►
From YouTube: Exact Timing and Oscillatory Dynamics (Part 5)
Description
Jeff answers common questions about exact timing and oscillatory dynamics in the brain.
You should have some background in HTM before understanding these videos. See http://numenta.org/htm-school/ for more videos explaining basic HTM theory.
Music: "Holy Roller" by YACHT
(used with permission from Free Music Archive)
A
Let's
talk
a
little
about
exact
timing,
I
know
what
you're
gonna
say
about
this,
but
I
just
want
you
to
say
it
again,
because
we
get
questions
about.
Why
aren't
we
modeling
exact
timing
and
what
about
facility
or
if
it's
a
little
Tori
dynamics
and
the
cell?
Why
don't
we
care
about
that?
Maybe
you
didn't
those.
B
Are
two
separate
questions?
Yes,
the
timing
is
not
the
same.
It
is
the
scylla,
Tori
dynamic
sure
you
know
theta
waves.
This
is
talking
about
time.
Okay,
all
right,
it's
clear
that
you
have
to
have
time
right.
So
if
I'm
grading
at
my
speech
right
now
requires
that
I
produce
a
set
of
motion,
motions
or
movements
of
musculature
contractions,
very
specific.
B
B
I
am
long
have
a
theory
about
where
the
timing
is
occurring
and
how
it's
occurring
that
and
I
even
wrote
about
there's
own
intelligence.
We've
talked
about
other
times,
I'm,
just
gonna
briefly
mention
it.
I
believe,
and
this
is
a
hypothesis.
It's
not
proven
that
that
is
stem
is
originating
timing.
Signals
are
originating
in
part
of
the
thalamus
called
the
matrix
cells,
which
is
not
the
same
cells.
B
We
were
talking
about
earlier
that
they
broadcast
over
a
broad
region
of
cortex
specific
to
particular
men
down,
so
auditory
won't
get
a
set
of
cells
in
the
matrix
element,
a
broad
kiss
up
with
the
auditory
region.
Another
one
might
give
some
of
the
visual
region,
and
so
on
and
I
believe
and
there's
no
evidence.
B
My
speculated
this
many
years
ago,
there's
no
evidence
that
those
matrix
cells
actually
produce
a
timing,
light
signal,
a
cascade
of
activations,
which
would
allow
those
regions
to
get
a
timing
signal
and
that
would
occur
on
the
apical
dendrites
of
those
cells.
So
this
timing
signal
appears
in
the
layer
one
of
these
cells
in
layer,
three
and
layer.
Five
apical
jump
right
to
go
up
there,
yeah
and
so
the
essentially
the
the
cells,
the
lantree
cells
in
layer.
B
Five
cells
would
have
just
this
turning
signal
and
in
the
case
of
Layton's
temporal
memory,
it
would
be
the
layer
of
three
cells
in
layer
three
beat
and
that
we
could
walk
through.
Indeed,
somehow
this
time
I
say
won't
work
where
the
sequence
memory
would
now
not
just
go
to.
The
next
element
would
go
to
the
next
element
at
the
correct
interval
of
time.
From
the
previous
element,
I
see.
B
B
A
A
B
Well,
terrorizing
about
we
haven't
implemented
in
HDM
yeah
I.
Keep
track
of
this
I
was
just
at
cosine.
A
couple
weeks
ago,
I
was
talking
to
a
woman
who's,
an
expert
on
these
matrix
cells
in
the
thalamus
I'm
drilling
her
about
this
and
I
ran
by
her
this
hypothesis.
She
thinks
it's
probably
right.
There's
all
this
evidence
suggest
insist
on
what's
going
on.
We
don't
model
that
today,
because
we
don't
it
seems
easy
to
do
and
we
don't
have
a
need
for
it.
B
If
I
were
working
on
a
machine
learning
problem
that
required
specific
timing,
I
have
to
implement
something
like
this.
At
the
moment
we
don't
have
to.
We
don't
have
a
problem
that
requires
it.
It
doesn't
seem
to
be
the
challenge.
It's
not
a
big
challenge.
For
me,
it
almost
seems
simple
at
this
point,
so
I'm
not
going
to
spend
time
on
and
implement
it
I'm
other
big
fish
to
fry
right
or
catch
or
whatever
each
other.
So.
B
B
B
A
B
Yeah
they're
sort
of
groups
of
cells
that
are
sort
of
synchronized
and
they're,
roughly
synchronizing
their
firing
pounds
roughly
okay.
So
if
I
look
at
all
the
spikes
are
to
be
generated
by
some
set
of
cells
between
two
regions
or
might
see
them
better,
they're
more
aligned
on
certain
weight
frequencies,
they
tend
be
spiking
on
certain
peaks.
Oh
I
see
right.
So
it's
like
that's
what
they
mean
by
individual
cells
with
these
spikes
coming
out,
and
sometimes
it
looks
like
they're,
somewhat
coordinated
all
right.
So
it's
not
a
random
liquor.
B
They're
conquering
on
waves
of
Peaks
of
things
like
that.
That
make
sense!
That's
what
the
oscillations
me?
Okay,
our
basic
belief
at
the
moment-
and
it's
just
my
hypothesis-
is
that
these
oscillations
are
essential
for
neural
tissue
to
operate,
but
they're
not
essential
from
an
information
processing.
Point
of
difference.
B
Yes,
one
of
the
the
basic
cores
of
hgm
theory
is
that
the
the
cells,
the
neurons,
create
these.
What
is
called
NMDA
spikes
in
the
dendrites
yeah
and
by
there
was
a
really
nice
article
that
came
out
recently
talking
about
how
these
surprisingly
of
these
enemy
of
spikes
were
occurring
everywhere.
All
the
time
which
we
predicted
push,
and
but
anyway,
one
of
the
hypothesis
that
these
NMDA
spikes
are
are
a
key
critical
processing
element
for
how
the
brain
works
and
an
enemy.
B
When
a
the
synapse
is
on
a
particular
dendrite,
a
small
section
of
dendrite
become
get
input
at
the
same
time
near
each
other
in
space
and
near
each
other
in
turn.
If
you
spread
them
out
too
far
apart,
they
say
this
one
comes
active
and
then
Owen
ten
milliseconds
later
not
ten
milliseconds,
but
another
10
mil
sit
doesn't
work
right.
They
gotta
also.
B
They
all
have
to
arrive
sort
of
simultaneous
within
a
few
milliseconds
of
each
other
right,
so
let's
say
that's
all
nature
could
come
up
with
okay.
So
how
is
it
going
to
make
sure
that
these
spikes
tend
to
arrive
at
the
same
time?
One
way
to
do
that
is
just
put
an
oscillatory
background
on
the
whole
thing
to
say:
I
say
well.
This
makes
everybody
certifier
news.
So
what
I
want
to
detect
a
set
of
cells?
Firing
I
need
to
arrive
at
the
same
time,
I'm
really
just
detecting
a
set.
A
B
There's
a
plumbing
problem,
it's
like
you,
know,
I,
think
the
hardware
I
might
add
a
capacitor
to
solve
some
problem.
You
know
yeah,
it
doesn't
really
matter.
It
doesn't
help
the
computer
compute
better.
It
deserve
make
sure
the
signals
arrive.
At
the
same
time
so
honest,
we
don't
really
need
to
worry.
We.
B
B
A
B
B
This
is
a
big
idea
and
it's
really
I
think
it
should
be
a
very
important
idea
in
brain
theory,
and
so
we
need
to
get
that
out
there
and
we
never
have
it
start
being
critiqued
and
you
know
getting
pushback
on
what's
right
and
what's
wrong
about
it.
But
the
next
thing
that
someone
will
be
able
to
get
out
of
this
will
be
a
paper
well,
of
course,
we'll
we'll
post
it
on
probably
by
our
archive
before
it
gets
accepted
for
publication.