
rhythm
under
that
would
be
like
the
default.
Let's
say
the
sake
of
killing
cops
you'd
be
the
normal
default
frequency.
Not
now
I
would
agree
with
you.
If
you
were
stuck.
If
you'd
accept
that
idea,
then
I
would
tell
a
fine.
D
So
the
coffeehouse
model
is
a
bunch
of
displacement
variants,
but
integral
to
it
is
the
normals
scaling
of
a
balance,
and
so
then
it
would
be
like
you
said,
then
it
would
be
like
a
separate
thing,
but
it's
not
stored
separately.
It's
just
it's
that's
the
best
you
learned
it
imagine.
I,
have
a
clock.
It's
a
world
clock
and
I've
learned
everything
related
to
this
world
clock
right.
D
What
everything
relative
to
a
meter
stick
and
now
so
in
some
sense,
I
I've
stored
it
in
meter
genus,
but
now
I
can
just
change
the
size
of
a
meter
or
I
can
change
how
fast
the
clock
lines.
So
it's
kind
of
in
between
it's
not
like
you
have
it.
The
knowledge
is
completely
independent
of
the
size
of
time.
It's
based
on
sort
of
clocks,
a
certain
in
certain
sizes,
but
I
can
change
that.
If
you
want
to
call
it
meet
the
size
of
a
meter
as
the
definition
of
the
object,
I.
C
D
D
D
D
D
D
B
A
So
I
could
have
written
these
instructions
out
differently
or
we're
draw
the
same
thing,
except
rather
than
moving
I'm
gonna
like
I'm
gonna,
see
this
image
like
my
flashing
currents
meant
multiple
cortical
columns
are
gonna.
Are
gonna
see
this,
like
suppose,
my
visual
field
is
kind
of
divided
up
into
cortical
columns.
A
Anyway,
so
we're
talking
about
like
yeah
predicted
it
seeing
this
image
should
kind
of
okay
here,
I'll
say
something
that
I'll
try
to
defectives
seeing
this
pendant
causes
causes
a
prediction
over
here.
Somehow
see
seeing
this
image
in
this
part
of
this
cortical
column
ought
to
cause
a
prediction
over
here.
Like
you
know,
we
want.
We
want
to
predict
yeah
it's
hard
for
me
to
talk
about
scale
in.
D
D
We
haven't
done
with
that
problem
at
all.
We've
just
assumed
in
flash
inference
everybody's
learning
some
component,
they
all
vote
yeah,
that's
it
I
know
this
is,
but
we
haven't
been
able
to
show
how
how
you
can
scale
to
them,
how
you
could
say:
okay
and
a
bigger
scale,
the
prediction
this
guy,
you
can
predict
something
over
here.
D
D
B
A
D
D
You
know
pushing
the
coffee
cup
the
voting
works
really
well,
because,
because
it's
this
is
an
on
scale,
object
and
your
fingers
were
right
on
top
there.
So,
but
when
it
comes
to
this
this
the
flashing
from
vision,
we
never
really
got
everything
working
carpet
in
the
system
and
and
now
we're
saying
hey.
This
is
another
problem
of
division.
One.
D
A
D
A
A
A
D
D
All
right,
well,
I,
don't
think
we
to
solve
this
mob
today,
but
I
think
now
I'm
going
to
add
that
most
of
the
problems
they
have
to
solve.
But
if
you
got
again
here,
we
came
about
them,
touch
and
touch.
We
don't
have
these
issues,
we
feel
like.
We
don't
have
these
issues,
sensing
of
distance
and
that
our
recently
a
few
weeks
ago
or
a
month
ago,
I
read
that
actually
touch
it,
something
in
a
distance.
It's
just
an
extreme
version
of
it.
D
So
I
think
there's
probably
this.
This
is
the
generic
problem
for
flesh
and
prints
at
scale.
It's
a
problem
and
I'm,
not
sure
if
the
costume
idea
solves
it
feels
like
I
can
see
how
the
columns
could
be
voting
on
the
scale,
each
column,
saying
I'm,
recognizing
something
and
there's
a
hypothesis
of
what
I'm
recognizing
and
so
all
the
columns
could
be
making
hypothesis
of
scale
and
and
then
saying,
okay.
Well,
let's,
let's
try
this
scale.
You
know
this
is
the
winning
scale.
D
A
A
A
It's
a
or
its
equivalent
to
like
how
Jeff
Fenton
will
use
the
example
with
you
see
an
eye
and
that
votes
for
the
idea
that
there
is
a
face
right
here
and
you
see
this
other
than
other
parts
than
other
parts
of
your
visual
cortex
sees
this
eye
and
it
votes
for
the
idea
that
there's
a
face
right
here.
So
you're
kind
of
voting
on
the
parent
object
at
a
pose.
D
D
A
A
A
D
D
D
D
A
D
D
D
C
Hooli
hierarchy
is
a
critical
part
of
the
answer
to
this
problem.
It
seems
that
we
have
this
positive
displacement
cells
for
these,
and
you
know
how
much
discussing
there
between
these
two
objects,
maybe
within
a
single
column
in
particular
that
there's
compositionality
there
as
well
still
some
objects
within
it.
So
we
had
some
notion
of
hierarchy
below
that
single
object
that
that
hierarchy,
blue
can
also
produce
the
displacement,
but
is
able
understand
it
with
this.
This
subject,
this
this
sub
objects,
there's
some
displacement
that
is
implied
by
you
know.
C
The
base
is
a
little
bit
is
like
one
new
Great
Chicago
way
from
the
top,
and
that
displacement
vector
can
then
be
taken
up
to
a
higher
level
hierarchy,
and
that's
the
exact
scale
it
about
you
they're,
just
like
a
multiplier.
The
little
scale
of
the
composition
of
the
whole
can
they
be
used
to
make
predictions
about
it
than
that
project
as
well.
It's.
D
C
A
C
C
B
C
D
Final
thought,
but
my
way
of
thinking
about
this
would
not
be
to
be
lying
hierarchy.
It
would
be
as
far
I
think
about
single
common
single
level.
Ask
how
of
these
problems
and
then
expand
down
to
how
they
vote.
You
solve
the
problem,
then
you
explain
that
value
good
progress.
Okay,
don't
I,
don't
want.
D
D
D
D
D
A
Anything
that
we
talked
about
before
that's
kind
of
just
suffer
from
all
this.
That
I
think
is
a
useful
thing
to
bring
it
the
idea,
and
you
might
have
first
I
forget
it
was
you
were
talking
to
someone
that
might
have
been
mount
castle
and
might
have
been
very
sherman
or
something,
but
the
idea
that
okay
cortical
columns,
we
talk
about
them
like
they're,
these
discrete
things,
but
sometimes
they
might
be
useful
to
think
of
cortical
columns
are.
A
It
is
all
kind
of
this
continuous
field,
this
continuous
whatever,
and
the
image
that
comes
into
the
retina
sort
of
determines,
where
the
cortical
columns
are
and
the
boundaries
between
cortical
columns
are
not
these
strict
things.
If
this
cylinder
lands
right
here
versus
if
it
lands
like
the
millimeter
over
or
something
like
that,
functionally
is
almost
like
cortical
columns
up
here,
where
they
need
to
appear.
It's
almost
like
am
I
making
sense,
I.
D
B
D
D
He
ended
up
in
the
end,
I
believe
believing
that
they
were
important.
There
were
discrete
vary
in
size
and
one
question
about
that:
they
could
be
different,
Marvin,
smaller
large,
ready
to
go
off
the
hierarchy,
but
but
I
mean
we
don't
know
the
answer,
the
rest
of
it
when
I
started
off
my
name,
I
think.
D
A
C
C
C
D
D
C
D
C
C
C
D
C
A
C
If
you
want
to
figure
out
the
death
exactly,
you
have
to
know
the
size,
you
don't
know
this
size,
but
if
you
keep
the
size
fixed,
just
like
each
other
you're
also
true
you're
also
correct
that
for
a
given
image,
if
you
move
point
five
degrees,
everything
is
done
at
five
degrees.
That's
also
correct.
I.
D
C
D
C
D
C
D
A
A
mic
this
and
this
beam
is
going
to
land
somewhere
else
somewhere
now.
One
weird
thing
is
that
confuses
me:
a
little
bit
is
you're.
Not
you
turn
your
eye,
you're,
not
just
going
to
rotate
about
this
point,
you're
going
to
rotate
about
some
part
back
here,
so
so
fine
you're,
the
actual
pupil
is
moving.
So
is
that
where
the
subtle
difference.
D
C
D
D
A
C
A
D
A
D
D
D
Yeah
so
we
know
it,
but
if
you're
getting
a
single
column,
we
don't
reflection
to
design
network.
So
the
questions:
how
does
it
seem
column
solve
all
these?
These
posed
issues
which
we
talked
about
they're
tough,
but
we
didn't
go
to
soccer
vision
and
I'm,
suggesting
we.
Maybe
we
just
focus
on
that
single
column
again
and
to
think
about
it
that
way,
and
then,
when
you
can
ask
how
the
single
column
you
don't
solve
all
these
problems,
to
figure
out
all
these
different
variables,
then
you
can
ask
Howard
opposed
to
do
that.