►
From YouTube: Neuralink Paper Review - Numenta Research Meeting
Description
Subutai will quickly review Elon Musk & Neuralink's new paper and we will discuss. https://www.biorxiv.org/content/10.1101/703801v1
A
B
A
A
A
A
A
So,
even
though
there
was
a
lot
of
hype,
it's
actually
not
an
unreasonable
thing
that
they're
doing
and
I
think
the
summary
you
had
a
really
nice
summary
here
so
basically,
summary
is
Norling,
picked
the
best
of
existing
technology
and
pushed
it
forward
in
a
number
of
important
dimensions
and
most
impressively
as
an
integrated
implantable
product
that
goes
beyond
the
current
state
of
the
art.
So
that's
a
pretty
positive
statement
from
a
scientist
and
there
were
several
others
like.
A
And
if
you
can
to
somewhat
this
kind
of
summarizes
it
it's
not
that
any
one
of
these
things
is
that
radically
new
thing,
but
what
they,
what
experimental
neuroscientists
often
do?
Is
they
get
these
parts
and
they
create
their
own
rigs
and
their
own
devices?
Here,
it's
like
a
really
nicely
packaged
thing
with
a
couple
of
really
nice
interesting
twists
that
I'll
get
to
it
a
bit,
but
it's
sort
of
a
it's
sort
of
packaged
almost
like
a
product
which
is
interesting
for
this
kind
of
stuff.
A
D
A
B
E
A
C
B
F
C
A
I
think
so
the
further
purpose
of
this
kind
of
live
streaming,
the
you
know
Norland,
is
not
the
only
firm
building
it,
but
there
are
the
farms,
but
certainly
he's
not
gonna
get
as
much
hype
as
Elon
Musk's.
That's
this
okay!
So
what
else?
So
the
overall
telling
I
think
was
pretty
positive.
Another
kind
of
meta
comment,
so
this
paper
was
posted
on
by
archive
I,
think
it
showed
up
this
morning
or
I
think
it
must
have
posted
it.
Yesterday.
C
C
C
E
A
E
A
A
A
A
Yeah
today,
I'm
not
a
single
business
like
so
yeah
there's
several
probes,
and
then
this
is
another
design
where
you
have
a
thing
to
call
it
a
tree
tree
structure
probe
where
these
channels
are
hanging
off
to
the
side.
So
you
can
get
a
higher
density
of
channels
here.
So
these
are
two
two
designs
that
they're
working
with.
C
D
A
E
E
C
They're
not
the
same
so
the
other
probe,
the
probe
the
chip
will
use
today
are
more
hardly
they're.
You
know
they're
a
little
spikes
and
stuff
and
they're
not
flexible
as
much
I
think,
and
so
probably
those
of
that,
even
when
they're
the
usability
changes
because
of
the
Brad's,
the
brain
moves
and
pulsates,
and
so
on,
you
lose
contact
with
the
same
cells.
The
machines
are
only
useful
to
the
extent
that
you're
contacting
the
same
cells
are
measuring
the
same
places
in
the
brain
and
I.
C
E
C
A
C
This
is
one
of
the
today
when
they
do
these
implants
and
people
and
so
on.
They
don't
last
very
long
at
all.
This
may
be.
It
could
be
a
couple
of
days
because
they
program
the
machine
to
recognize
these
patterns
and
and
then
two
days
later,
they're
putting
up
different
patterns
different
places
in
the
brain.
So
then
it
didn't
have
to
train
all
over
again.
So
it's
you
know,
yeah.
A
A
C
A
A
A
C
A
A
A
C
A
So
this
is
an
existing
thread
and
now
there's
a
couple
of
things
to
notice
in
this
video
I'll
just
play
this
first,
so
here's
the
robot
coming
in
and
inserts
a
thread,
and
you
see
all
this
flashing
going
on
and
there's
a
delay
and
then
the
next
thread
so
what's
going
on
here,
the
robot
is
pretty
impressive.
What's
going
on
is
that
the
robot?
You
know
automatically
picks
up
the
the
thread
with
this.
You
know
you
see
this
little
loop
here,
so
it
picks
it
up.
Just
like
a
sewing
machine.
A
It's
like
a
needle
in
my
sewing
machine
and
then
it
kind
of
plans,
the
trajectory
and
all
those
flashing
lights
that
are
going
on
is
it's
sort
of
registering
against
fiduciary
points
in
the
brain,
and
it
can
do
a
very
precise
than
that
alignment
and
and
figure
out
where
to
do
it
and
they
want
to
avoid
the
vasculature,
so
the
blood
vessels
and
stuff
like
that,
they
then
map
those
out
and
it
can
avoid
that.
It
apparently
also
tracks
the
motion
of
the
brain
and
and
can
use
you
know.
Second,
very
precisely
insert
these
things.
A
A
It
basically
retton's
is
independent,
it
can
do
it,
despite
that
it
can
track
the
motions.
So
it's
a
pretty
impressive
device
and
I,
don't
think
that's
been
done
before
so
you
can
see
that's
what
all
this
flashing
is
going
on
and
the
part
of
the
delay
is
just
that.
It's
planning
out
exactly
where
to
put
things
and
time
time
everything
correctly.
E
A
B
A
A
A
So
they
yeah
that's
a
good
point.
They
do
not
do
spike
sorting,
so
they
don't
track
individual
neurons.
What
they
do
is
they
do
spike
detection,
so
they
can
detect
when
a
neuron
is
fired,
but
for
each
contact
there
are
multiple
neurons
around
it
and
to
typically
actually
isolate
neurons.
You
have
to
look
at
multiple
probes
and
do
this
spike,
sorting
business
I
think
you
can
do
it
with
a
single
probe
too,
but.
A
Could
yeah
it's
all
computationally
expensive,
so
you
can't
do
that
in
real
time
right.
So
this
is
oh,
they
can't
do
they
can't,
despite
sorting
in
real
time,
but
they
can
do
spike
detection
in
real
time,
and
so
that's-
and
this
is
pretty
common
in
BMI.
You
don't
do
this
by
sorting
you
just
look
at
spikes,
then
you
assume
we
can
get
enough
signal
from
that.
To
do.
D
C
Yeah
was
it
matters
in
some
applications
and
it
doesn't
matter
technologies
and
depending
what
their
ultimate
goal
here
is,
and
you
know
proposed
original
to
my
goal.
Was
this
whole
sort
of
merging
your
brain
with
super
intelligent
computers
in
that
would
likely
require
the
real,
very
specific
neuron
to
neuron
connection,
if
you're
doing
the
prosthetic
we
play,
you
know
control
like
for
the
injured
person
with
an
arm
or
something
like
that-
probably
not
yeah.
B
C
A
D
E
E
E
C
E
C
You
don't
really
know
which
ones
you're
reading
from
you're
reading
from
a
bunch
of
them,
and
so
this
is
my
point
when
you
they're,
just
gonna
put
a
voltage
or
some
current
at
some
point
in
the
brain
and
it's
gonna
act,
it'd
be
you're,
a
bunch
of
dendrites
will
be
a
bunch
of
axons
will
be
a
go
to
cell
bodies.
You'll
have
a
sort
of
a
massive
tacked
on
on
a
great
number
of
different
things,
and
it.
D
C
D
B
B
C
C
C
What
the
tissue
does
before
they
remove
it
or
maybe
they'll
remove
it,
so
they
would.
They
would
just
an
open
brain
and
a
live
patient,
so
you
will
awake
on
the
operating
table.
You
can't
feel
this.
They
would
do
a
current
injection
into
a
small
area
of
the
brain
and
they
would
ask
you
what
you
did
anything
if
you
noticed
it.
C
You
might
say:
oh
I,
hear
a
song
or
I
just
saw
my
grandmother
or
whatever
I
just
had
a
vision
of
something
or
there
was
blue,
and
so
they
can
tell
that's
a
vision
thing
or
whatever,
and
then
they
were
trying
to
isolate
areas
in
my
related
to
language.
So
they
want
to
pick
an
area
that
doesn't
look
like.
It
was
cut
you
they
want
to
exercise
a
certain
amount
of
tissue,
but
they
don't
want
an
exercise
tissue
that
would
leave
you
deaf
or
you
know
the
language
deficits
so.
C
C
C
We
should
be
able
to
plug
our
brains
directly
into
the
computer
so
that
we
not
left
behind
so
as
the
guys
get
super
intelligent,
we
get
to
go
along
with
them
and
for
that
they're
Korea
time
to
have
a
very
high
bandwidth,
high
capacity,
very
targeted,
bi-directional
communication
at
a
minimum.
So
that's
why
the
originally
they
want
to
be
able
to
go
both
ways.
I
think
in
the
short
term,
there's
going
to
be
this
opportunity
for
both
neuroscience
research
and
prosthetics
and
in
those
cases,
especially
on
the
prosthetic
you
don't
need
to
integrate
the
brain.
C
F
C
Heard
anyone
doing
that,
maybe
they
are
better
but
they're,
still
destroying
it
again.
It's
if
someone
can
control
an
arm
or
move
a
mouse,
something
like
that
and
they're
so
far
from
I.
Unless
you
know
something
I
know
I've
never
heard
anyone
saying:
oh
now
we
got
it
working.
Let's
worry
about.
You
know
the
proprioception
part
I.
E
C
B
A
C
E
C
D
C
E
A
E
E
E
C
A
E
A
E
A
E
C
Research
and
open
brain
surgery
and
people
are
quadriplegics
and
then
there's
this
dream
of
the
future
and
the
dream
of
the
future.
Is
you
just
go
in
and
say:
hey
I
want
a
new
brain
implant
and
they
go
sure
Lucas.
You
know
sit
down
here
and
you
know
bingo
and
it's
all
done
you
don't
EE
brain
opened
up
and
you
know
I
think
that's
where
they
want
to
go
yeah
because
he
wants
to
get
to
this
to
be
like.
This
is
a
this.
Is
the
future
of
humanity.
A
A
D
E
A
E
B
E
C
C
Wrote
about
flipping
a
brains
that
are
what
about
and
I
wrote
about
this
thing,
and
so
I
was
a
little
nervous
like
actually
you
know,
but
I
think
I.
Think
I
don't
need
a
lot
of
anything.
I
wrote
my
take
on
this,
because
you're
interested
was
that
this
idea
of
uploading,
your
brain
to
prevent
being
left
behind
by
super
AI,
think
it's
ridiculous,
because
I
don't
think
it
is
gonna,
be
super,
AI
and
I
think
you
need
to
be
worried
about
left
behind,
but
it
not
not
even
sure
you
want
to
do
that.
C
F
C
I
mean
anything
that
we've
done
to
help
the
sort
of
the
medical
world
would
be
great
and
it's
gonna
go
on
anyway.
So
who
knows?
Maybe
in
the
future
we'll
find
out
that
the
technology
that's
developed
for
this
kind
of
quadriplegic,
stuff
becomes
available
to
us,
for
you
know,
maybe
additional
sensory
inputs
that
we
find
useful.
It's
not
a
it's
less
of
a
merging
with
the
computer
and
it's
more
of.
B
C
Hard
to
believe
that
you
could
actually
just
totally
merge
with
the
computers
up,
don't
be
sort
of
like
you,
you,
the
computer,
you
have
to
be
modeling
on
your
cortex,
you
know
they're
the
right
kind
connections
and
it
all
can
be
very,
very
specific.
So
these
are
sort
of
broad.
These
are
sort
of
very
adult
intermittent
subsets.
They
just
you
know,
check
some
current
or
they
they
abstract
some
signals,
but
not
at
the
level
of
you
know
a
Billy
neurons
and
you
know
high
fidelity
and
that
kind
of
stuff.
You
think
this
type
of
work.