►
From YouTube: Bridging HPC and Quantum Computing - Panel
Description
Bridging HPC and Quantum Computing - Panel
B
Okay,
great
well
yeah,
hello,
everyone
and
welcome,
and
thanks
for
for
attending
so.
A
C
B
I
am
a
an
engineer
here
at
nurse.
That's
coming
at
this
very
much
from
the
classical
Computing
perspective
and
I
thought
helpful,
just
to
very
quickly
sort
of
set
the
context
for
for
this
panel.
So
my
my
work
at
nurse
involves
supporting
real-time
data
analysis
out
of
experiments.
So
imagine
you
have
a
particle
accelerator
or
microscope,
and
you
want
a
very
rapidly
analyze
data,
so
you
can
make
decisions
about
the
experiment
and
so
I'm.
This
is
also
my
perspective.
B
Around
Quantum,
Computing,
obviously
I'm
not
saying
that
this
is
the
only
perspective.
I
just
want
to
set
some
context,
all
right,
and
so
we
have
an
awesome
panel
assembled
here
and
I,
think
it
would
be
really
helpful
if
all
the
panelists
sons
kazra,
who
has
already
introduced
himself,
so
it
would
be
really
useful
to
have
all
the
panelists
introduce
themselves
and
just
briefly
give
their
background.
B
D
B
D
Well,
it's
very
nice
to
be
here.
I
I
appreciate
this
invitation
to
participate.
My
name
is
Alex
McCaskey
for
those
of
you,
I
haven't
met
yet
a
little
bit
of
background
about
me.
So
my
research
Focus
over
the
last
you
know
eight
nine
years
or
so
has
really
been
on
languages,
compilers
and
Associated
software
tooling
for
heterogeneous
Quantum,
classical
Computing.
D
So
really
thinking
about,
like
you,
know,
kind
of
moving
away
from
this
remote
execution
model
that
we're
so
familiar
with
currently
and
what
happens
when
we
do
start
to
integrate
classical
and
Quantum
Resources
together
in
a
kind
of
more
familiar
fashion.
Like
we've
seen
in
recent
heterogeneous
classical
systems,
can
we
really
use
a
qpu
as
an
accelerator
to
an
existing
heterogeneous,
Computing
environment?
That's
kind
of
what
I've
been
focused
on
just
to
kind
of
go
down
the
list
here.
D
I
got
my
graduate
degrees
in
pure
physics
at
Virginia
Tech,
where
I
was
focused
a
lot
on
modeling
electron
transport
across
single
molecule
magnets,
and
so
this
was
an
opportunity
for
me
to
really
kind
of
hone
in
on
some
interesting
kind
of
qubit
potential,
qubit
Technologies,
but
really
to
kind
of
build
out
a
lot
of
my
background
in
software
engineering
and
computational
physics
that
I
that
I
have
since
leveraged
kind
of
in
future
or
jobs.
After
my
graduate
degrees
were
completed.
D
So
after
that
I
moved
on
to
Oak
Ridge
National
Laboratory
I
was
born
and
raised
in
Tennessee.
So
this
was
a
very
good
fit
for
me.
I
I
started
in
2014
as
staff
at
in
the
computer
science
and
math
division
at
Oak,
Ridge
National
Lab
in
the
computer
science,
research
group
and
then
later
on
in
a
group
that
was
formed
for
Beyond,
more
technology.
So,
specifically
thinking
about
things
like
Quantum
and
neuromorphic
I
did
a
lot
of
work
in
leading
kind
of
a
lot
of
the
software
efforts
around
Oak
Ridge.
D
At
the
time,
I
was
the
software
research
lead
for
the
quantum
Computing
Institute
at
Oak
Ridge
National
Laboratory
I
also
had
an
opportunity
to
lead
a
couple
of
research
thrusts
in
different
doe
projects.
That
ORNL
had
received
funding
for
specifically
the
quantum
Computing
application
team
project,
as
well
as
the
advanced
research
and
Quantum
Computing
project.
D
So
then,
after
Oak
Ridge
I
actually
have
been
at
Nvidia
about
a
year
coming
up,
November
1st,
it's
been
about
a
year
and
I'm
the
quantum
Computing
software
architect
here
at
Nvidia
and
I'm.
The
technical
lead
for
this
new
programming
model
called
Coda,
the
quantum
optimized
device
architecture,
where
we're
really
trying
to
learn
a
lot
from
the
Cuda
world
and
kind
of
apply
that
in
the
kind
of
quantum
Computing
world.
So
so
I'll
go
ahead
and
stop
that's
kind
of
my
brief
background
and,
like
I,
said
happy
to
be
here.
E
D
F
B
E
Okay,
so
hi
everybody,
I'm,
Nick,
kamuki
I
am
the
CTO
at
queera
Computing,
so
I
actually
could
have
slide
together
to
introduce
query
Computing,
but
I
will
say
a
few
things
about
myself:
I'm,
an
atomic
physicist
physicist
by
training,
so
I've
I've
been
doing
that
for
about
25
years
since
I
met
Steve
Chu,
who
helped
invent
laser
cooling.
That's
one
of
the
core
technologies
that
we
use
to
build
quantum
computers
here
at
queer
Computing.
So
we
actually
do
a
range
of
different
things.
E
The
company
started
in
about
2019
we're
up
to
about
40
people
now
and
we
do
basically
the
full
stack
of
quantum
Computing.
All
the
way
down
to
the
hardware
and
the
hardware
that
we
build
is
based
off
of
neutral
atom
technology.
So
you
can
see
an
example
of
that
down.
The
bottom
left
corner
of
the
slide,
where
you
can
see
a
grid
of
little
purple
dots
each
one
of
those
dots
is
actually
a
single
atom.
E
We
also
build
tools
that
allow
us
to
understand
how
these
Quantum
processes
that
are
so
large
in
scale
behave.
Most
of
them
truthfully
are
above
the
simulation
threshold,
they're
not
easy
to
simulate
with
classical
Hardware,
but
we
do
have
tools,
software
tools
that
will
help
you
to
simulate
our
systems
up
to
the
60
or
70
Cubit
scale.
We
also
have
people
in-house
that
work
on
designing
algorithms
and
fitting
to
applications
on
the
specific
Hardware
as
well.
E
So
if
you
call
us
up
say
that
you
are
interested
in
particular
problem,
we
have
a
team
of
people
that
will
think
hard
about
that
problem
and
how
to
implement
Quantum
algorithms,
on
this
neutral
item
architecture
that
we
built.
So
you
know
we
have
a
very
solid
scientific
founding.
We
work
very
closely
with
research
groups
at
Harvard
and
MIT.
The
company
was
actually
founded
by
people
who
came
out
of
Laboratories
of
Misha
Lucan
vladen,
bulatich,
Marcus,
Greiner
and
Dirk
England
as
well,
and
you
can
right
now
start
using
these
machines
with
us.
E
So
our
product
lineup
consists
of
their
Quantum
processor,
we're
launching
our
first
device,
which
is
called
the
killer
over
the
next
couple
of
months,
and
then
also
the
software
tools
that
help
with
that.
So
you
can
look
online
and
find
all
of
those
kinds
of
things,
but
I
think
I
will
not
take
too
much
of
the
panel's
time
up
with
more
introduction
for
myself
or
my
company
and
hand
it
off
to
the
next.
B
Thank
you,
nice.
Next
up
we
have
Anna.
G
Hello,
let
me
share
my
slide
really
quickly.
Okay,
so
all
right
so
hello,
my
name
is
Anastasia
would
call
I
go
for
Anna.
So
I
am
a
career
scientist
at
the
career
texture
group
at
amcr,
lbnl,
so
I.
Just
we'll
quickly.
Tell
you
about
my
background.
G
I,
don't
don't
have
any
background
in
Quantum,
Computing
I
got
my
digital
design
and
electronics
degree
at
National,
Technical
University
at
Kiev,
and
then
I
got
a
microelectronics
master
in
Montpelier
France
and
then
I
I
had
my
PhD
in
HPC
architecture
simulation
in
France
as
well.
So
as
you
can
see,
I
don't
not
necessarily
a
Quantum
scientist,
but
when
I
joined
the
lab
I
was
I,
expressed
my
interest
and
and
been
lucky
enough
to
be
invited
to
participate
in
Quantum
projects.
G
So
briefly,
my
research
interests
are
so
I'm
I'm
working
on
designing
the
control
hardware
for
Quantum
and
among
other
efforts.
My
current
project
include
the
advanced
Quantum
test
by
where
I'm
developing
Quantum
Isa
as
kind
of
an
efficient
software
and
Hardware
interface.
So
if
you
ever
heard
about
Quasar
so
I
am
who
is
working
on
that
so
I'm
also
interested
in
in
the
computer
architecture
group
that
is
now
lead
by
John
shelf.
G
So,
as
you
well
know,
he
he's
working
a
lot
in
the
Beyond
Wars,
Technologies
and
architecture.
So
superconductor
and
electronics
is
one
of
the
of
the
potentially
promising
technology.
So
I've
been
involved
in
this.
So
we
have
a
super
tools,
IR
program
and
within
this
program
we
also
exploring,
if
we
can,
you
know,
do
cryogenic
control
for
Quantum
the
equation
or
something
like
that
and
yeah.
G
So
I'm
also
interested
in
different
architectures
at
the
edge
and
and
we
are
using
a
lot
of
fpgas
and
new
technologies
related
to
the
fpgs
to
building
the
prototypes
and
exploring
different
architectures.
So
that's
my
broad
background.
B
Thank
you
very
much.
Next
up
we
have
Ravi.
F
F
My
work
focused
on
multi-mode
circuit,
Crown
electrodynamics,
so
specifically
I'm
using
resonators
as
as
a
Quantum
memories
here
at
Berkeley
lab
I'm,
the
head
of
measurement
at
the
advanced
Quantum
test
bed,
which
you
heard
about
from
Castro's
talk,
and
so
we'll
go
too
much
into
detail
there,
but
just
to
highlight
some
of
the
research
efforts
that
I've
been
involved
in
as
part
of
the
aqt
generally
three
broad
areas,
Quantum
control,
so
being
able
to
enable
the
sort
of
Novel
types
of
of
control
that
are
capable
on
superconducting
Quantum
processors,
including
work
with
ternary
Quantum
logic
and
the
multi-cubic
gates
that
cash
were
highlighted
earlier.
F
Furthermore,
we've
also
been
working
on
understanding
how
how
noise
affects
Quantum
processors,
one
of
the
most
important
issues,
in
my
opinion,
with
Quantum
processors
and
along
these
efforts,
both
being
able
to
Benchmark
Quantum
processors
and
understand
how
these
errors
behave
in
the
context
of
algorithms
on
cells
and
along
those
lines,
we've
been
working
on
techniques
to
sort
of
optimize
how
these
algorithms
perform
given
the
capabilities
of
our
processors
and
also
the
noise
that
affects
them.
B
Thank
you
very
much
and
finally,
we
have
Norm.
H
B
H
I'm
Norm
Tubman,
thanks
for
having
me
on
the
panel,
it's
great
to
be
here.
I'm,
a
research
scientist
at
NASA,
Ames
I
have
two
or
three
quick
slides.
So
a
lot
of
people
introduce
their
background.
My
background
as
an
undergrad
was
computer.
Science.
I
did
my
PhD
in
physics
at
Northwestern,
and
then
I
did
post-docs
both
in
physics
and
computational
chemistry.
H
All
my
work
is
more
on
the
computational
side,
so
I've
been
working
on
algorithms
to
solve
chemistry,
physics,
problems,
lattice
models,
things
like
that
things
that
have
to
do
with
hamiltonian
simulation
classically
and
in
that
regard
you
know,
I
spent
a
long
time
working
with
HPC
resources,
trying
to
figure
out
how
to
optimize
these
algorithms,
where
the
largest
systems
were
the
most
complicated
problems
we
can
solve
on
the
classical
side
and
in
the
and
I
have
some
examples.
On
the
slide
of
various
molecules.
We've
looked
at
over
the
years.
H
We've
done
some
having
Benchmark
heavy
benchmarking.
Some
simple
molecules
like
Benzene
on
the
bottom
left.
There
we've
done
some
more
complicated
transition,
metal
stuff,
like
our
poor
friend
in
the
middle
there,
and
even
larger,
like
organic
Frameworks,
and
things
like
that.
So
there's
a
lot
of
Material
Science
problems
in
chemistry,
and
you
know,
physics,
lattice
models
that
you
know
I've
been
interested
in
in
my
studies,
a
quick
summary
of
what
you
know,
I'm
working
on
right
now
in
terms
of
papers
and
trying
to
understand
how
all
this
fits
on
to
qualm
computers.
H
You
know
I'd,
say
probably
I've
listed
some
three
questions
here
that
have
been
sort
of
the
focus
of
my
research
in
the
last
few
years
and
a
lot
of
it's
on
where's
this
interface
between
what
we
can
do
with
classical
Computing
and
where
will
Quantum
Computing
really
pick
up
and
help
us
push
forward
into
new
directions
and
I
have
a
couple
papers
listed
here
that
sort
of
represent
various?
H
You
know
activities
that
I
do
in
terms
of
my
research,
which
is
developing
new,
classical
algorithms,
an
example
of
something
we
called
ASCII,
which
is
a
configuration
interaction,
type
method
that
we
developed
in
the
last
few
years.
We
developed
new
Quantum
algorithms
in
my
group,
so
I
have
an
example
here
that
I
worked
on
with
people
on
lbl
that
uses
various
applied
math
techniques
to
try
and
accelerate
the
way
we
can
get
convergence
to
ground,
States
and
other
properties
of
chemical
systems.
H
I
do
a
lot
of
benchmarking,
with
large
groups
of
people
who
have
their
own
classical
or
Quantum
algorithms,
and
we
try
and
compare
the
efficiency
of
algorithms
trying
to
understand
what
the
state
of
the
art
is
and
I
also
develop.
A
lot
of
Open
Source
software
with
various
collaborators
on
the
side
which
I
have
here
on
the
bottom
left
an
example
of
quantum
Monte
Carlo
code.
We
wrote
called
qmc
pack
a
couple
of
years
ago.
So
that's
a
quick
summary
of
me
and
the
different
research
techniques
that
I'm
involved
with.
B
H
B
All
right,
so,
thanks
to
the
panel
for
introducing
themselves,
I
I've,
said
in
the
chat
that
I
encourage
the
audience
to
ask
questions
in
the
in
the
chat
as
we
now
transition
to
the
discussion.
Part
I
also
I
would
like
to
encourage
the
audience
to
raise
their
hand
if
they
want
to
ask
a
question-
and
we
can
you
know
Shepherd.
This
share
with
the
questions.
B
B
B
G
So
I
will
start
from
the
end,
so
you
you
said:
what's
gonna
change,
you
know
when
we
transition
from
this
to
something
different,
so
I
think
that
the
the
important
the
main
difference
is
going
to
be
that
the
way
it's
working
right
now
and
the
way
it's
probably
gonna
work
and
the
nisk
most
of
the
nisk
air
computers,
it's
very
experimentalist
kind
of
oriented,
so
it's
different
from
the
way
that
people
are
using
nurse
resources,
as
I
understand
so
being
like
an
experimental
Computing
means
that
that
you
need
to
have
a
lot
of
support
from
the
experimentalist
from
the
physicists.
G
So
you
cannot
do
this
alone.
You
need
to
have
you
need
to
have
help
then
yeah,
so
there's
no
kind
of
clear
protocols.
G
There
is
no
clear
standards,
most
of
the
most
of
the
problems
that
I
demonstrated
right
now
they
are
oriented
at
like
proof
of
concept
and
not
in
actually
solving
the
problem
and
and
the
metrics
that
you
know,
I
kind
of
I
used
for
those
experimental
Computing
are
also
oriented
into
like
exploring
the
quality
of
those
computations
and
not
actually,
you
know
Computing
like
solving
the
problem.
So
all
of
those
you
know
points
include,
we
won't
be
able
to
transition
or
well,
not
not
because
of
that.
G
Well
and
those
are
the
kind
of
major
differences
that
I
see
in
between,
like
the
experimental
Computing
and
like
the
regular
Computing.
So
once
we
once
we
transition,
we'll
we'll
see
like
something
different,
so
I'll
just
stop
talking
here.
Yeah
I'll
take
all
the
time,
if
not.
B
Does
anyone
else
want
to
volunteer
to
give
an
answer
before
I
call
on
someone
foreign.
H
I'm
happy
to
say
something:
I
mean
you
gave
a
multi-parp
question,
I'll,
try
and
say
a
couple
things
about
the
first
part.
So
you
asked:
what's
the
role
of
HBC
in
designing
building
and
operating
quantum
computers?
And
you
know
I'm,
not
part
of
a
bunch
of
different
collaborations.
We're
part
of
you
know.
One
thing
that
I'm
a
part
of
is
a
a
doe
Center
at
fermilab.
H
It's
called
sqms
and
we're
using
various
levels
of
computing,
not
all
of
it's
HPC,
but
no,
no
multi-core,
processors
at
the
very
least,
and
some
of
it
is
HBC
to
do
different
parts
of
Designing,
a
clone
computer.
There.
One
aspect
I'd
say:
is
that
we're
looking
at
Material
science
questions
and
this
isn't
my
expertise
per
se,
but
people
are
starting
to
do
density,
functional
Theory
calculation,
phase
field
calculations
within
our
Center
to
try
and
help
experimentalists
build
better
Hardware
to
do
better
design
of
materials
and
I.
H
Think
that's
that
might
not
be
underappreciated,
but
it
gets
less
press
in
terms
of
where
HPC
can
really
help
out,
but
they're.
Definitely
Material
Science
problems
that
we're
interested
in
the
other
aspect.
I'd
say
that
we're
looking
at
now-
and
this
you
know-
comes
up
a
lot,
there's
a
there's,
a
lot
of
different
architectures
out
there
and
within
our
Center
at
sqms,
we're
looking
at
QD
systems
and
trying
to
design,
Gates
and-
and
you
know,
efficient
gate
sets
and
the
pulses
for
these
Gates
also
takes
a
quite
a
bit
of
computing
power.
H
So
I
just
want
to
bring
up
those
two
things
that
you
know
have
been
really
discussed
within
some
of
the
things
that
I'm
working
on
recently,
but
there's
certainly
a
lot
of
other
places
where
some
of
the
other
panelists
might
discuss
in
terms
of
Designing.
New
algorithms
and
testing
out
q-day
architectures
also.
E
Yeah,
actually,
maybe
I
can
follow
up
there.
I
think
Norman
had
a
grab
both
of
most
of
these
points,
but
I
can
think
of
kind
of
basically
five
areas
that
the
quantum
Computing
and
HPC
or
Superior
Computing
kind
of
intersect.
One
is,
you
know,
doing
things
with
a
quantum
computer
that
you
would
otherwise
do
with
HPC
systems
and
I
think
there's
some
great
pie,
charts
from
the
usage
of
nurse
facilities
that
that
really
help.
E
You
understand
like
what
areas
to
to
attack
or
things
like
that,
the
second
one
is
and,
and
people
touched
on
this
already
simulating
the
performance
of
quantum
computers
and
using
them
to
affect
the
design
and
so
forth.
The
third
is
hybrid
Quantum
computing,
and
you
know
I'm
not
sure
anybody's
really
put
their
finger
on
the
exact
use
case
or
need
for
having
massively
scaled,
classical
Computing
working
in
tandem
with
Quantum
Computing,
but
I
think
it's
probably
not
long
before.
E
We
do
that,
and
you
know
that's
essentially
using
high
performance
Computing
to
make
Quantum
Computing
better.
In
a
sense,
you
could
also
do
hybrid
in
a
different
way,
which
is
using
the
quantum
computer
to
make
the
high
performance
Computing
better
and
I.
Think
there
you
might
think
of
quantum
Computing,
as
essentially
being
a
hardware
accelerator
that
I'm
very
specific
problems
you
know,
might
be
able
to
give
an
advantage.
Obviously
we're
not
quite
there
yet,
but.
I
E
Fifth,
one
and
I
think
this
one
is
really
interesting,
especially
now
is
that
supercomputing
facilities
form
a
great
kind
of
framework
for
figuring
out
how
to
construct
things
like
user
facilities,
where
you
have
you
know
scientifically
literate
people
who
are
using
these
these
kind
of
Novel
machines,
but
need
a
lot
of
interaction
with
the
people
on
the
ground
who
are
building
the
hardware
to
try
to
find
a
way
to
make
it
work
and
I
think
that's.
E
D
Put
that
on
I
tend
to
agree
with
everything
that
you
just
mentioned:
I
think
that
when
I
see
when
I
heard
this
part
with
this
question,
first
I
thought
all
right:
building
designing
and
operating
Quantum
Computing.
Where
does
HPC
fit
into
this
and
I
think
in
the
near
term,
you're
really
going
to
be
reliant
on
HPC
and
multi-gpu
architectures
to
allow
programmers
to
interface
with
that
potential
Quantum
resource
in
an
emulated
way
right
and
so
starting
to
think
about.
D
You
know
what
kind
of
algorithms
I
can
start
to
experiment
with
maybe
they're
kind
of
purely
Quantum
algorithms,
but
maybe
they
are
hybrid-
that
incorporate
some
kind
of
pre
and
post
processing
on
GPU
resources,
GPU
and
CPU
resources,
but
the
HPC
system
really
sits
in
as
an
emulated,
qpu
right
and
so
you're.
Starting
to
see
this
with
some
of
the
the
simulation
technologies
that
have
come
out
in
the
last
few
years.
You
know
one
that
I
was
involved
in
at
Oak
Ridge.
D
Was
this
tensor
Network
Quantum
virtual
machine
simulator
that
kind
of
leveraged
different?
You
know
tensor
decompositions
to
kind
of
scale
out
how
much
how
many
qubits
you
could
potentially
consider
in
a
simulation,
but
now,
of
course,
with
Nvidia.
We
have
you
know
great
back-ends
for
existing
simulators
with
Q
Quantum
and
Q
tensor
net
and
Q
State
Bank,
and
so
I
really
think
that
in
the
next
five
years,
five
to
ten
years,
potentially
that
the
HPC
resource,
the
classical
HPC
resource
will
be.
D
This
perfect
kind
of
you
know
testing
ground
for
us
early
on
in
algorithm
development,
but
also
in
programming
model
development.
How
do
I
interface
users
with
that
combined
resource?
How
do
I
interface
with
them
and
provide
them
an
efficient
programming
model
foreign.
B
What
do
you
like
to
give?
Maybe
your
your
more
experimental
oriented
perception.
F
Yeah
I'd
be
happy
to
so
along
the
lines
of
I'll.
Give
this
perspective
specifically
for
superintendo
doing
qubit
platforms.
F
From
our
perspective,
what
makes
I
guess
arguments
different
than
sort
of
many
of
the
the
atomic
type
of
qubits
is
that
all
of
our
qubits
are
unique
in
the
sense
that
they're
macroscopic
objects
and
to
do
this
they
can
have
variation
due
to
fabrication,
inaccuracy
and
really
getting
this
designed
to
performance
chain.
F
Secure
I
think
requires
a
really
accurate
simulations,
especially
if
you
start
going
to
scale
and
we
quickly
find
that
when
we
design
these
things,
if
we
try
to
lay
out
you
know
many
qubits
on
a
single
chip,
it
can
quickly
get
very
computationally
expensive
to
simulate
the
electromagnetics
of
that
system,
and
you
know
we
have
proxies
and
we
can
do
it
quite
well.
We
can
do
it.
F
We
think
we've
shown
with
our
performance
below
the
one
percent
level
and
even
down
to
the
point,
one
percent,
but
if
we
want
to
get
to
the
point
where
we're
making
fault
tolerant
quantum
computers,
we
really
need
to
push
that
even
further
and
push
that
on
scale
even
further.
And
that
can
be
a
challenge.
F
I
will
be
able
to
do
that
with
HPC
I
think
we
need
to
be
smart
without
we
need
some
combination
of
of
proxies
that
can
scale
to
larger
systems
and
also
be
able
to
throw
a
lot
of
computer
to
edit.
B
So
so
what
you
say
that
that
is
the
biggest
challenge
for
scaling,
Quantum
Hardware
in
terms
of
just
number
of
qubits
or
gate
depth,
or
are
there
other
scaling
challenges
that
you
foresee.
F
So
I
think
you're
pointing
at
what
is
the
central
issue.
Is
that
that
we
have
good
ways
of
understanding
how
errors
are
at
the
local
scale,
and
we
really
have
a
good
sense
of
of
being
able
to
optimize
on
that
scale.
F
F
However,
understanding
how
errors
transform
evolve,
as
we
scale
up
it's
really
challenging,
and
in
particularly
for
supporting
qubit
some,
because
the
environment
for
each
qubit
can
be
considerably
different
and
the
qubits
themselves
can
have
variants
being
able
to
predict
how
that
will
scale
is
it
can
be
a
real
Challenge
and
if
we're
trying
to
build
say
you
know
a
thousand
qubits
per
logical,
qubit,
then
transitioning
to
that
that
fault,
tolerant
limit
may
be
more
challenging
than
just
understanding
what
the
local
air
landscape
is.
B
Right
thanks.
Maybe
this
is
a
good
segue
to
Nate.
Can
you
give
us
your
perspective
about
neutral
atoms
and
why
you
believe
in
this
technology.
E
Yeah
obviously,
I
put
a
big
personal
bet
on
it.
So
I,
you
know,
I
might
give
a
little
bit
of
a
biased
response,
but
you
know
I
in
some
sense,
some
of
the
strengths
are
are
Robbie
alluded
to,
which
is
the
with
neutral
atoms.
We
don't
have
to
work
too
hard
to
make
every
atom
identical.
E
In
fact,
you
know
that's
that's
the
way
they
come
by
Design,
so
you
know
we
we
don't
spend
as
much
time
learning
every
individual
qubit
in
fact
we're
prone
to
throw
them
away
and
every
quarter
second
or
so,
and
and
load
up
some
new
ones
with
the
confidence
that
you
know
very
much
like
the
second
is
defined
by
oscillations
and
Adam's.
You
know
our
qubits
are
that
good?
We
have
other
problems
with
scaling
things
up,
so
I
mean
I.
I,
don't
want
to
give
you
a
too
optimistic
of
a
view.
E
E
So
programmability
is
really
a
challenge
for
us
and
so
much
so
that
the
first
devices
that
we're
making
now
are
not
fully
programmable
devices,
we're
using
them
for
Quantum
simulation
things
like
that,
and
in
fact
we
don't
have
a
single
control
Channel
per
per
atom
in
our
machine
right
now,
but
we're
working
hard
to
do
that.
So
you
know
we.
We
have
a
an
uphill
battle
in
that
sense
of
now
that
we
have
been
able
to
find
a
way
to
get
to
very
large
scale.
E
Now
we
have
a
different
kind
of
engineering
challenge,
which
is
how
do
you
control
all
those
qubits?
So
we
have
to
look
into
new
Hardware
to
do
that,
we
have
people
on
our
team
who
are
using
integrative
photonics,
which
is
itself
kind
of
an
edge
technology,
to
try
to
build
laser
beams
that
can
control
atoms
precisely
enough
to
do
High,
Fidelity
Gates
and
do
that
at
scale.
E
You
know
we've,
the
technology
is
still
moving
very
fast.
We
had
a
big
Discovery
within
the
last
year
of
how
to
move
atoms
around
in
real
time
and
preserve
their
Quantum
coherence.
That's
that's
huge
because
it
does.
E
It
means
that
we
no
longer
need
a
single
laser
control
Beam
for
every
atom
in
the
device,
but
we
can
move
them
to
the
control
means
so
you're,
starting
to
see
some
of
the
kinds
of
things
that
that
were
essential
for
making
classical
Computing
move
forward,
which
is
you
know,
essentially,
how
do
you
Multiplex
control
in
a
quantum
computer
and
now
that
we're
starting
to
solve
some
of
those
challenges?
E
I
think
we're
overcoming
the
primary
hurdle
of
scaling
up
control
in
in
the
atomic
systems,
so
I
mean
the
simple
answer:
is
it's
easy
to
go
to
scale
so
big
that
you
know
we
got
to
build
the
machine
to
see
what
it's
going
to
do.
I
think
that's
exciting!
E
It
means
that
that
every
machine
that
we
put
out
you
know
it's
it's
Unique
on
Earth,
even
super
Computing
facilities
can't
tell
you
what
they're
going
to
do
that's
exciting,
but
the
next
step
is
is
to
try
to
get
the
control
in
so
that
we
can
actually
fully
program
the
devices.
B
Right,
thank
you,
and
actually
this
is
there's
a
question
directed
to
Alex
in
the
chat
and
I
think
it
can
actually
be
generalized
to
to
other
people.
B
So
please,
chairman,
if
you
have
any
thoughts,
so
the
the
question
sings
the
prices
of
cool
Quantum
and
then
asks
so
then
Kareem
asks
if
you
can
open
source
examples
and
then
also
mentions
that
resources,
like
the
manual
scarce
with
with
these
examples,
so
maybe
also
more
broadly
speaking,
I
think
a
lot
of
us
in
HPC.
Have
this
worry
that
you
know
it's
very
hard
to
find
resources
about
learning,
Quantum
Computing,
so
maybe
Alex.
D
Yeah,
no
definitely
so
you
know
I'm
I'm,
not
on
the
Q
Quantum
engineering
team
per
se,
I'm
more
on
the
coda
programming
model
side,
but
I
do
work
with
those
folks.
All
the
time
and
I
do
know
that
there
are
public
examples.
D
Github
under
the
Nvidia
organization,
so
you
can
go
check
those
out
the
the
figures
that
were
the
Benchmark
figures
that
were
shown
this
GTC
I
believe
are
currently
in
you
know
somebody
else,
probably
you
know
one
of
the
product
people
could
correct
me,
but
they're
they're,
still
internal,
but
I'm
sure
that
they'll
probably
be
you
know,
open
sourced
at
some
point
in
the
future
and
I'm
sure
that
there
will
be
a
you
know.
Python
example
for
that
as
well.
D
No
I
think
that
the
kind
of
overall
documentation
and
textbook
question
for
interfacing
classical
HPC
domain
computational
scientists
with
Quantum
is
a
really
good
question
and
I
think
that
you're
starting
to
see
it
with
you
know
a
few
efforts
here
and
there
I
think
that
one
resource
I
like
to
point
people
to
is
the
the
quiz
kit
textbook
is
a
really
nice
resource
for
kind
of
learning.
D
A
lot
of
the
early
kind
of
textbook
algorithms
and
algorithm
implementations
out
there
and
then
kind
of
taking
that
and
porting
it
to
other
Frameworks
is
relatively
straightforward,
but
I
think
that
it
gets
that
more
of
a
more
general
question
about
kind
of
Workforce.
Development
and
you
know
kind
of
developing
programming
models
that
are
at
a
high
enough
level,
that
kind
of
get
us
away
from
the
kind
of
core
Quantum
circuit
construction,
kind
of
methodology
and
kind
of
bring
it
more
toward
the.
D
You
know,
typical
types
that
we're
interested
in
like
integers
and
floats,
and
can
I
do
algorithms
on
integers
and
floats
in
the
quantum
register
instead
of
the
classical
register,
but
then
also
kind
of
thinking
about
presenting
being
these
hybrid
algorithms,
hybrid,
Quantum,
classical
algorithms,
in
more
of
a
application-centric
way
and
abstracting
Away
the
details
of
the
underlying
Quantum,
you
know
Isa,
essentially
right,
like
the
the
gate
set.
You
know
those
are
things
that
we're
currently
working
on
at
Nvidia
and
I
know.
A
lot
of
people
are
working
on.
D
You
know
kind
of
across
the
field,
but
that
I
think
will
be
necessary
to
really
kind
of
bring
the
existing
HPC
domain.
Computational
science
and
research
community
on
board,
with
leveraging
Quantum
co-processing
in
these
existing
heterogeneous
workflows.
B
Yeah,
thank
you.
Any
anyone
else
have
perspectives
on
this.
B
If
not
I'm
going
to
actually
ask
Norma
question:
are
a
variational
algorithms
ever
going
to
work
at
scale
and
and
with
that
we
mean
show
Quantum,
Advantage
and,
and
also
are
there
alternatives
to
hybrid
algorithms?
Sorry,
are
they
alternative,
hybrid
algorithms
that
are
promising.
H
Hi,
okay,
this
is
I
feel
like
this
is
a
tricky
question
and
there's
going
to
be
a
lot
of
opinions
on
this
and
I
know.
Cuera
has
some
papers
on
this
also
where
they've
tried
to
do
some
variational
stuff
at
pretty
large
system
sizes.
So.
H
Everybody
yeah
what
I'd
say
is
I
think,
there's
a
lot
of
different
promising
routes
in
which
variational
algorithms
might
be
useful
right
now,
I'd
say
other
than
well.
Let's
just
say
if
we
were
to
focus
strictly
on,
say
quantum
chemistry
applications
trying
to
see
where
things
are
right
now
on
Quantum,
Hardware,
I,
don't
think
I've
seen
almost
anything
on
superconducting
Hardware,
at
least
that's
more
than
like
10
to
16
qubits,
and
if,
if
that's
indication,
it
might
be
quite
hard
to
scale
up
to
order
system
sizes.
H
But
that
being
said,
we
do
have
a
lot
of
tools
on
which
we
can
optimize
in
the
presence
of
noise.
There's,
no
inherent
reason.
We
can't
scale
larger
and
I
do
think
it's
a
fundamental
algorithm
that
is
needed
now
and
might
also
be
needed
in
the
future.
Even
when
we
do
have
full
tone
Hardware
we're
going
to
need
to
do
state
preparation.
Quite
broadly,
there
are
lots
of
different
algorithms
to
do
it.
Variational
algorithms
still
might
be
prominent,
even
as
we
go
forward
with
Quantum
Hardware
advancing,
but
I
also
think.
H
There's
no
part
of
the
question
was:
are
there
a
lot
of
different
strategies
that
in
which
you
maybe
change
up
the
algorithm
or
change
things
a
little
bit
to
make
these
sorts
of
algorithms
more
useful,
I?
Think
there's
a
lot
to
be
said
by
pre-training,
where
you
start
making
use
of
clever
classical
algorithms
to
get
circuits
that
are
are
somewhat
optimized.
They
get
you
off.
Baron
plateaus
and
these
other
sorts
of
things
that
have
been
talked
about
in
the
literature
and
then
you
use
quantum
computers
as
a
refinement
step.
H
There's
also
been
discussions
of
using
no
decent
wave
function
or
decent
State
preparation
as
input
into
other
sorts
of
algorithms
out
there,
such
as
Monte
Carlo
techniques,
and
things
like
that.
H
So
I
am
optimistic
that
we
will
be
able
to
see
some
Quantum
Advantage
with
variational
algorithms
and
I'd
say
the
main
X
Factor
isn't
necessarily
whether
it
will
be
useful
or
not
it's
whether
there
are
other
algorithms
that
overtake
it
in
some
sense,
when
we
go
to
The
Fault,
tolerant
error,
whether
or
not
you
know-
and
this
is
part
of
my
research
to
some
extent-
I
develop
new
algorithms.
All
the
time
other
people
are
developing
new
algorithms
all
the
time.
It's
not
a
static
field.
H
So,
if
we're
to
just
freeze
all
the
other
algorithms
right
now,
I
bet
variational
algorithms
will
have,
you
know,
will
eventually
cat
will
eventually
be
extremely
useful
in
the
future.
If
people
keep
developing
new
algorithms,
especially
as
we
move
to
The
Fault
on
our
error,
we
might
have
a
whole
another
set
of
techniques
that
we
can
use.
There
are
already
a
lot
of
other
algorithms
also
that
can
be
used
in
Fault,
tolerant
error.
B
Right,
thank
you
and,
and
I
also
noticed
that
Kareem
has
turned
this
video
on.
So
maybe
do
you
have
a
a
follow-up
question
and
also
would
encourage
the
rest
of
the
panel
to
chime
in
with
perspectives
on
variation,
algorithms,.
I
Thank
you
thank
much
I
appreciated
about
that.
Maybe,
if
I
follow
up
with
the
algorithm,
what
do
you
think
about
qpe
taking
charge
or
replacing
a
variational
algorithms
at
some
point.
H
H
H
I
honestly,
don't
know,
what's
going
to
happen
with
those
three
algorithms
in
The,
Fault,
tolerant
error.
There's
a
lot
of
discussion
that
you're
not
going
to
want
to
do
variational
algorithms
because
of
all
the
measurements
you
have
to
make
when
you're
in
default,
tolerant
error
and
that
the
time
Evolution
type
algorithms
like
qaoa,
like
adiabatics
data
preparation,
might
be
better,
but
I
still
think
it's
up
in
the
air
a
little
bit
and
I
still
think
the
variational
approaches
might
make.
H
Even
if
you
don't
need
them
necessarily
to
make
the
shortest
circuits
possible
your
default
torn
error,
you
can
go
as
long
as
you
want.
You
still
have
other
issues
to
consider
like
how
long
do
your
algorithms
run
just
because
you're
in
default
torrent
error
doesn't
mean
that
you're
going
to
wait
two
years
for
an
algorithm
to
finish
so
you
might
want
still
want
short
circuits.
You
still
might
want
to
do
the
shorter
circuit
you
can,
you
might
be
only
be
able
to
get
to
that
with
a
variational
type
of
algorithm.
H
So
there's
lots
of
trade-offs
that
there's
lots
of
trade-offs
to
consider
and
I'm
I'm
still
investigating
myself
within
my
own
research
to
try
and
figure
out
the
answer
to
that
question.
B
No
thank
you
I.
Think
like
looking
at
the
time,
I
would
like
to
ask
casra
the
final
question,
and,
and
perhaps
this
is
a
bit
of
a
question
I
suppose
that's
a
little
bit
more
peppy,
so
I'd
be
interested
to
hear
what
your
perspective
is
so
in
for
classical
Computing,
the
bang
for
your
buck
I.E,
the
return
on
investment
has
been
proven
and
I
would
be
really
interested
to
hear
your
perspective.
What
that
would
look
like
for
Quantum
computing.
B
Oh
wow,
okay,
so
the
so
you
can
kind
of
Imagine.
A
this
idea
of
an
HPC
facility
is
a
user
facility
right.
So
it
represents
a
fairly
large
investment
of
money
and
it
returns
both
signs
at
scale
and
also
science.
That
can't
otherwise
be
done.
B
If
you
now
replace
the
the
HPC,
the
classical
HPC
facility,
with
either
Quantum
augmented
facility
or
a
Quantum
facility.
What
does
your
return
on
investment?
Look
like.
J
Okay,
let
let
me
try
to
answer
the
closest
question
that
I
think
applies.
It's
look,
we
are,
and
we
have
been
over
the
last
couple
of
decades
in
in
very
uncertain
Waters
when
it
comes
to
Quantum
Computing
right.
It
took
more
than
two
decades
since
the
proposition
of
quantum
Computing
for
the
very
first
algorithms
to
come
out
to
start
coming
out
in
the
form
of
Shores
algorithm
and
search
algorithms
and
whatnot,
and
still
to
this
day,
the
best
quantum
algorithm
designers.
J
Don't
really
have
good
intuition
for
exactly
how
to
design
Quantum
algorithms
right.
This
is
basically
we're
still
very
much
learning
and
the
answer
to
your
question
really
largely
does
depend
on
on
algorithm
development.
What
we
do
know
for
sure
is
that
there
are
complexity,
classes,
that
kind
of
separate
Quantum,
heart
and
Quantum
Museum,
classical
heart
and
and
all
of
that
and
there
we
do
know
that
there
are,
you
know,
classes
of
problems
that
will
be
vastly
easier
solved
through
Quantum
Computing.
J
What
we
don't
really
know
is
exactly
you
know,
all
of
that
landscape
in
detail.
Where
are
the
boundaries,
and
especially
in
the
nisk
era,
and
as
we
approach
fault
tolerance?
Where
are
those
boundaries?
Where
do
hybrid
algorithms
lie?
What
role
do
they
play?
J
And
you
know
there
are
a
lot
of
people
who
have
no
go
theorems
and
talk
about
how
things
won't
scale
and
won't
work
out
and
the
reality
is
you
know.
History
shows
us
that
it
always
looks
like
that
until
somebody
does
it
right,
and
so
we
we're
basically
we're
big
Believers,
that
we
will
be
able
to
access
those
complexity,
classes
and
bang
for
buck
will
not
be
really
a
question.
J
B
Oh,
thank
you
I.
Think
I
really
like
that
answer.
So
I
think
we
are
at
the
top
of
the
hour.
I
want
to
once
again
thank
all
the
panelists
for
that
time
and
their
interesting
perspectives
and
their
expertise
and
I
would
like
to
thank
the
the
audience
for
you
know,
having
showing
interest
and
and
and
you
know
just
attending
and
asking
questions
all
right
back
to
you.
Katie
and
Dan.
C
Can
I
just
say
one
thing
so
I
want
to
thank
all
the
panelists.
Also.
This
was
a
really
great
discussion
and
I
really
appreciated
a
month.
Many
others.
The
the
one
comment
about
a
place
like
nurse
and
Department
of
energy
having
this
infrastructure
that
that
we
have
the
the
knowledge
and
the
the
know-how
and
and
we
you
know,
we
interact
with
the
end
user,
with
the
person
trying
to
trying
to
do
the
computation
and
the
the
people
that
are
actually
running
developing
and
really
have
the
detailed
knowledge
of
the
systems.
D
C
One
thing
I
forgot
to
mention
when
I
spoke
earlier,
is
you
know
this?
This
is
part
of
a
larger
effort
that
nurse
God
does,
which
is
looking
at
all
kinds
of
advanced
Computing
Technologies
from
both
the
near-term
and
media
term
and
the
long
term
it's
kind
of
the
longer
term
outside,
but
that
group
is
led
by
Nick
right
get
nurse
and
they
really
did.
That
group
really
does
excellent
job.
Looking
looking
at
these
things,
and
so
if.
C
A
Yeah,
thank
you,
Richard
and
thank
you
again
to
all
the
panelists
and
Johannes
for
moderating
the
discussion.
We'll
have
a
16
minute
breakdown,
we'll
we
will
be
back
at
20
past
the
hours
for
for
our
technical.