►
From YouTube: HTM Hackers' Hangout - Jun 1, 2018
Description
https://discourse.numenta.org/t/htm-hackers-hangout-jun-1-2018/3939
HTM Hackers’ Hangout is a live monthly Google Hangout held for our online community. Anyone is free to join in the discussion either by connecting directly to the hangout or commenting on the YouTube video during the live stream.
If you have something specific you’d like to discuss, or if you just want to learn more about the HTM Community, please join HTM Forum at https://discourse.numenta.org. We have active discussions about HTM theory, research, implementations, and applications.
A
B
B
B
One
of
them
is
an
extensions
of
the
columns
paper
that
we
published
last
year,
where
we
speculated
and
theorized
that
a
cortical
column
and
learn
complete
models
of
objects
and
that
multiple
columns
together
in
resolve
ambiguities
and
so
on.
A
big
piece
of
that
paper
was
figuring
out.
What
we
assumed
that
there
was
a
location
signal
that
told
you
or
informed
you
of
locations
of
features
that
are
actually
on
an
object
in
the
objects,
reference
name,
and
that
allows
you
to
kind
of
make
predictions
as
you
as
you
move.
B
So
if
I'm
moving
from
here
to
the
bottom
of
the
mug
I
can
I
can
make
really
good
predictions
of
what
I'm
going
to
feel
here,
because
I
know
about
this
object.
I
know
where
features
are
on
this
object,
an
independent
of
how
the
object
is
kind
of
oriented
with
respect
to
my
body
and
to
my
sensors
I
can
still
make
a
prediction.
B
So
over
the
last
year,
Marcus
and
others
have
been
really
working
on
that,
and
you
probably
heard
from
previous
hangouts
and
other
stuff
that
we've
been
looking
at
something
called
grid
cells
which
is
grid
cells,
are
a
type
of
and
the
neurons
that
are
in
the
hippocampal
area,
in
the
entorhinal
cortex
and
so
Matt's
got
a
great
blog
post
on
this.
I
won't
go
into
detail,
but
basically
grid
cells.
Allow
you
to
compute
locations
in
an
environment
in
the
space
of
an
environment.
B
So
we
propose
that
there
are
grid
cell,
like
analogs
in
cortical
columns,
and
that
these
grid
cell
type
neurons
are
actually
computing,
the
location,
the
features
and
and
features
in
and
up
better
explain
just
like
what
they
do
in
the
hippocampal
areas,
where
they
compute
the
location
of
the
animal
in
the
environment
reference
space.
So
this
is
this
paper
kind
of
feels
in
that
missing
piece
that
was
in
the
in
the
columns
paper
from
last
year.
We're
working
through
a
lot
of
the
details
of
how
you
implement
this.
What
the
properties
are.
B
B
Of
writing
up
right
now,
so
hopefully
we'll
publish
that
sometime
later
this
year,
these
papers
always
take
a
lot
longer
than
you
actually
think
they
would.
So
that's
one
thing
we're
doing.
Jeff
is
actually
writing
a
larger,
bigger
picture
paper
on
the
implications
of
having
grid
cells
in
cortical
columns.
B
So
there's
a
lot
of
different
implications
of
this,
and
once
you
realize
that
cortical
columns
not
only
get
sensory
input
but
also
compute
locations
of
objects,
and
it
kind
of
allows
to
give
you
a
framework
for
a
lot
of
other
computation
that
might
be
going
on
in
college,
so
he's
kind
of
working
through
those
implications
and
think
of
it
from
a
very
high
level
view
what
is
a
cortical
column
doing?
How
does
cut
art
there's
a
proposal
of
grid
cells
and
cortical
columns,
change,
I,
think
about
what
the
cortex
is
right.
B
So
that's
another
paper,
a
third
paper
we're
working
on
is
headed
by
mirko's,
guess,
scientists
here
and
that's
looking
at
mathematical
implication
than
a
mathematical
sort
of
theory
around
grid
cells.
So
that's
I
would
say
a
little
bit
earlier
in
the
process
and
it's
you
know,
he's
sort
of
working
through
exactly
what
would
be
in
that
paper
and
I.
Think
Marcus
is
involved
in
that
as
well.
B
B
What
else
we
have
a
conference
that
we're
going
to
in
Seattle
in
July
called
the
cns
conference?
It's
a
computational
neuroscience
conference,
that's
held
in
different
places
around
the
world
every
year
and
we
have
four
different
presentations
in
July
in
Seattle.
So
that's
going
to
be
a
lot
of
fun
so
pretty
busy
summer
to
interns
here
as
well,
working
on
different
aspects
of
our
research,
a
lot.
A
B
B
Sometimes
it
gets
rejected
right
away,
in
which
case
you
have
to
start
from
yeah
with
another
journal.
If
it
goes
for
review,
which
is
more
common,
it
takes
about
two
to
three
months,
depending
on
the
journal,
to
get
your
first
set
of
comments
back
and
then
you
have
to
address
them
and
you
send
it
again,
and
usually
you
got
another
the
comments
you
send
it
back.
So
from
the
time
you
first
submitted
it's
somewhere
around
four
to
eight
months
before
it
actually
appears
depending
and
that's
where
a
fairly
fast
Journal.
B
B
Luckily,
however,
however,
I
will
say
that
what
we
also
do
at
the
time
we
submit
our
first
version
to
the
journals
we
always
tend
to
put
that
same
version
into
a
preprint
server.
So
a
bio
archive
our
archive,
so
the
community
will
get
access
to
it
much
sooner,
but
in
terms
of
the
end
and
a
peer-reviewed
publication
and
what
appears
it's
a
slow
process
all
right,
so
stay
tuned.
B
C
A
A
I'm
excited,
okay.
Speaking
of
you
know,
information
from
research
getting
to
the
community
I
think
well,
we'll
have
a
nice
way
to
do
that.
Coming
up
so
I've
been
doing
all
these
interviews
with
neuroscientists,
but
I'm
planning
on
stopping
that.
First
of
all,
it
takes
a
lot
of
time
and
and
I.
Think
I
can
do
it
much
easier
in
a
podcast
format
and
I
think
it'd
be
less
intimidating
for
people
and
I
think
it'd
be
able
to
get
more
people
to
do
it.
A
So
Christine
aver,
our
vp
of
marketing,
is
working
she's,
going
to
be
in
charge
of
a
podcast,
doesn't
move
forward
and
I'm
gonna
help
with
content
and
interviews
and
stuff,
and
so
she's
already
got
a
list
of
interesting
things
to
talk
about,
and
so
eventually
I'd
like
to
pull
in
Jeff
into
the
podcast
and
he'll
he'll.
He
always
gives
interesting
insights
into
what's
going
on
in
research
as
well.
So
so
that's
a
good
way.
A
I
think
that
we
can
sort
of
expose
some
of
some
of
our
some
of
our
thinking
and
thinking
of
other
people
that
were
interacting
with
tubes
I.
Think
that's
that's
especially
interesting
when
we
have
scientists
coming
here
or
people
in
industry
coming
here
to
talk
about
how
what
they're
doing
relates
to
what
we're
doing.
That's
that's
exciting,
so
I
think.
A
B
You
so,
with
the
interview
with
the
Northside,
the
video
stuff,
you
pretty
much
we're
limited
by
people
who
are
coming
here
so
that
if
you
have
to
kind
of
physically
videotape
that
with
the
podcast,
are
you
gonna?
What
do
you
think
about
doing
sort
of
more
remote
interviews,
since
this
is
just
audio
interviews
or
or
some
other
thought
about,
that
cuz
they're,
a
bunch
of
people
we
collaborate
with
that
are
not
able
to
come
here
on
a
regular
basis,
the
sky's,
the
limit,
the
sky's,
the
limit?
Okay,
all
right!
So
yeah!
A
Right,
yeah,
of
course,
it's
very
organic.
This
type
of
stuff
I
mean
honestly
I
like
to
let
you
guys
keep
heads
down
on
research
and
I'll,
keep
heads
down
on
my
stuff
and
you
generally
wouldn't
we
come
up
and
look
at
each
other
like
hey
good
job,
so
alright.
Moving
on
to
the
next
topic,
I'm
gonna
share
my
screen,
quick
update
on
what
I've
been
working
on
there
we
go
so
I
always
want
to
share
as
the
open
source
community
manager
hold
on
I
just
button.
A
Mouse
isn't
one
okay
words!
I!
You
see
my
screen
now:
yeah,
alright,
there's
some
code
on
it.
So
I
love
sharing
code,
the
open
source
guy.
So
here's
what
I've
been
working
on.
It's
an
open
source
repo
that
contains
a
bunch
of
visualizations.
Why
is
that
I?
Don't
you
can
see
this?
That's
called
building
HTM
systems
and
what
it
is
it's
it's
in
our
hgmd
github
group,
and
it
contains
all
of
the
visualizations
that
I'm
ending
up
packaging
and
putting
into
this.
A
You
know
this
document
that
I'm
working
on
I
thought
I.
Think
I
talked
about
this
at
one
of
the
last
hackers
hangout.
So
all
these
visualizations
as
interactive
stuff
is
completely
open
source
and
you
can
run
it
really
easily.
You
can
see
them
all
really
easily.
You
don't
have
to
run
a
server
you
just
if
you
can
get
this
coat
off.
Github
there's
this
one
file
called
widgets
and
if
you
open
it
just
from
your
file
system,
it
looks
like
this
all.
A
A
Development
of
new
widgets,
here's
what
I'm
currently
working
on
it's
cyclic,
encoding
that
they
don't
pay
attention-
that's
not
quite
a
category
yet
but
I'm
going
to
show
how
you
can
do
category
encoding
with
this
like
like
a
coder
or
less
discreet
types
of
encoding
as
well.
So
this
is
type
of
stuff
I'm
working
on
to
support
this
building,
HTM
systems
documents-
and
anybody
can
can
run
this
just
by
getting
the
code.
A
Oh,
it's
got
all
of
the
visualizations
from
my
grid
cell
blog
post
to
are
here
so
it
basically
any
visualizations
I'm
doing
in
d3,
for
for
community
for
educational
stuff
is
going
to
be
in
this
repository.
So
I
just
want
to
share
that.
If
anybody
cuz
I
know
a
lot
of
people
in
the
and
the
HTM
hackers
community
are
interested
in
visualizations,
you
can
hook
this
up
to
any
HTML
honestly
I'm,
going
to
write
an
HTM
in
JavaScript
called
simple
HTML.
A
It's
going
to
be
sort
of
the
driver
behind
some
of
these
visualizations,
but
for
right
now
it's
just
encoders!
There's,
there's,
there's
really
no
no
HTM
going
on
yet
ok!
The
last
thing
I
want
to
talk
about
was
nupoc
1.0
0.5
just
got
released
that
got
released
because
Mac
OS
newer
versions
had
a
compatibility
problem
with
captain
proto,
so
we
needed
to
upgrade
captain
proto
in
the
core
and
PI
Captain,
Pete
and
nupoc,
and
so
we
cut
a
new
new
pic
or
we
put
a
new
new,
that's
for
ram.
A
So
there's
a
few
installation
problems
that
I've
seen
reporting
on
Mac
OS,
Mac
OS
that
have
to
that
looks
eerily
like
serialization
issues,
I'm
hoping
that
this
new
version
will
clear
those
up.
So,
if
you're
having
any
installation
issues
on
Mac
OS
at
all
and
they
throw
like
KJ
exceptions
or
something
like
that,
that's
typically
some
type
of
a
cat,
proto
air
update
two
new
big
1.0
five.
That's
it
I'm
out
of
updates
for
those
hackers
hangout.
Does
anybody
have
anything
else
to
say
before
we
close
out?