►
From YouTube: Numenta Research Meeting, May 29, 2019 - Part 2
Description
My broadcasting software crashed in the middle of this stream, so I had to cut it up into pieces. Sorry about the previous botched video. This one has sound throughout.
-- Watch live at https://www.twitch.tv/rhyolight_
B
A
There
was
a
chair,
although
you
can
see
this
chair
and
there's
no
reference
for
anything
else.
Well,
I.
What
would
I
held
on
the
chair,
I
would
say.
Oh
there's,
a
chair,
I
would
say
it's
distance
from
me
and
I
would
say
it's
orientation
other
really
good
position
in
any
kind
of
angle.
The
reference
track.
It's
inching
as
I
turn,
my
body
I,
don't
perceive
that
the
chairs
position
is
moving
so
clearly,
it's
I
know
that
and
then
we
talked
about
having
two
objects
that
are
like
that
and
and
then
what
would
their?
A
How
would
we
do
this
and
what,
if
they,
what
if
they
were
on
a
rug?
So
now
I
were
wrong
and
I
have
two
chairs
on
the
right
now.
Somebody
didn't
feel
that
all
of
a
sudden
that
that
I'm
using
the
Rockets
of
reference
right
as
we
talked
about
that
that
would
be
also
in
reference
to
him
so
I'm
going
beyond
your
claim.
C
A
C
A
A
A
A
A
We
have
the
same
issue
in
displacements.
Right
like
an
individual
column,
is
looking
at
a
point,
object
and
a
moment
later
it's
a
different
out
there
and
you
can
say
what
is
the
displacement
back
column
but
the
product
of
particular.
You
know
things
like
just
there's
no
point,
you
can't
do
a
point-to-point
comparison
and
all
this
about
it
and
I.
A
Think,
though
we
had
a
kink
in
space,
what
you
can
do
is
say
is
a
compliment,
doing
a
particular
point
and
then
long
later,
to
know
that
this
thing
issue
existed
in
the
dispersed
themselves
and
now
the
existing
angular
displacement
as
well.
So
that's
the
best
I
can
say
that
we
haven't
walked
through
this
vector,
but
there
was
a
lot
of
parallels
between
I
think
between
what
we
call
the
displacement
of
vocation
which
have
the
same
dish
year
about
there's.
A
Normally,
there's
no
point,
there's
no
work
of
many
of
these
things
now,
in
the
same,
is
your
about
the
angular
position
of
these
two.
Is
that
there's
no
point?
You
can't
say
the
whole
object
position.
You
can't
say:
there's
a
displacement
of
angle
between
these
two.
If
you
find
displacement
is
saying
I
hope
I
said
them
is
company
a
same.
A
In
this
case,
it's
from
its
from
this
position
here
right,
yeah
and
I
can't
say,
be
treating
the
opposite
number
in
the
displacement
cell
for
location.
The
way
we,
the
latest
phrase
is
you
can't
ask
you
know
is
that
is
that
the
displacement
of
an
object
or
an
object
is
there's
like
on
one
point
of
objectivity
that
the
object
itself
doesn't
have
any
displacement
captures
the
relative
positions.
These
things,
nothing
specific.
A
A
Is
happening
here,
I
can't
say
just
like
I
can't
say
you
know:
two
objects
are
so
many
you
know
inches
apart
for
the
displacement.
I
can't
say
these
two
objects
are
so
many
degrees
apart.
This
right,
all
I
can
say.
Is
that
any
particular
column?
If
you
know
your,
if
you
know
where
it's
on
this
object,
they
give
you
a
displacement
on
though
I'll
know,
where
is
it
on
this
object?
I
try
to
overall
we're
going
to
try
to
make
the
parallel
between
an
orientation.
A
A
A
Enough
so
they're
still
they
just
basically
yeah
yeah.
The
displacement
can't
be
measured
in
angles.
Just
like
this,
the
displacement
you
can't
measure
this
displacement
degrees.
In
the
same
way,
you
can't
measure
the
displacement
to
to
existing,
especially
you
can't
measure
in
terms
of
you
can't
say
how
many
inches
apart.
Two
things
are.
It
just
came
to
that
there
is
no
origin
to
say
unless
they
of
us
the
displacement
from
here
here
in
our
current,
it's
all
displacement,
I
can't
say
how
its
three
inches.
A
F
A
There's
no,
you
can't
put
a
metric
on.
You
can't
say
it's
inches
in
just
the
same
way
here.
I
cannot
say
how
many
degrees
apart
these
are
you
haven't
careful
about
that?
You
there's?
No,
it
isn't.
It
does
not
measure
of
sometimes
like
the
Vista
this,
but
there's.
No,
you
can't
say
this
is
30
degrees.
Rotated
from
this.
That's
that's,
that's
not
possible.
Yeah.
E
A
This
point
you
don't
think
not
starting
with
it,
has
to
be
based
on
this
point
of
view
here,
right
I
have
to
calculate
the
displacement
eg.
These
two
all
I
know
is
I'm
measuring
from
here.
I
have
I,
have
information
I
get
from
here;
I've
information
I
get
them
here,
but
I
move
and
so
I
I
move
through
an
angular
distance.
My
step
service
I
knew
something
of
the
distance
and
but
I
need
to
figure
out
the
displacement.
A
The
displacement
these
two
first
in
an
angular
space
and
then
in
and
then
in
a
metric
space
and
you'll.
Not
do
that
so
I
don't
want
to.
This
is
a
topic
that
needs
to
be
delved
into.
What
I'm
excited
about
is
the
fact
that
I
think
this
is
a
tractable
tractable
problem
that
I
can
define
the
learning
of
an
object.
A
A
I've
drawn
these
two
hours
here
there
there
must
be
a
set
of
a
preferred
orientation
for
them
for
the
parent
object
and
I
have
to
share
that
I
have
that
that
must
be
shared
between
me.
I
have
to
know
my
my
angular
position
of
knowledge
up
to
this
room
in
the
same
way
that
these
things
have
to
learn
their
relative
orientation
villages
of
the
room.
So
there
was
a
common
grounding
here
in
orientation.
If
you
will,
between
these
two
thousand
other
inclusive
them.
A
I
am
not
proposing
the
method
for
how
to
do
this
yet
I,
just
I'm
gonna
chop
today,
I
just
I'm
just
letting
it
sink
in
a
bit
about
the
problem.
You
know
the
problem
we're
going
to
do
is
we're
going
to
come
up
with
a
a
an
inference
in
point
in
orientations.
Things
like
angular
movement
and
turn
that
into
and
so
and
then
we're
going
to
turn
that
into
a
displacement
in
certain
orientation
and
grid
cells
based
in
the
object
space.
A
So
it's
it's
a
it
just
defines
a
transform
that
has
to
occur
and
I
can
take
remove
the
entire
concept
of
moving
my
body
around.
That's
can
be
written
as
business
problem
where
that's
what
we
started,
that
that
was
the
main
movement
we
had
and
now
saying.
Oh
this,
actually
not
good
at
all.
I
can
just
not
move
our
body
knowledge
into
the
objects,
we're
gonna,
be
stationary
and
holding
a
new
thing
of
the
movements
and
we'll
be
able
to
build
this
entire
object
model.
That
mean
that's
just
a
statement
of
problem.
It's
not.
C
A
Objects
is
the
eye
and
say
that
that's
part
of
the
object
definition.
Okay,
but
that's
that's
part
of
the
problem
right.
The
problem
is
you're
gonna,
it's
not
part
of
the
coverage.
The
object
definition
has
no
sense
at
all
about
your
location
of
your
location.
I
started
up
front
right.
The
object
definition
is
no
concept
of
the
location
of
the
viewer,
so
we,
but
we
start
with
a
viewer
position
and
we
have
to
get.
We
have
to
take
this
viewer,
centric,
America
and
converted
into
an
object
center
position
and
that's
the
task.
A
But
to
me
this
is
even
not
allowed
to
do.
This
is
a
huge
advance,
because
I
now
have
a
different
way
of
thinking
about
the
problem.
I
eliminated
the
idea
that
I'm
we
were
moving
this
guy
and
I
just
need
to
figure
out
what
are
the
one
of
the
transformations
that
have
to
occur
to
get
from
this
position.
This
is
sort
of
view
into
an
optimal.
F
C
The
context
where
I
brought
up
determine
angular
displacement
was
like
if
the
orientation
of
these
objects
relative
to
this
parent
reference
right
now,
I
was
saying:
there's
an
angular
displacement
between
this
and
that's
your
displacement
like.
If
I
do
taxing
on
this
object
and
I'm
this
object,
then
I'm
saying
you
can
represent
the
orientation
of
this
relative
to
the
parent.
A
A
A
A
A
There
really
is
just
a
bunch
of
child
objects,
even
though
even
this
room
or
rug
is
just
another
one
of
them
and
as
you
go
between
these
you're
learning
the
displacements
between
these
guys
so
either
one
year
learning
the
displacement
earning
the
child
of
the
parent
or
you're
learning
to
displacement
team,
the
individual
children
and
Mark
and
I
discusses
a
bet.
There's
there's
advantages
both
of
these
and
it
seemed
like
we
trying
to
do
both.
A
So
if
I
just
come
into
the
room,
you
know
I
see
an
arrangement
of
plates
on
the
table.
I
might
just
learn
that
arrangement
might
run
like
this
and
I.
Don't
wear
any
developed
into
some
parent
object.
I,
don't
make
it
part
of
the
table.
It's
just
a
temporary
arrangement
of
things.
It's
like
I've
walked
into
that
white
room
and
I
see
three
chairs
range
relative,
each
other
I'm.
Just
remember
the
arrangement,
but
obviously,
if
I,
if
I,
see
those
three
chairs
on
top
of
a
world
like
this
or
in
a
room.
F
A
Know
I
might
be
tables
in
a
room.
I
might
learn
them
as
a
part
of
this
room
like
there
is
a
parent
and
there's
children.
This
is
far
more
efficient
in
terms
of
storage
and
it's
also
really
what
you
need
for
compositional
structure.
So
you
can
be
certain
that
this
is
occurring
and
I.
Think
I
think
what
can
happen.
A
I
haven't
worked
this
out
yet,
but
I
think
what
the
basic
idea
is
that
when
you
attend
an
object,
object
object,
you're,
going
to
be
figuring
out
the
displacement
between
them
in
all
cases,
and-
and
there
can
be
a
time
where
the
where
the
that
one
of
these
is
preferential,
is
like.
If
I
was
in
a
room
or
on
a
rug,
it
would
be
sort
of
a
preferential
one,
so
as
I
attend,
I
would
also
be
I
only
be
calculating
back
to
this.
A
One
I
haven't
worked
out
for
yet,
but
this
is
another
issue
we
definitely
dealt
with
and
I
have
a
lot
of
thought.
Experiments
I've
been
doing
trying
to
tease
apart
exactly
what's
going
on
there,
but
my
hope
is
that
that
we
could
show
that
the
the
same
single
cellular
mechanism
can
do
both
of
these,
and,
if
there's
something
that
suggests
one
of
these
objects,
it
should
be
a
parent
object.
Then
it
assumes
that
role.
A
But
if
there's
no
suggestion
that
one,
we
should
be
a
parent
object
and
then
the
system
defaults
to
just
guess
think:
there's
a
single
mechanism
that
does
both
of
these
the
physical
attributes
of
the
the
actual
objects
define
where
there's
one's
the
parent.
Just
to
give
me
an
example
of
that
meant
to
here's
like
a
wrong
and
three
chairs
on
it:
okay,
and
if
I
see
that
I
would
think
that
I
would
the
rug
is
going
to
stop
me
from
this
parent
role.
A
I,
don't
see
everything
relative
to
that
rock
if
I
had
a
small
rug
and
I
stuck
it
over
in
the
corner.
Here,
like
this
night,
three
chairs
in
a
bar
over
here,
we
don't
longer
feel
that
way,
even
though
there's
still
a
rug
and
there's
still
a
space
around
the
rug,
and
you
can
feel
I
could
help.
This
is
the
parent
objects,
and
these
things
were
just
sort
of
nearby
in
the
space
of
the
rug.
But
it's
something
physical
about
the
rug
itself,
which
suggests
that
it's.
A
A
How
witness
we
decide
that
so
there
is
these
two
different
methods:
Marcus,
if
you
throw
this
one
bit,
we
could
do
this
with
the
Union.
This
is
what
we
do
pooling.
We
can
find
an
object,
that's
displacements,
and
this
one
because
you
might
define
an
object,
there's
just
a
million
of
displacements,
but
I
had
to
think
the
same
mechanism
working
both.
So
that's
what
I'm
gonna
work
on
I
had
a
couple
of
seems.
E
Like
there's
another
thing,
when
we
talk
about
this,
it's
not
maybe
not
quite
captured
by
what's
on
there,
you're
kind
of
defining
the
rug
as
being
the
parent
object
now
individual
piece.
But
if
you
think
about
a
coffee
cup,
there's
there's
you
have
the
cylinder
and
you
have
the
handle,
there's
nothing:
physical,
that's
the
coffee
cup,
it's
an
abstract
notion
of
a
coffee
cup
and
we're
saying
the
handle
whenever
you
have
a
handle.
Next,
do
you
know.
F
A
So
it's
it's!
You
know
it's
just
like
some
random
thing
that
we
associate
almost
explore
that,
but
first
of
all
we
can
agree
that
the
momentum
logo
is
smaller
than
the
coffee.
There's
a
parent-child
relationship
right
there,
and
if
there
was
a
coffee
cup
in
the
logo,
that
would
be
smaller
than
the
logos
of
the
parent-child
relationship
and
now
the
question
bringing
up
is
the
handle
versus
the
cylinder
yeah.
A
That
case
yeah,
so
I,
don't
know
it's
a
good
question.
Here's
the
use
some
thoughts
about
that.
First
of
all,
we
do
see
a
lot
of
cylinders
in
the
world
without
handles
that,
as
we
see
cups
without
handles
all
the
time,
we
rarely
see
a
handle
without
a
cup
so
I
just
there
isn't.
There
is
a
discrepancy
or
dichotomy
between
those
they're
not
equal
to
somehow.
So,
let's,
even
though
this
one's
not
contained
in
this,
there
is
clearly
a
difference
between
those
two
Oh
Bob.
This
is
bigger
than
this
one.
They
are
attached.
A
E
A
B
A
That's
how
we've
even
phrased
it
there
is
a
you
know:
it's
a
cup
close
to
handle
it's
not
a
handle,
plus
a
cup
so
mentally.
In
our
line.
There
is
a
difference.
This
is
smaller,
I.
Don't
having
answered
your
questions,
I'm,
just
exploring
the
idea
a
little
bit
now
you
can
talk
about
a
mixing
idea
like
this.
Here's
other
imagine
we
had
four
chairs
and
I
tried.
Imagine
this
is
just
this
is
a
continuing
different
expense.
Relation
of
the
same
problem,
I
had
four
chairs
I.
A
Imagine
I
just
had
those
four
chairs
randomly
just
dropped
in
the
room.
Random,
be
very
quite
you
remember
that
configuration
like.
What's
the
configuration
of
those
chairs
I
think
that's
hard
to
do.
It's
like
it's
just
there's,
no
there's
no
obvious
reference
frame
and
which
to
say
that's
the
reference
frame
that
place.
Those
chairs
it'd
be
very
difficult
to
remember
that
configuration
all
right,
I,
put
the
chairs
down
and
such
that
there
were.
A
Three
of
them
were
in
a
clearly
sit
erect
your
position
like
this
and
the
fourth
one
could
be
anywhere
at
random
for
Bella.
To
do
that
I'd
be
very
easy
to
remember
that
that
arrangement
I
would
see
these
three
chairs,
it's
defining
some
of
the
parents
in
some
sense,
it's
it's
like
they're
cannons,
a
triangle
or
some
like
that.
You.
A
A
rectangle
here
isn't
one
because
some
of
the
objects
so
beg
to
define
one,
and
so
it's
like
it's
interesting
just
to
notice
that
okay,
four
chairs
and
one
arrangement
fortune
of
different
news,
four
chairs,
let
you
easily
remember
them
and
I-
think
I
can
easily
remember,
because
I
would
just
imply
that
this
is
so.
The
parent
object
is
these
three
together
seem
to
be
together.
They
don't
seem
to
be
random
in
the
range
and
therefore,
these
guys
have
preference
over
this
guy.
I
don't
see
this
guy.
A
Is
the
parent
I
see
these
three
defining
a
parent
region,
so
I
mean
there's
a
little
thought
experiment
you
can
do
like
that
which
so
the
tease
apart
like
well.
There
is
no
larger
object
here,
but
clearly
I
see,
there's
an
implied,
1
and,
and
so
I'm
now
I
would
be
able
to
blur
this
and
I
didn't
have
an
implied
one
I.
Just
can't
learn
it's
really
really
hard
for
randomly
placed
chairs
among
what
their
configurations
so
I
think
there's
just
clues
in
this
I.
Don't.
A
And
that's
all
I
can
say
about
that,
is
it
there
seems
to
be
able
to,
and
so
this
reminds
a
little
bit
of
the
coffee
cup
in
the
sense
that
well
iPod,
look
at
it's
just
two
objects:
a
lot
of
them
clearly
seems
to
be
more
dominant
than
the
other
and
it
says,
suggests.
Kidney
telomere
and
I
have
everything
and
I
existed
the
world
independent
this
little
guy.
So
pick
me
up
on
the
reference,
even
though
it's
not
contained
in
the
other
one,
so
it'd
be
a
little
bit
like
saying:
I
had.
A
And
then
I
had
four
chairs.
You
know
positions
around
the
outside
of
preachers
like
that.
How
about
I
remember
the
Ducks
position,
those
things
well,
I
would
probably
see
it
as
these
three
chair
fell
through
the
rug.
That's
something
this
is
the
parent
I'm
sad,
this
room.
They
ran
to
the
chairs
on
the
outside
of
the
right,
but
but
that
mean
it
again.
Why
is
that
different
than
this
one
down
here?
A
This
one
I
have
three
chairs
on
a
rug,
but
in
this
case
the
rug
is
small
on
this
case
like
a
baby
and
there's
something
about
that.
I
think
that
sort
of
closed
you
into
which
one
of
these
is
more
dominant.
In
some
sense,
it
gives
your
suggestion
which
one's
warmer
Clinton,
which
one
gives
you
the
rain,
maybe
just
tries
to
figure
out
what
should
be
the
parent
reference
frame
and.
A
And
goes
with
that,
so
these
are
think
these
are
all
problems,
I
guess,
obviously,
now
the
problem
that
I'm
going
to
work
on
is
two
here:
one
is
what
is
just
a
problem
representation.
How
do
I
represent
an
object
where
I
have
a
set
of
displacements
relative
to
some
power
of
the
reference
frame
and
those
displacements
have
to
include
location
orientation
and
scale.
C
F
C
Another
is
being
and
the
reference
frame
of
the
of
the
we'll
call
this
a
child
object,
or
the
third
option
is
that
it's
that
this
is
this
wasn't
immediately
obvious
that,
like
a
vector,
can
also
be
in
a
location
relative
to
an
object
can
be
in
a
totally.
It
can
be.
An
apparent
reference
frame,
and
that's
kind
of
what
we're
getting
at
here
is
that
this
part
of
what
you're
getting
at
you
is
that
these
green
arrows
aren't
in
the
reference
frame
of
the
sensor
or.
A
C
C
Partly
why
it
took
so
long
to
find
it
said
it
would
go
to
a
different
animal
I've
drawn
something
like
this
on
the
part
here.
I'll
use,
I'll
use
shape
it's
not
as
symmetric
or
I'll
do
something
that
can
be
rotated,
so
object,
vector
cells
can
be
imagined
really
each
one
of
them
represents
a
possible
green
arrow,
so
I'll
draw
them
in
green
for
that
easy.
C
C
C
And
I
guess:
I'll,
just
the
environments,
a
big
square
around
it.
Now
these
cells
is
interesting,
just
it.
This
is
gonna
all
come
back
to
things.
I
said
here:
if
you
rotate
the
object,
the
cells,
the
fire
killed,
stay
fix
the
firing
fields,
don't
rotate
with
the
object,
if
you
rotate
the
environment,
then
or
if
you
clip
the
book.
If
you
put
this
in
a
different
environment
where
the
animal's
head
Direction
cells
are
rotated
compared
to
this
one,
then.
C
F
F
E
E
A
C
A
Sort
of,
like
my
experience
when
I
was
doing
this
no
thought
experiment
here
and
I
was
imagine
looking
at
this
chair.
Even
when
I
turn
my
body.
My
perception
of
the
of
the
the
chairs
position
isn't
changing
its
changing
relative
in
my
face,
but
my
sense
of
where
it
is
I,
don't
feel
it
when
I
do
this
I
don't
feel
that
chair
is
moving
right.
I
know
that
chair
is
not
moving,
even
though
it's
sneaking
around
like
this.
So
it's
I
think
that's
that's
captured
analysis.
Yes,.
A
A
C
A
C
So
what
I
want
to
immediately
do
when
I
see
this
as
I'm
like
okay,
so
I
can
I
have
a
cell
that
responds
when
I
met
this
this
distance
in
direction
from
this
object
and
when
I
attend
to
this
other
one
I
get
a
different
one
of
these
arrows.
Well,
naturally,
what
I
want
to
do
is
take
the
displacement
between
those
arrows
to
try
to
get
the
to
represent
the
relative
location
between
these
two.
These
two
objects
in
a
way
that's
invariant
to
my
location,
that's
right!
That
is
the
key!
So.
A
C
E
B
C
So
so
what
I
want
when
I'm
glad
they
did
in
the
study
or
they
they
did
to
some
extent,
but
they
didn't
it
didn't
really
push
on
it
very
hard.
So
I
still
don't
know.
The
big
important
question
is:
are
these
arrows
to
some
reference
point
on
the
object?
This
is
some
common
point
on
the
object
or
are
they
to
the
perimeter
of
the
object?
Are
they
do
some.
C
C
F
A
C
C
C
E
D
A
A
A
C
C
And
so
I
drew
this
conclusion
that
that
it
must
be
relative
to
the
perimeter,
because,
basically,
in
the
small
print
of
their
paper,
they
said
that
that
the
distance
of
the
fields
relative
to
the
perimeter
of
the
object
was
was
highly
correlated.
It
didn't
make
any
any
actual
statement
that,
like
it
wasn't
when
we
tried
with
the
center.
F
C
D
C
A
A
E
Typically,
you
know
when
you
use
origins
to
compute
the
relative
displacement
between
two
reference
frames,
but
we're
saying
there's
no
origin,
but
yet
you
can't
have
relative
displacements
between
these
location
spaces
right
yeah.
Do
we
need
say
the
same
analogy
could
called
for
orientations.
Do
we
we?
Yes,
we
need
a
reference
orientation.
We
may
not
need
a
reference
orientation
at
all.
It
may
just
be
that
their
orientation
spaces
and
we
can
talk
about
the
relative
orientation
of
these
orientation
spaces
without
any
single
reference
or.
E
A
C
F
A
E
F
A
Remember,
I've
always
made
the
comment
that
orientation
cells
in
the
cortex
seem
to
go
for
all
the
columns.
Its
remind
me
that
it's
like
saying
that
the
orientation
cells
and
I'm
going
to
be
used
for
myself
and
for
these
objects
info
rooms
and
how
are
all
anchored
together
or
something
like
that.
It's
I'm
not
saying
that
exactly,
but
they
there's
no
preference,
but
they
get
anchored
together,
yeah.
A
So,
whenever
I'm
in
this
room,
even
though
I'm
moving
around
Welch's
room,
I'm
gonna
have
the
same
cells
facing
this
way
and
the
objects
in
the
room,
I'll
even
say
God.
Now
all
these
things
that
somehow
the
skin
reaction
to
room
there
is
a
preference
to
render
into
the
central
environment
of
the
water
object
the
room
and
that's
going
to
share
these
things.
A
There
is
no
preferred
you're
right,
there's
no
there's
no
really
way
of
saying
this
is
the
right
direction.
It
could
be
that
could
have
drawn
that
green
I,
like
that
yeah
I,
can
find
the
green.
A
lot
of
this
is
just
that,
whatever
I'm
in
that
room,
the
cell
that
represent
this,
it's
all
relative
to
the
same
yeah.
B
A
B
We
make
that
one
experiment
in
the
right
or
any
place.
You
know
the
reference
frame
needs
to
be
like
that.
Existing
physical
object,
okay,
feel
like
a
point
that
imagine,
like
I,
see
a
closer
sure
yeah,
but
I
just
meant
there's
like
a
closer.
There
is
a
center
of
it.
Closer
and
I
can
see,
there's
a
there's,
a
placement
of
location
and
orientation
regarding
to
that
of
the
question.
But
this
is
something
I'm.
Imagining.
D
A
A
A
A
In
our
language,
you
might
see,
there's
a
bunch
of
columns
that
are
seeing
these
angular
and
developing
and
they're
voting,
even
though
there
are
columns
that
are
missing
and
I'm
throwing
this
enough.
Columns
are
voting
same
triangle
that
we
see
trying
to
yeah
that'd,
be
doubly
neuroscience
explanation
for
that
writing
card
voting
mechanism,
so
yeah,
so
it
didn't
have
to
really
just
it.
Just
have
to
be
applied
enough
implied
that
there's
enough
voting
going
on.
That's
enough
of
these
cells.
B
G
A
A
A
So
I
think
what
basically
goes
on
as
we
as
we
attend
to
different
objects.
We
we
calculate
the
displacement
between
them
constantly
and
that's
the
displacement
gene
them
relative
to
a
parent
object.
If
there
is
a
parent
object,
there
is
no
parent
object,
somehow
it's
just
between
the
individual
pieces,
but
it's
the
act
of
attending
the
different
objects,
but
like
the
different
chairs
in
the
room.
That
literally
just
builds
up
a
set
of
displacement
voltage
of
the
Paramount
day,
and
it
doesn't
matter
where
you
do
that
problem.
C
A
location
and
orientation
in
such
a
way
where,
like
right
now,
you're
looking
at
this
marker
and
I,
mean
the
specifics
and
RedNet
now
right
now,
it's
it's
location,
staying
constant,
but
it's
orientation
is
changing
yeah
and
does
that
happen
in
the
brain
is?
Is
there
is
that
possible
for
an
orientation
of
something
to
change
without
without
its
location,
changing.
A
F
A
A
C
A
A
A
So
now
we
ask
the
question:
if
I
have
a
thousand
visual
prompts
observing
this
object
as
it
rotates,
individual
columns
will
see
things
that's
further
or
closer
to
the
right
and
and
will
be
seen,
different
parts
of
object
and
so
I
think
if
I,
if
I,
if
I
do
a
temporal
pulley
go
through
all
these
changes
or
something
that's
stable,
submitter
does
not
imply
that
there's
a
point
it
just.
It
just
implies
that
there's
a
stable
representation
for
that
thing.
E
A
A
D
C
A
C
G
C
C
F
D
E
Think
it's
it's
from
what
I
don't
know.
The
literature
is
one
of
isolation,
but
I
think
it's
very
similar
to
that.
But
it's
it's
not
specific
to
the
object
identity,
but
it
is
specific
to
the
object,
kind
of
shape
and
things
like
that.
So
if
you
have
two
different
completely
different
rectangles,
it
would
still
be
the
same,
but
something
that
it's
curvy
might
be
represented
differently,
because.
A
There's
another
thing
about:
that's:
not
what
I
want
to
reach
for
an
optic
right,
you
might
say
well,
I,
need
to
know
what
it
is
and
I
want
to
say.
No,
you
have
to
know
where
the
optic
is
relative.
Each
part
of
your
skin
that's
going
to
touch
each
party
out
there.
Each
one
seems
to
know
where
the
part
of
the
object
is
going
to
touch
you
don't.
This
is
a
centralized
distance
of
location.
A
It's
more
like
every
part
of
your
hand,
has
a
position
relative
to
an
individual
part
of
the
top,
and
each
one
has
to
figure
out
how
to
disconnect
grab
its
particular
part
of
the
cup.
There's
no
central
location
of
where
that
is,
and
that's
enough
to
trip
the
motor
hatred.
The
trick
come
over
here
is
that
these
fingers
moving
independently
and
it's
not
like.
Oh
my
handles
at
XYZ
and
cops
at
XYZ.
It's
known
as
my
fingertip
is
someplace
and
the
fingers
don't
touch
to
some
place
and
there
is
no
vocal
there's,
no
centralized
position.
A
A
A
A
A
E
A
A
A
A
A
D
A
E
A
A
And
I
have
adult
the
distance
so
I'm,
giving
you
a
magic.
It
is
I'm
going
to
give
you
I'm
going
to
give
you
your
XYZ
XYZ
position
in
the
room
I'm
going
to
give
you
a
distance
Y
on
the
distance
to
and
Delta
change
of
angle,
those
those
are
the
things
I'm
giving
you
that's
that's
what
I
can
directly
measure
and.
A
A
A
A
A
B
E
C
F
F
A
Then,
when
I'm
here
does
another
cell,
it
doesn't
there's
another
cell
that
that
it's
not
this
one.
This
is
not
there's
a
cell,
be
that
also
responds
to
that.
Yes
again,
and
so
here
I
have
this
couple.
So
here
I
have
two
cells
that
we
come
out
now.
You
can
think
there's
really
two
populations
become
active
population.
A
population
B
and
I
mean.
A
A
Because
this
cell
can
there's
a
cell
that
responds
to
this
and
a
different
cells,
but
I'm
here,
there's
a
cell
respond
to
this
and
there's
another
cell
responsiveness
and
so
really
the
way
I
think
about
it.
There's
a
spark
population.
The
response,
when
I'm
here
attending
to
this
and
as
a
sparse
population,
responds
with
here
10.
A
A
F
A
And,
and
so
what
we're
saying
here
is
what
do
our
old
displacement
cells?
Two
are
all
displacement
cells
go
between?
You
have
either
you
know
space
space
number
one
space
number
two
and
the
movement,
and
you
can
go
either
way
and
say,
give
me
space
my
locations
based
on
one
communication,
space
number,
two.
Okay,
so
that's
that's
the
same
thing
here:
I
have
a
population
of
cells,
I
have
a
movement
which
is
a
rotation
or
you
know,
and
and
now
populations.
A
And
now
there's
a
difference,
so
basically
I
will
say,
given
this
poster
moment.
I
can
pick
this,
and
given
this
plus
a
movement
I
can
predict
this.
That's
what
we
need
in
some
sense
and
but
much
theory
that
here's
what
we
want
to
have
happen,
that
if
I'm
over
here
and
I'm
doing
this,
that
if
this
displacement
one
works
here,
displacement
one
would
work
here.
That
would
be
the
goal
that
would
be
saying.
I'm
now
calculate
the
displacement
you
do
to
these
two
objects.
A
That's
independent
of
my
position
that
this
displacement
vector
will
represent
the
relationship
and
by
that
I
mean
I.
Can
at
one
location
move
under
my
look.
I
know
where
I'm
gonna
be
another
one.
This
will
all
just
paralleling
the
existing
displacement
models
that
would
be
I,
think
might
be
the
government,
but
that
to
me
that
should
be
the
goal.
Whatever
argue
to
calculate
displacement
between
this,
this
population
and
this
population,
I
should
get
the
same
displacement
when
I
go
from
this
population.
A
A
This
object
that
result
is
very
helpful
because
now
I
can
say:
oh
this
populations
element
of
population
of
cells,
and
you
said-
and
you
said
it's
just
like:
oh
yeah,
it's
displacement
between
those
two
or
you're
saying
they
still
hold
on
saying-
is
population
the
population
and
now
another
thing
the
same
thing
that
I'm
just
saying
if
I
calculate
this
is
a
different
set
of
cells,
and
this
is
the
difference
in
itself.
But
the
displacement
between
yourself
in
this
stuff
should
be
the
same
as
the
displacement
in
their
cells
and
is
no
that's.
F
F
A
Where
you
were
the
end
up
with
the
same
display
themselves
after
a
particular
movement
over
the
same
displacement
cells
at
the
same
decision,
that
is
a
very
nice
problem.
Definition
there
are
existing
mechanism,
might
work,
think
about
it,
but
instead
of
take
the
existing
mechanism,
we
assume
there's
a
bunch
of
grid
cell
modules.
A
A
Okay
but
I'm
afraid
the
problem
terms
of
object
represents
which
we
don't
exist
right
and
object
of
yourself.
If,
as
you
said,
it's
the
Truman
Robin,
we
don't
have
to
ask
how
that
occurred.
Necessarily
right,
we
don't
object
ourselves
come
about
necessarily,
but
if
I
postulate
them
our
object,
vector
cells
that
behave
like
I
said
then
forget
about
the
perimeter.
Stop.
A
C
F
A
C
And
that's
that's
just
our
interpretation
or
putting
over
it.
That
might
be
correct.
They
did
experiment
with
multiple
objects
and
they
did
find
that
that
when
there
are
multiple
objects,
the
firing
map
of
the
cells
behaves
in
the
expected
way.
Where
but
we're
like
this
cell
fires
here
and
here's
the
cell
fires
here
and
here
they
didn't
do
any
tribe.
They
didn't
try
to.
They
didn't
try
to
do
any
temporal
untangling.
They
didn't
try
to
make
sense
of.
A
F
A
A
I
knew
it
a
chair.
I
can
now
attend
to
the
simple
subset
of
the
chairs.
If
I
haven't
learned
the
chair,
yeah
and
so
I
will
doing
that,
all
these
things
will
be
changing.
So
if
there's
a
lot
of
internal
mental
stuff
going
on
while
representing
what's
happening
here,
you
know
so
you
know
even
in
this
box,
like
you
just
say,
oh
this
list
attend
to
this
this
little
symbol
over
here
in
the
corner
and
then
I
would
get
a
different
set
of
object
response.
C
A
A
F
A
You
know
now
that
now
this
one's
bigger
than
some
smaller
I
think
that's.
That's
all
part
of
the
the
I
think
this
was
growing
out
of
the
thalamus.
You
know
that
whole
scaling
I
can
visit
the
comments
that,
as
you
do,
these
attention
things
Thomas
is
calculating
the
distance
or
the
six
eight
or
Donald's
pocket
in
the
distance.
It's
complicated,
but
beautiful.
A
Yeah
love
it
if
we
phrase
the
problem
in
terms
of
these
displacements
and
start
there
that
these
have
to
be
in
Brooklyn,
then
then
these
complications
might
resolve
themselves
as
we
go
through
the
process,
because
here's
a
fact
I'm
saying
this
is
a
bet.
This
is
going
to
be
able
to
that.
That's
the
definition
of
what
we
need
and-
and
there
are
all
these
other
complications
here-
are
going
to
happen.
A
We
have
to
start
from
that
fact
and
that
will
help
us
guide
us
to
the
right
solution
and,
of
course,
at
any
point
in
time.
There's
multiple
columns
recognizing
this
and
there's
multiple
columns
recognizing
this.
So
there's
this
population
going
on,
but
he's
commented
on
said
about
how
to
correct
ourselves.
D
A
Might
have
a
single
column
that
looks
up
here
and
it
has
to
move
a
little
bit
to
infer
this
object
and
so
now
has
a
stable
representation
of
the
object.
Who
knows
what?
Because,
even
though
it's
angle
here
is
changing
and
what
point
of
the
object
is
looking
at
is
changing
and
then
it
swings
over
here
and
now
it's
going
to
recognize
this
object,
even
though
it's
positioning,
it
might
be
changing.
So
it's
like
there's
a
if
I
can
have
a
stable
repetition
of
this
object.
A
A
G
A
Don't
we
need
that?
We
need
this,
but
the
point
is
we
need
to
have
not
focusing
on
the
exact
spots,
we're
focusing
on
the
fact
that
you
need
to
have
the
same
displacement
representation,
regardless
of
where
your
activated
that's
part
of
the
solution,
not
not
part
of
the
problem
statement,
the
problem
statement.
We
have
to
have
the
same
displacement
and
whether
we
don't
know
we
have
exactly
what
it
means
on
the
same
spot.