Dan: Hello and welcome
to the quantum divide.

This is the podcast that talks about
the literal divide between classical it.

And quantum technology.

And the fact that these two domains.

Are will and need to
become closer together.

Ultimately, this podcast is
going to be a combination of

one-to-one podcast recordings.

Myself and my esteemed colleague, Steve.

Discussing different topics around
quantum networking, quantum computing.

And so on.

And w we'll pepper, the agenda with a
number of interview topics I'm looking

forward to bringing in a whole broad
selection of individuals working in and

around the quantum technology space.

That should.

Really bring some fantastic.

Conversations and
interesting debate as we go.

We're going to try and focus on.

Networking topics.

Quantum networking actually is
more futuristic than perhaps

the computing element of it.

But we're going to try
and focus on that domain.

But we're bound to experience many
different tangents both in podcast

topics and conversation as we go on.

And just as a reminder I'm Dan home.

I'm not classically trained in quantum
physics or physics of any sort, really.

So I'll be the one on the call.

In the conversations, asking the
stupid questions and hopefully

those stupid questions serve as a
platform for you to help understand

some of the topics that we discuss.

Enjoy.

Steve: Okay.

So let's see.

So I guess we start with the discussing
why Cloud quantum computing is a good

idea and why that's gonna be the.

Only way to access quantum
computers in the near future.

Dan: Yeah I thought we'd record this
episode to talk about cloud computing

because, the global hyperscalers,
they're really embedded into the

way a lot of computing is done these
days in many different ways where

it's hosting infrastructure or just
providing access to compute capabilities

to process individual functions.

But the benefits of cloud
computing are clearly gonna.

And already are simplifying and
democratizing access to quantum computers.

So the benefits of
elasticity cost savings.

Let's start with that.

It, I guess it costs quite a lot to
build and manage a quantum computer.

Yeah.

I think if you Google for that,
Other search engines are available,

obviously, but if you search for
that there's normally at least six

zeros on the end of it, but I'm
sure it's more complex than that.

Cause the teams required
to manage the thing.

May, maybe it's worth us just
digging into that, first of all,

like what makes managing and owning
a quantum computer expensive.

Steve: Yeah, I guess the first
things are the environment that

the quantum computer has to sit in.

So firstly, that has to be a
controlled environment, probably has

to be relatively secure environment
limiting access to who can get into

these spaces in the first place.

And that's just, okay,
that's just the start.

Where does the quantum computer go?

Then when you think about the
technology, it's needs to be

used to do the quantum computing.

It probably needs to be cooled
to extremely low levels.

And then again, you need some
kind of refrigeration schemes

that are very difficult to build.

Probably a lot of chemicals used
to perform the refrigeration,

the tubing, all the connections
in place just to PR provide the

chemicals into the refrigerator.

So you need good infrastructure in the
first place, and then all that needs

to be maintained and refilled , and all
the, watching every step along the way.

Okay.

And then that's just the
physical components of it.

And then you need the
electronic components as well.

And then electronic components
and the software then need

hardware, very precise hardware,
very advanced technology as well.

And all that needs a very specialized
person to do that type of thing.

You can't just pick someone
up off the street to start

building your quantum computer.

So all those things considered,
I can understand why six

zeros are in that number.

Dan: Yeah, easily.

I, I've read, I often see the term sub
kelvin, Which means going all the way

down to between zero and one kelvin.

Which is incredibly low temperatures.

So yeah, the hardware to
that is extremely specialist.

And expensive to maintain.

Yeah.

So it's not something you're
gonna buy and on your desk at home

Steve: I was gonna say, although there
are some efforts towards building desktop

quantum computers, but generally you're
not gonna have one in your house because

it's, the environment is not good enough.

It's too much infrastructure
at the moment, although we,

Dan: It's a bit early
for that, I would say.

I would agree.

And from what I can tell the accuracy
of the measurements and things and the

precision that's needed in the hardware
is it's a lab environment, isn't it?

It's a science lab environment.

Steve: Yeah.

Dan: People have to wear white
coats with pens in their top pocket.

And walk around with clipboards.

Steve: It's a lot involved with
maintaining a quantum computer.

And then, yeah, then there's the ability
to access them even within the lab.

So you can't just go up to the quantum
computer and plug in your SB stick

and run a program, most likely.

You need ways to interface
with the quantum computer.

And those ways are also not easy.

It's not the average person who can
interface with the quantum computer.

So generally you need some F P G A
programming or something like that, to,

to really, I guess once you have the
FPGAs set up, you can interface with the

FPGAs and that simplifies things too.

But generally it's not the average
person who's gonna have one.

It's not like an easy access to either
programming them or maintaining them.

Yeah.

Dan: You just gotta look in the
media and pictures of quantum

computers and obviously it's a big
frame of cooling apparatus with

ultimately a chip at the bottom
which often you won't be able to see.

But are the FPGAs, you're talking about
driving the signals into that chip?

Basically they're specialized hardware
which send the pulses or whether

it be light or electromagnetism
or microwaves or something at the

individual qubits in the chip.

That whole chain of activity
is extremely specialist, right?

It's not something you can buy off

the shelf.

Steve: That's right.

Yeah.

I think every company probably does it.

Different at the moment and I would be
doubtful that there are any standards

that exist yet for that as well.

I think it's too new to have any
standards, so you can't just go

to any manufacturer and know that
whatever they give you is gonna work.

You have to probably develop your own
software on top every time you might get

the F P G, but how to program the F P G
A, I think this is completely open still.

Yeah.

As all.

Yeah.

Dan: Although,

Steve: It just gets harder and harder
to get a personalized on a computer

cuz there's so many components
that you have to do on your own.

Dan: Absolutely.

Although there was promising news
this week actually from Intel.

I dunno, did you see that?

Intel's first silicon spin Quantum qubit
device released to the research community.

Steve: Oh, I didn't see that.

Actually.

Dan: Oh, of course you were on holiday.

Yeah.

I forgive You then.

I forgive you.

You should look up Intel Tunnel Falls.

It's, they've released basically
a silicon qubit device.

And I guess that's one step towards
standardization when silicon vendors

or semiconductor manufacturers are
developing something that can be

used by different organizations

to, to then make the computers right.

Steve: Yeah.

Yeah, it makes sense.

Yeah.

Having a one chip that everyone's
using, that'll make things a lot

easier, especially standardizing things.

Yeah.

Simplifies the process.

Dan: Its only one step
in a very long, so It.

Steve: Yeah.

But that's why I think at least for
the next, I'd say 10 years, people will

just access computers to the cloud.

Dan: Yeah, thank you for bringing
it back to the topic of the pod.

Yeah, the way The, cloud the
cloud provides agility and

global access to resources
instantly and that kind of thing.

And that's exactly what is
needed in the quantum world, cuz.

It's very expensive to run the
computer, the quantum computer operated.

Whereas The hyperscalers, they have
these cloud platforms which are already

extremely mature when it comes to managing
platforms and infrastructure as a service.

So they're perfectly set up to
act as a kind of go between.

I.

Some quantum computer vendors may
have their own platform as well.

I dunno whether there's a layering
of platforms there, but in your

experience, what's, what does
the market look like there?

I, it seems that, the top three
hyperscalers are offering some

quantum computing services.

Some type with different quantum
computing vendors, but does that mean

they have bought the quantum computers
and they're running them in their data

center, or are they just acting as a.

Steve: I think with the exception of
ibm, the remainder of the companies

offering that service are using it as
a, they're acting as the middleman.

So they collect the information
for running quantum algorithms,

so the program instructions and
the metadata around the program.

Then they send that off to whoever's
quantum computer needs to run.

And I think each company who's hosting
their quantum computer on those

platforms, they probably have those
onsite in their own in their own labs.

So they're just acting as, they're
basically taking their platforms

like, you know, AWS who has
this whole infrastructure for.

Serverless and all those kind of like
micro services that they perform.

And I think they just take that
infrastructure and they just point

it at the quantum computer in a
very simplistic explanation, but

with much more detail of course.

But in a way I think
that's what they're doing.

They're using their infrastructure
to just recycle and then apply

it to the quantum computers.

I think same with Microsoft.

But with ibm they have their
own quantum computers and I.

I think they do it a
little bit differently.

I think they've really started from
scratch a way that they, yeah, they

set up their whole infrastructure
bottom up, and it's all designed

for quantum computing at least.

Okay.

Just in that particular
offering that they have.

Dan: It's not consistent
at the moment, right?

There's lots of different
options out there.

Perhaps we can just explore a
few That you are familiar with.

But yeah, I've seen different
quantum computing vendors available

on multiple cloud platforms.

And it makes sense that's
the way they do it.

And I suspect most of the quantum
computer providers have their own SDKs.

So Qiskit obviously is the IBM one.

The circ is related to Google somehow,
but then it seems most of the services are

offered through some kind of Python front
end, like a Jupiter notebook or something.

With the STK enabled, I guess that's the
quickest way to get access to running.

If you run an algorithm you define it.

You write it in Python or something
with the libraries from, or SDKs from a

particular platform you can then simulate
that locally and test it runs and so on.

And then you can put that through a ju
to notebook to a third party, right?

That's the way I've seen it done.

But is there more of a
programmatic approach to that?

Cause obviously The notebook approach
is great for developing and testing

around and playing with things.

But if you want to do something a
bit more serious and programmatic,

there must be APIs that can be used
through the SDK in a piece of code

that you write locally for a program.

If you want to write a program that is
predominantly classical, but it needs

to run Something that performs better
on a quantum computer, say a search

algorithm across some unstructured
data then what does that look like?

What's the optimal way of doing that?

Which, what other options have you got?

Steve: I think the best offering
in terms of APIs right now from my

experience is the IBM one because
they offer, a way to access their

computers through the Python scripts.

So you don't need Jupyter Notebooks.

So most of the time you have to go
on the particular platform using the

Jupyter Notebook like you said, and
then you write your code in there and

then you're already on their servers.

And then you can run an interface
with the quantum computers that way.

But with the IBM version,
you get your API key.

You just run some commands to
get the connection established.

And then you can write quite complex
programs where you use the quantum

computers within your programs.

So you can do a lot of the
classical parts, and then you send

off a job to the quantum part.

You wait for the job to return, and then
you can continue with your classical

parts and you can write quite complex.

Software in that way.

And it's nice that they offer the a p
i that you don't have to always connect

to the the Jupyter Notebooks and stuff.

But there are some pitfalls of that, that
I've experienced myself, some of them that

you need a stable inter net connection,
especially if you don't have direct

access to the quantum computers though,
if you have to queue up your job, for

example, and you need the output of the.

The job to continue your algorithm, then
sometimes it could be a bit challenging.

The outputs are saved on like
J files on the web interface.

But if you are queuing and you want
to just take the measurement results

and apply them into your classical
algorithm and you lose internet

connection, then you have to restart
and you might have to queue up again.

So you lose those results.

The job is in a queue, but you don't have
connection to the Python script anymore.

And sometimes these things
can take 10 hours to run.

They take quite some time.

And if you disconnect in those 10 hours,
you might need to start from scratch.

So it's a nice program, but if you're
not in a stable internet connection,

you could have some issues, even
though it works really well in theory.

That could be the benefit of using
the Jupyter Notebooks is you just

connect to the Jupyter Notebook and the
connections are probably much more stable.

They're not over the internet, they're
probably just over a local network.

So as the pros and cons of that,

Dan: Yeah, definitely.

Yeah.

So what makes the wait so long?

Is it the queue?

Typical?

Quickly that there's a high
demand for people running

need, needing to schedule jobs.

Because I would think and here comes
a stupid question, I would think

most quantum algorithms would run
very quickly, like milliseconds.

But are there quantum algorithms which
have been developed, which are complex

enough that it takes hours to run?

Steve: So the algorithm itself to execute
one shot of the algorithm, probably, yeah.

Like you said, it takes like
maybe millisecond millisecond

range, maybe even less.

So you have to repeat that many times.

So that's already.

A hundred thousand times, let's say,
before you get some meaningful statistics.

Okay.

Then, but even while you're running
the algorithms, there's things

that happen behind the scenes.

So I think they often run like
calibration on the hardware they have

some that pause your job internally,
or you might have to, you run your

job once, you need to run something
else and you need to go back in line.

There's a lot of things that are
not just executing the algorithm

that come into play with the timing.

I think also involved is conversion
of the algorithm to machine

instruction and optimization
processes that happen before running.

So there's like a lot of
pre-processing, post-processing.

I think the shortest part is probably
running on the quantum computer.

But yeah, there's a lot of things
that surround that little part

that take a lot of time as well.

So it's

Dan: That's fascinating.

And in your experience, do you get.

Any visibility of that in any other
platforms other than seeing that

jobs got a very long delay or it's.

Steve: Yeah, I think I would guess
a lot of people don't say what

they're doing behind the scenes.

Like IBM has some explanation,
but it doesn't say exactly

what's the status of your job.

It could have changed
in the last two years.

I haven't used it in
about, yeah, two years.

Maybe it's different now, but from my
experience it's kinda like queued done.

Mistake, something like that.

And then you can see the last maintenance
schedule on each quantum computer.

You could see the status of the
quantum computer is up or down,

how many jobs are queued already.

So you can have an idea, like if you want
to use that particular quantum computer,

how long before you get access to it.

But there's also different
ways of queuing that they have.

So if you pay for the service,
you can skip the queue sometimes.

You might even get a designated slot,
you can actually reserve the quantum

computer for some block of time.

In that sense, no one can access
the quantum computer except for you.

Of course, that's not to come for free,
so you have to be a partner of the

IBM Quantum Network so in general,
it's, I'm actually not exactly sure

why it takes so long, but just my
assumptions that there's a lot of things

happening behind the

Dan: Could be many reasons.

Yeah.

Steve: For basic algorithms,
it took like hours.

Dan: Okay.

And in terms of programming
languages, now I've mentioned

Python cause it's just everywhere.

It's ubiquitous and used in so many diff
on so many different hardware platforms.

I assume that the programmer
language we use as long as there's

an stk, it doesn't really matter.

As long as it can convert the,
in fact, let's talk about that.

What is it that the programming
language needs to do with an

sdk and what does the SDK do?

I take it, it's if the programming
language can describe your quantum circuit

in terms of gates, which is the way you
describe the algorithm in, a hardware

abstraction, on what each qubit is doing.

But then that needs to be transcoded
somehow it needs to be converted

into different formats, so do
you know about the STKs enough

to talk about them in that way?

Steve: A little bit.

Yeah.

I think, like you said, it's, it
doesn't really matter what programming

language you use because it's, at
the moment, programming quantum

computers is at the gate level.

So you're just writing machine
instruction piece by piece . Yeah.

So if when you write these machine
instructions, they probably get converted

into this language called Chasm.

And Chasm just is a kind of a
unified way of sending instructions

to the quantum computers.

So is it not Python anymore?

It's like very close to assembly level.

Doesn't really matter.

Like which language you write the
gates in and then you send it to Chasm.

And Chasm is the language that
probably goes to the computers anyway.

The only thing that's a benefit
of the SDKs is how they perform

optimization of the gates.

So you write your gate
instructions and they might not

be the optimal set of gates.

So each SDK has it's
optimization software.

So IBM has some, I think continuum also
has some software regarding that, and

each one performs a little different.

They could make your gates a lot shorter,
and then you get better performance when

you execute on the quantum computers.

So that's, I think that's the main
advantage, the optimization parts.

But after that, it's, it doesn't matter.

Once you have the circuit, if it's per
optimal, if you write it in Python,

write it in, see it's gonna be the
same thing at the end of the day.

You'll just go down the chasm level.

Yeah.

Dan: Okay.

And the outputs tend to just
be probabilities, right?

Is that right?

Your code abstracts what the outcomes are
to whatever it is the algorithms doing.

But is there a standard way that
the outputs are spat out from.

All the computers, different computers.

Steve: That also I think is a.

SDK specific thing.

So from my experience, I've
used a bit of the QC ware Forge,

and I've used the IBM platform.

Both of them have, okay, this is two years
ago, so it could be changed, I don't know.

But when I was working with those
platforms, the outputs were different

and you had to program what to do with
the outputs a little bit differently.

You probably do your own post processing.

These days it might be different, but
like I said, it's it's been a while,

but IBM tends to have a cleaner output.

They have this, adjacent output
and you just use that adjacent

output and you extract different
properties of the output.

So you have the measurement
results and the statistics.

You have things like the runtime,
how long it took all kinds

of details about the output.

So it has very detailed
outputs in with those.

You can do things like your own
error mitigation, for example.

You can try other things.

So you get a lot of freedom
with with the IBM outputs.

I don't know if every platform offers
that level of detail, but they should.

They don't.

I think it's an important thing to have.

Yeah.

But the outputs are, yeah, there's no
standard for the outputs, but generally

you at least get the output state
in the computational basis with the

number of times that state occurred.

So if you run your algorithm a thousand
times, It'll tell you, you've got

state one 10 times state two 500
times, et cetera, until you get to

the total number of shots and then you
have your statistics and you can do

the post-processing however you need.

That would be the minimum.

I think the minimum you get is that, and
then the rest is nice things to have.

Dan: It's, you know what I'm getting
from this is it's still very low level.

Both the inputs on the outputs And.

It makes me think, what, are the
decisions that need to be made or the

things that need to be considered in
order to decide if something warrants

using a quantum cloud service?

I guess that's the same question as
warrants using a quantum computer.

Does it have to be something that
can be optimized using some of

the main algorithms like Dot Joe?

I mentioned Grover those kind of things.

And then how would you know, or how
would you test that it's worthwhile?

I guess that's what simulation is for.

Would you just personally,
would you simulate it locally?

Would you simulate it in the cloud?

Would how do you go about that?

Steve: There's a trade off point, I think
when you can, when you probably wanna

run it in the cloud versus locally, and
that just depends on the system size.

Because they offer things like G P U
accelerated quantum computing, and if

you don't have that, then the cloud
really offers a meaningful service there.

So you run simulation of 20, 30 qubits and
you don't, you don't have to wait so long.

Maybe your, the computers we
have, don't, our laptop probably

can't run a 30 cubit simulation.

They'll probably have no option
but to send it to the cloud.

There's one thing I would say is
a warning though, is sometimes.

When you send your algorithms to the
IBM Cloud or the n a cloud, you're

basically telling them exactly what you
wanna execute and you are giving them

a lot of information about what you
want to, what your algorithm is doing.

So there you might think about what to
run locally first, in case you don't

want to give away your algorithm.

Maybe you have a new algorithm for doing
something important, and then you wanna

simulate in one of these cloud platforms.

When you upload that algorithm
to their platform, you tell

them exactly what you've done.

And maybe they, it's clear to them,
but they can read that, right?

It's not private.

So that's a thing where people
talking about blind quantum computing.

So you don't have to give away the
algorithm, that's a little different.

You need some kind of
quantum communication.

There's an argument to make.

Are you willing to give away your
algorithm and then yes or no, it dictates

if you wanna run in the cloud or not.

But yeah, generally it's
hard to run anything locally.

You need power is the
hard things to compute.

Dan: Blind quantum computing.

That sounds like a great
topic for a future pod.

So just focusing on that
same question again.

So it really just comes down to compute.

And would you recommend simulating
always many times in advance to get

a feel for whether your algorithm is
worthy of the quantum computer or.

Steve: Yeah, I think it's necessary,
at least in some way, at least

a small scale, to know that the
algorithm is gonna work because.

You are consuming so many resources
by sending to the quantum computer.

It might take five hours to get something
that takes half a millisecond on your

computer, so you're waiting magnitudes
longer for very simple results.

So you always want to simulate locally
first because it's not worth not doing.

You wait way too long and you might be
paying a lot of money as well to do it.

So I think at the moment it's
probably not even worth sending

to the quantum computers anyway.

I think the use of the today's
quantum computers is, it's not

good enough to validate the
price of using one at the moment.

It's good for testing,
it's good for experiment.

If you have the access, you
know that it's gonna work.

The systems all work.

Your APIs work, but, Using
those results for something

important, I would say you don't.

You don't get that yet.

Definitely not.

So simulation will tell you that
it's gonna work in the future.

When the quantum computers are good,
sending to the coding, computers tell you

that your code is configured correctly,
but the results, they don't overlap.

One is meaningful results.

The simulation results one
will give you garbage results.

But you know that your
programs work in practice.

So that's the trade off, I think.

Dan: It's a bit hit and miss.

From what I've heard is it's the
optimization problems are the ones where

there's been more of a successful outcome.

Steve: I think actually, what was the last
week, IBM had that big announcement too.

They ran the I have to read it myself.

I didn't get a chance yet,
but was about a simulation.

On their code computers, I
outperform the classical computer.

So those kind of things are important.

Those things aren't known to
outperform classical algorithms,

practical use cases of them.

I have to read up more.

I, but I'm thinking that
there might not be one yet.

At this early stage, doing anything
better than classical computer on a

quantum computer is a breakthrough.

So having the practical part
of that is not necessary yet.

It's about building meaningful computers.

Dan: Yeah, you've done it again.

You've introduced me to something
that I know nothing about.

Boon sampling, restricted model of
non universal quantum computation,

introduced by Scott Sson and Alex ov.

I dunno what that is.

That gives me some.

Thing to go away and read about.

Thanks.

Steve: I don't know much about it
either, but I know that it's what

they execute on the Google, like
the Google supremacy experiments and

also there's been a few iterations
of people reproducing that experiment

.
Dan: So where do you think the cloud
services approach for providing

access to quantum computing is going?

I think it's clear that it's gonna.

Be the main method for a long time.

But do you think it could move away
from the hyperscalers or do you

think the hyperscalers will go all
in and certainly we see Microsoft

working on their own quantum computer.

Same with Google developing their
own chips, a series of chips.

Do you think it's just gonna
keep going in that same

direction or you know what other.

Influences on the market are
there that could affect that.

Steve: I think until they have
something like room temperature,

qubits, then it has to be like that.

I don't see how the average small
scale businesses who wanna install a

quantum computer for their own company.

Even in that sense, it's too hard.

I think it's maybe it's too expensive,
it's too specialist for a company

who's not making millions in
profits and they can't afford that.

So I think until there's room temperature
qubits, it has to be cloud-based.

That's my opinion.

And I think the people who are gonna do
it are just gonna be the companies who

already have the cloud platforms in place.

Like Microsoft, Amazon, ibm, to
be honest, I'm surprised IBM has

done it so well because I think what
Amazon has most of the profits simply

from the cloud services and it's
already a huge company and they just

really dominate in cloud services.

Compared to starting from scratch.

Yeah, my opinion is it's gonna
be the hyperscalers, the biggest

companies around that will continue
to offer the cloud services.

The smaller companies or building quantum
computers will be under their umbrella.

I think it'll continue like that
until, room temperature qubits

come out that we don't need to
store these quantum computers in

such a specialized environment.

Dan: Yeah.

Thanks.

I, IBM's an interesting one.

You should definitely try and get
somebody from IBM to come talk to us.

The they've got a history in mainframe.

And super computing.

They do have their own Cloud,
There's IBM Cloud, and of

course they also own Red Hat.

So they're very embedded into massive
applications like Oracle and sap.

So they've got an interesting
place in the market, slight quite

unique compared to the other
hub scalers and cloud providers.

And yeah, they've definitely
gone all in on Quantum.

The community around ibm
quantum is pretty good.

They're pumping out a lot of training
materials and labs and really trying

to stimulate the technical community,
which I think is quite honorable.

But cuz it's quite costly
to that kinda thing.

Steve: Yep, definitely.

These things are giving most
of it away for free too.

So all these high quality resources,
videos, lectures, textbooks, software

APIs, this stuff definitely can't come
cheap, but I think they have managed

to at least offset some of that with
their IBM M Quantum Network, that

it's quite expensive for companies
to access the quantum computers.

I don't know if they're cutting a profit
off that, but definitely offsetting

some of those costs for I can imagine.

Dan: I'm sure, and it's definitely
a long game for them, right?

For anybody in Quantum really.

So let's think about real
world applications for a minute.

In terms of a developer wanting to, to
use quantum powered algorithm in his

or her code to solve difficult problems
MP problems, then, are we looking at

we often hear about financial modeling,
drug discovery, those kind of things.

What is it about the cloud services
that leans it towards that,

that those kind of use cases?

Is it just that they're the kind
of use cases with the difficult

challenges at the moment and the
cloud provides the flexibility,

the scalability elastic services

if you are running something
on a single quantum computer

via the cloud, you can read.

You could quite easily use multiple
ones, which by changing your

code and maybe paying a bit more.

Whereas if you're doing that in
your own environment, obviously

that would be extremely difficult.

Oh, let's just build
another quantum computer.

It's not gonna be that simple.

But yeah.

Yeah.

In terms of real world use
cases, what are your thoughts?

Steve: So to me, it's always a
tricky thing to answer what's gonna

be the use case for quantum peers?

I know that the chemistry is there.

Quantum simulation is there.

Solving classical algorithms with quantum
peers, I think is very challenging at

the moment because the data loading
problems where you have to load

classical data to the quantum computer.

So with the cloud quantum computers,
especially in the chemistry, I think what

you're gonna get more out of is probably
the size of the quantum computers.

Maybe we do have room temperature qubits,
but do we, does that mean we have a

hundred thousand of them, or can we still
execute those in a room temperature?

So maybe if we have a hundred
thousand qubits, maybe those are also

distributed across multiple processors.

So just having that like ease of access,
I think that's the benefit, especially

if you need something large scale.

So if you have a chemistry problem,
you need a thousand qubits, maybe

they don't sell that kind of computer
that you can buy in the near future.

It's always gonna be like one step ahead.

Cloud quantum computers always be
bigger, more of them faster, like

something's gonna be better than the
ones you can buy for your own business.

It's kinda like super computing, right?

Like supercomputing clusters.

Universities have tons of these things,
but the average person doesn't have,

30 towers, 30 racks of computers.

And so it's kinda like that.

There's always someone who will
need more power, bigger, faster.

And there'll be some people who
just need what they need at home,

or they need for their own use,
like a laptop power equivalent.

Dan: It's one of the great things that
the cloud has brought right, is to

democratize access to resources like that.

It's good for both sides.

It's good for the developer who wants
to access those kind of resources

easily at low, a relatively low cost.

It's also good for the owner
of the compute platform or

whatever the service is.

To monetize it globally quite
easily using the platforms that are.

Steve: I think that's how it has to be
somewhere now because it's so difficult.

Even the problem that already exists
classically too, the high performance

stuff is already cloud-based
mainframes and those kind of things.

I think there'll be a lot of
analogies between the two.

A lot of analogies between
high performance computing

and quantum computing.

At least until there's some new
breakthrough and I think it's might

come, like I believe there's gonna
be a big breakthrough in cubit

technologies in the next 10 years.

I don't know if superconducting
qubits will survive, but at

least they're working now.

The technology's still getting
better with them too, but I think

there'll be some breakthrough coming.

It's The superior qubit that
just takes over everything.

Dan: Yeah, I mean certainly there's
a lot of investment and hype in some

cases, but that means there's a lot of
investment and interest in the field.

It's definitely a bit of a
golden era for Quantum, I think.

It's all going in the right direction
for something like that to happen.

You're right about high
performance computing.

That's ultimately what It's right.

It's a niche in.

High performance computing to solve
problems which are uniqely difficult

to compute using classical computing.

So in a way it's it's a slice of
the HPC market what I see anyway.

But is it gonna grow beyond
that, I guess potentially.

But it's so far away.

We just my crystal ball doesn't
see that far into the future.

Steve: Yeah.

Yeah.

I think in the meantime, bringing it
back to the networks to talk about,

like how to communicate with those
computers from a, non-local environment.

And I think until they make a
way to make things private it's

gonna be tricky to adopt as well.

For most people, they don't wanna
give away their proprietary data.

They don't wanna send the data to
the high performance computers.

No one wants to send their private
information to the companies who are

selling that as a product, especially
when they're directly competing.

So I think until the communication
protocols somehow mask those

things, it might also be, that
might also be the bottleneck.

It could be that no one uses
them, even if they're there, cuz

it's just not private enough.

Dan: Yeah, no, that's fascinating to know.

So even though the technology is there
to democratize and make it available

everywhere, it doesn't become appealing
because of the lack of privacy.

Now let's just touch a little bit on.

What solutions there
could be to solve that.

So you mentioned before the quantum
communication being necessary to somehow

hide the algorithm that's running,
you called it blind quantum computing.

I think, at a high level what does
the architecture look like there?

Would the cloud provider,
quantum computer provider.

We'd need to accept some kind of
fiber, point to point connection

from a customer, and then accept the
requests encoded in qubits somehow.

This could get extremely
deep very quickly.

I know, but is that ultimately
what you think would happen?

What's your initial thought?

Steve: Yeah.

So in the blind quantum computing,
but from what I can recall it's

about a client who has very
simplistic operations for quantum.

Like they can prepare quantum states
and measure quantum states, but maybe

they can't manipulate the quantum
states in between those two steps.

And they send a quantum
states to a cloud provider who

performs the algorithmic parts.

And then sends the quantum state
back or makes a measurement on their

side that they don't understand.

Then they send the measurement results
back, and then the person who sent

the quantum states, they know what
to do with the measurement results.

They can basically decode them in
a sense, and that way the provider

doesn't know anything that happened.

They just perform the execution.

They don't know what the algorithm is.

They don't know what the
measure results mean.

So what that means is they have to have
some kind of point-to-point communication

with the client and it's not completely
contained in a local environment.

So there are some non-local operations
happening that would make things quite

complicated cuz you now you need quantum
internet or something, or quantum

networks and that's a big step to take
for something like quantum computing.

So firstly, we need to know what
to do with quantum computers.

Then we need to put in the quantum
network and then it's safe to

use for the average person.

But yeah, that's a long
road ahead, I think.

Dan: Yeah, I know these discussions
are extremely futuristic.

We're reaching right almost
to sci-fi level here.

But Yeah it's an extension of the
hardware manifestation of your algorithm

across the network to the third party.

How would they not know what
the algorithm looked like?

. They'd have to be measuring
each of the interactions and

therefore that would break down.

When you're implementing an algorithm
on a chip, is there a shadow of what

the algorithm looked like in any way?

I guess that all comes down to you
can't measure it, but it may come

down to the hardware manufacturer.

It leaves anything behind.

Steve: Something I have to read up on
again, but it's based on this concept

like ZedX calculus, and I think it's
basically the circuitry is fixed.

For the quantum algorithm
and what is changing is the

angles of the manipulation.

So you're not sending a series
of gates, but a series of angles

and the gates are already fixed.

So it might be like gate one apply, pi
over two, gate two apply pi over three.

And somehow the combination of these
series of gates implements the algorithm

and the, but the angles themselves
mean essentially nothing cuz you

don't know what the input state is.

And the measurement result is
based on the input state and

all the operations combined.

So the output probably is meaningless
as well, unless you have statistics.

So when we come back to this topic,
I'll make sure I understand it better,

but that's my suspicion right now.

Dan: that's fine.

I I think that means the answer to
probably no, and that makes sense

because ultimately the position of
the qubit at any point in time it's

very transitive, it's gonna disappear
very quickly and there are many in a

particular state and all interacting
with each other in a, for a very short

period of time is the way I see it.

So there's no way to really capture that.

Okay.

There's a lot of unknowns
here, aren't there?

There's a lot of stuff based in
research and , unless you're deep in

it every day, then, so understand that
you're not gonna have it all to hand.

Is there anything else you wanna add on
around the tools and resources provided

by cloud providers or cloud-based
quantum computing services that would

be interesting to discuss briefly.

Steve: One thing that's I
found interesting recently

is the execution pipelines.

You can construct using
these software tools.

But this is something I'm also
not so familiar with, but I've

been reading a little bit.

Basically they call it quantum DevOps.

You can set up complex infrastructure
to execute quantum algorithms and I

don't know, in a complex way, it's
not just sending your algorithm

and getting the results back.

You can already put things
in place to preimpose process

and also resource management.

This is a topic that is emerging.

It's like a new idea where you're seeing
people talk, talking about quantum DevOps,

and that's gonna be a way to manage the
complexity of the quantum algorithms and

the different structure they have to in
comparison to the classical counterparts.

Dan: So to use the word DevOps,
it's got to leverage something from.

Traditional DevOps world, whether it
be the pipeline management of changes

and queue schedule changes and so on.

Testing, automated testing perhaps.

And also is there any benefit sometimes
in spinning up a Temporary infrastructure.

Cause I wonder whether there's an
overlap with infrastructure as code

and spinning something up that needs
to be done classically in cloud because

you don't have the compute locally.

But then that is also reaching,
I guess that's an extension of

what we're already talking about.

As if you are running a a classical
algorithm locally and you're

using a remote quantum computing
service, you could then put

a classical bit in the cloud.

And define it all in essentially
define it in code so it can be

stood up and executed automatically.

Steve: Yeah, I think it
could simplify a lot.

If you wanna write an API that
has quantum components and you

take out inputs and the whole
input could be basically structured

like the classical way, and then.

They put in environments when
they're needed, take them down

when they're not needed to save
resources on the quantum, free up

your reservations or make reservations
when you know the load is coming up.

I think that's what's coming.

It's gonna be, it is.

I think it's no different
than classical DevOps.

. I think all the ideas are
gonna be exactly the same.

It's just a quantum
computing component to that.

Dan: Yeah.

Fascinating.

That'd be interesting to to
discuss further at some point.

Okay.

Should we wrap it up for today?

Steve: Yeah.

I think we, we discussed a lot.

It's

Dan: Yeah, lots of lots of thoughts going
on, pinging off in my head at the moment.

Need to go and lie down in
a dark room again for a few

minutes to try and calm down.

Steve: Yeah,

Dan: Good to talk

to you, Steve.

Steve: it's just Monday.

Still,

Dan: very much.

Yeah, I know.

Steve: you're gonna need your.

Dan: Good start to the week.

Yeah.

I'd like to take this moment to thank
you for listening to the podcast.

Quantum networking is such a broad domain
especially considering the breadth of

quantum physics and quantum computing all
as an undercurrent easily to get sucked

into So much is still in the research
realm which can make it really tough for

a curious it guy to know where to start.

So hit subscribe or follow me on your
podcast platform and I'll do my best

to bring you more prevalent topics
in the world of quantum networking.

Spread the word.

It would really help us out.