►
From YouTube: HTM Hackers’ Hangout (Apr 7, 2017)
Description
- Documentation efforts / Wiki migration
- HTM Chat with Jeff
- Temporal Pooling issues (Matt)
- NuPIC 1.0 update
A
Alright,
we
are,
why
welcome
to
the
HTM
hangar
tack
out,
it's
April,
seventh
and
just
so
happens
that
my
kids
walked
in
the
door,
so
you're
probably
going
to
hear
them
in
the
background.
Let's
they
just
got
home
from
school
there.
They
are
kids,
birds,
dog
source,
ignore
all
that
stuff.
I
wanted
to
talk
about
couple
things.
Then
we've
got
a
blend
Gastner
on
on
here.
Maybe
he
wants
to
talk
about
a
couple
things
we'll
see,
but
let
me
run
through
my
list
first,
just
as
an
update,
some
of
things
I'm
I've
been
doing.
A
One
is,
if
you're
on
HTM
forum,
you've
noticed
I've
created
a
quite
load
new
topics
that
aren't
really
knew
their
old
topics
that
I'm
moving
from
the
old
github
wiki
for
new
pic
and
putting
them
into
discourse.
So
I
wanted
to
try
and
clean
up
all
of
the
surface
area
of
all
the
possible
documentation
and
sources
for
information
for
new
pics.
So
I
am
trying
to
deprecate
this
old
wiki
clean
up
all
the
content
to
leave
the
things
that
are
all
obsolete
and
move
the
things
that
I
think
might
valuable
in
2h.
A
Damn
forum
where
we
can
continue
to
discuss
them
just
like
we
do
any
other
forum
post,
but
it
still
still
be
wiki,
so
anybody
could
edit
them
and
that
is
ongoing.
I
think
I
have
I'm.
Pretty
close
I
got
a
few
more
wiki
pages
to
figure
out
what
to
do
with,
but
I'm
almost
done
so.
If
you're
getting
a
bunch
of
annoying
notifications
emails,
that's
almost
over.
So
thank
you
for
hanging
in
there.
A
The
other
thing
I
wanted
to
know.
If
you
didn't
notice
that
there's
a
really
I
think
interesting
discussion
between
me
and
death
on
our
YouTube
channel
to
search
for
HTM
chat
with
Jeff
Hawkins
or
something
and
I
sat
down
with
Jeff
for
about
an
hour
and
a
half
and
just
talked
about
since
remoter
inference
theory
how
it
fits
in
with
existing
HTM
theory,
and
we
had
a
really
nice
conversation.
I
broke
it
out
into
playlists,
like
eight
videos
or
something.
So
if
you
haven't
seen
that,
hopefully
you
can
check
that
out.
A
Another
thing
since
Marcus
oops
here
we
talked
last
time
a
lot
about
temporal
sequences
and
trying
to
identify
and
extract
them
and
something
I
totally
missed.
I've
realized,
since
that
last
discussion
is
basically
what
I'm
talking
about
is
temporal
cooling
and
what
I
want
to
do.
This
I
want
identify
these
sequences.
It's
I
think
that's
temporal
flowing,
so
it's
not
something
that
we
can
do
right
now
in
the
current
state
of
New
Creek.
A
So
I
just
wanted
to
clarify
that,
and
the
last
thing
on
my
things
to
talk
about
is
we're:
making
good
progress,
I
think
towards
a
new
pic
1
point:
0,
I'm
pretty
sure
it's
going
to
happen
this
year.
I,
don't
know
it's
not
going
to
give
any
browsers,
but
I'm
pretty
sure
is
gonna
happen
this
year,
I'm
spending
a
bunch
of
time
on
documentation,
putting
a
quick
start
guide
and
we're
renaming
things
as
we're
writing
the
documentation.
If
something
doesn't
look
right
or
it
seems
fishy
here,
there's
an
API
and
it
should
be
exposed.
A
That's
the
only
way
to
do
something
we
are
fixing
that
stuff
as
we
go
along.
So
we
could
have
a
nice
example
of
how
to
do
this
with
each
different
API
that
we
expose
a
a
clean
way.
So
that's,
it
includes
also
more
name
changes
that
will
be
coming,
so
the
code
base
is
going
to
rock
a
little
bit
over
the
past
over
the
next
few
months
as
we
clean
this
stuff
up.
A
So
we
can
present
a
nice
clean
signature,
API
signature
to
the
public
and
that's
primarily
what
I'm
going
to
be
working
on
over
the
next
couple
months.
I
really
want
to
get
the
API
cleaned
up
and
documented
so
that
when
we
go
one
point,
oh,
we
can
say
so
without
any
hesitation
we're
ready
to
go
one
point
other
than
that.
I
am
going
to
I'm
still
working
on
HTM
school
episodes,
but
I
haven't
done
anything
think
since
the
last
episode,
so
I'm
not
really
ready
to
work
on
it.
Yet
I'm
still
thinking
about
it.
A
C
I
just
want
to
thank
everybody
over
at
New
mentum
for
all
the
questions
you
answer
on
the
forums,
all
everything
you
answer
in
hacker
hangouts,
like
I've,
asked
so
many
questions
and
gotten
so
many
answers.
My
mind
is
blown
that
you
guys
are
so
nice
about
this
up
and
sooner
or
later,
I
will
have
a
finished
project
and
I
will
pay
you
back,
but
it's
still
going
to
be
a
little
while
that's
it
well,.
A
A
I
think
he
sometimes
just
joined
in
to
listen,
which
is
fine,
I,
don't
see
any
comments.
Let
me
just
check
the
YouTube
page
really
see.
Chandan
is
on
the
YouTube
page.
What
are
the
objectives
of
nuclear
one
point
when
he
plans
to
launch
on
tensor
thing?
Ok,
so
that's
from
Camden
much
to
online
here
no
plans
to
integrate
with
tensor
cloaks
I
think
the
community
wants
to
do
hi
chandan
in
place
a
long
time
tends
to
drink.
A
Now
you
should
turn
off
with
a
YouTube
page
and
then
principal
in
London,
I'm,
gonna,
I'm
gonna
meet
you,
so
you
can
get
your
YouTube
channel
turned
awfully
something
that
this
happened
to
me.
So
I'll,
try
address
your
question.
The
no
new
mint
is
not
planning
to
write
any
integrations
with
tensorflow,
but
I
know
that
there's
some
some
work
on
that
in
the
community,
so
somebody
out
there
might
think
we've
done
a
little
investigation,
but
we
don't
need
it
at
the
moment
to
do
it.
A
So
we're
not
working
out
right
now
and
the
objectives
for
one
point:
oh
I
think
I
wrote
this
like
years
ago,
but
it
was,
we've
done
a
lot
of
them.
The
main
thing
was
to
extract
the
c++
from
the
Python,
which
we've
done
and
I
think
we
have
a
really
nice
clear
distinction.
There
got
all
our
sweet
stuff
over
on
the
seaside,
I
think
so
we
have
a
Python
on
the
code
base.
I
think
if
that
was
a
requirement,
I
think
we're
in
good
shape,
for
that
we've
got
a
clean
dependency
for
new
cricket.
D
Another
important
thing
to
point
out
about
a
one
point
out
release
would
be
stability
and
I
don't
mean
stability
of
the
code.
You
know
at
runtime,
I
mean
changes
in
the
API
is
and
stuff
like
that.
So
you
have
some
some
more
concrete
barriers
between
different
components
and
will
be
more
mindful
of
breaking
changes
and
stuff.
Like
that,
that's.
A
A
very
good
point:
it
would
be
great
if
we
could
all
spend
some
time
to
look
at
the
current
API
and
look
for
the
fragility
in
it,
because
it
may
be
there's
ways
that
we
can
save
ourselves
a
lot
of
work
in
the
future
by
doing
that
work
right
now.
That's
a
very
that's
a
good
point,
because
once
we
go
one
point,
oh
you
can
go
back
and
if
we
have
to
break
change,
make
breaking
changes.
A
B
Answer
this
Marcus
I'd
say
that
there
are
a
series
of
experiments
that
are
just
like
that
have
been
just
quietly
posted
to
new
pic
research
and
that's
going
to
continue
but
like
presenting
these
and
like
deciding
how
to
talk
about
it
publicly
is
just
like
that
can
be
very
time
consuming,
and
it's
just
not
something
we're
spending
time
on
right
now.
So,
like
we
have
new
pic
research
that
you
can
watch
the
commits
there
and
we'd
just
kind
of
run.