Jun 10, 2020
________________________________________
TRANSCRIPT For this episode of The AI in Education Podcast
Series: 3
Episode: 3
This transcript was auto-generated. If you spot any important errors, do feel free to email the podcast hosts for corrections.
Hi, welcome to the AI and education podcast. Hi Lee, how are
you?
I'm good, Dan. Had a great long weekend. Looking forward to getting
started.
Fantastic. Now, today we're going to look at something which is
really, really cool and really cutting edge. Quantum computing. I
can't wait. This is going to be so exciting.
Yeah, this is a big topic. Uh hopefully we can Is it in the time we
have, mate?
I know. Yeah, exactly. Well, like as a as a computer scientist, I
suppose, and and an ex-teer of of computer uh technologies, I used
to do it quite a lot around binary um and binary and classical
computing and and I think it's really I've had so many
conversations with people around quantum and it just just the
general conversations blow people's minds and I think hopefully we
can break apart some of those um uh detailed kind of uh background
to do with quantum today but also kind of bust some of the myths as
well because it does it does take you to a different zone mentally
when you start thinking about it. So when we when we talk about uh
modernday computers and and classical computing we're just talking
about binary machines right?
Yeah. Yeah. Look largely and and it's an it's a good place to start
is that kind of what we all understand today because you know and I
do a few talks on quantum and I'm by no means a quantum expert but
you start to get you're trying to explain quantum and you get to a
point where suddenly your mind sort of fries a little bit and you
and your brain starts to melt and you go well hold on a second but
that doesn't make sense and there's a piece of that which is
because frankly quantum sort of doesn't make sense in the physical
world we live in um but it's good to start with that binary
background I think you're right you know that's a place to kind of
ground our knowledge
obviously I used to actually walk up to a light switch and just say
look it's on or it's off and there's so many of these switches
inside uh a computer and you know through Mo's law and the amount
of um technology we've created to make more and more intricate um
circuits and processes and the way the really smart people have
architected a lot of these uh pieces of hardware over time that
we've seen a lot of the um uh moves towards like the expansion of
uh the use of these technologies and actually all built on a simple
premise that it's just ones and zeros. Um so So that that's where
we started, right? And that's where lots of computers today are
still running. My machine now is running in a classical system with
lots of smarts.
Yeah. Look, and that's that's a really good point that I guess, you
know, not everybody kind of acknowledges is the computers that that
you know, I was born in the 70s. I grew up with computers and the
computers that we started using, the home computers of the early
days, all built around that initial original x86 architecture, the
80886s from Intel,
uh, you know, that that's still fundamentally the same architecture
of the computer that you and I are using today to record this
podcast. It's still that construct of a binary based computer and
it you know if you kind of start to think about it and you ask
yourself well why binary why was that the case back in the you know
the sort of 60s when we started getting to this point and it really
comes down to because it doesn't make any sense you know binary
maths is not a way we think logically think you know we would think
normally in tens because of our you know decimals because of our 10
fingers that's how we arrived at that point
yeah
but but if we think back to you know binary at the time. Why were
computers built with binary systems? It was really down to
basically the cost of it was so expensive to build silicon based
you know chips at that point. Uh physics you know we in order to
measure a value in a computer that one or zero that you talk about
we have to measure an electrical signal ultimately that's what it
boils down to on the chip. There's a transistor or something that's
off off or on. And it just at that point in time physics led us to
build these things that could really only be measured in extreme.
yes or no on or off and so we needed that very simple design so you
know that's the principle why binary made sense but
it has it has become the foundation for computing zeros and ones
and I remember programming if you might remember Dan in
assembler
where you think about
loading data onto the stack pushing and poking or pu or pushing
onto the stack and then reading data off the stack and the
registers and the registers were essentially
values that you read zeros and ones and that tell the computer
that was so stressful oh god this brings back the chill To me it
does
but I suppose that pushed pushed a lot of limitations now and the
scale of the issues based on those computers. What were the kind of
main issues bring to a play?
That's that's exactly right and that's that's and I think you
mentioned it earlier the you know uh Moors law Gordon Moore from
Intel's perspective as as the market grew is that there was this
sort of continual growth. So every two years the number of
transistors on a on a silicon chip on a on a board would double.
and and this idea was actually that you we were continually scaling
out computational capability. So every time we thought about the
next generation of chips um classic physics would come into into
play and we would say okay because if you think about what's on a
silicon chip from a computer it's just initially thousands of and
now millions of and now billions of transistors and each of those
transistors can make can hold a state of zero and one and that's
binary computing and and so Moore's law was essentially you know
when we're building replicating these chips that say I don't
remember what they started at maybe you know um I don't they were
like 15 nometers or whatever they were when we first started
there's so many chips you can fit onto a particular piece of
silicon but as Gordon's law more Mo's law states that that as you
get smaller and smaller in that scale you continue to build out
more of them but we eventually hit this problem which is we move
out of the physical domain where you can physically build a
transistor down to seven you know nanomicron sizes onto these chips
where we start getting almost into the sub subatomic particle level
where we're actually now building things so small they are
constitutionally the size of an atomic particle as opposed to a
physical thing that we can measure a state on.
Yeah.
And that's that's what Mo's law has done. And you know I think now
we talk about Mo's law reaching its point in 2025 of of completion
where we've hit the point where law no longer applies. But that's
to your point that's fundamentally how our computers have grown and
scaled over the years. And what's happened is is if you understand
the way computers work is that we've shrunk the number of
transistors onto the chip, but we've increased what we call the
clock speed.
So, you know, your computers of old used to run at maybe five mehz.
My old Commodore 64 used to run at 3.5 MHz,
you know. So,
megahertz is the uh you know, the amount of cycles you can do as
well, right? And how many instructions you can process?
That's right. But it's Yes. So, it's exactly that the you know in
every second how many times with the crystal chip that's running
manages that process. How many times will it cycle through the bit
reading of the chip?
So, you've got this kind of faster cycle speed. So, we go from, you
know, 3.5 megahertz chips of the ' 70s to where we are today where
we're talking about 5.5 gigahertz chips reading 5.5 billion cycles
in a single second. Phenomenal. Yeah.
Yeah. And and compress that with a sort of a compression of the
number of transistors. This is the exponential growth of Mo's law.
We can do more
in a smaller space with more transistors. But I suppose what what
we do try to do and what people have done is then try to side track
from the Mosow element and then add GPUs in. So often with the
games units that we creating like Xbox and things like that when
people are doing film editing, you know, they they then bring in
specific graphics processing units to try to get around the fact
that the CPU is kind of frying. So they push a lot of the work out
to dedicated processing units, I suppose.
Yeah. Which is a change of architecture and that's that real moment
where we move away from say your traditional x86 compute chips the
the sort of the traditional CPU chip which although we're doing
more and more tasks with every clock cycle we're still doing it in
a zero process it's still micros secondsonds apart but essentially
the computer calculates something and then calculates the next
thing and then calculates the next thing and it's doing it one
after the other you don't have them have things happen at the same
time you know truest sense of it
so yeah so so PUS came along and and sort of moved into different
architectures that can do computation in a different way but it's
still computation in the physical sense. It's still measuring bits
ones and zero
and is that and that's impacting on the growth of some of the
technologies now that we talking about in this podcast like AI for
example, right?
Yeah. Look, and that's important that we bring this back to that AI
story as we want to is that you know a as AI has grown and we've
talked about this in other episodes the growth of AI as a as a
mainstream approach to technology and and you know problem solving.
It quickly under it quickly led us to recognize the constraints of
of the CPU in the x86 architecture. Some may argue that the
acceleration of mos law and the and the end of mos law has been
somewhat driven by this different differentiated desire to do
different types of compute which GPUs FPGAs you know build programs
enable us to do because it's and and there's a great video we
should put it in the notes um because bit of a mythbusters spam,
but it really visualizes this problem of when you give a CPU a task
to do and CPU say in the example, paint a face and a CPU can go
really really quickly with every single flick of the clock speed
and paint every single dot of the picture. But when you give a GPU
that problem, which is designed to kind of break a problem down
into multiple tasks and then do all those tasks at once, instead of
telling it to paint all the little dots, you just say to it, "Paint
these 6,000 dots now." And it paints all the dots and see in the
video we'll put in the link but I think it's a really really
visceral way to see that difference between traditional computers
and the limitations they provide and how GPUs and when you think
about that in an AI sense the important the reason why that's
important for AI is because of course AI is all about taking data
and running experiments if you like or learning processes over that
data to learn pathways to learn out and that's where you need that
massive move away from serialization into kind of you
parallelization of the data process
and and that's So we now start to kind of break into the kind of
quantum realm, right? So So let's start to let's sort of dig into
this. Holds on everybody. Your minds are about to explode a little
bit. But but I suppose one one thing to um tackle from the outset
is obviously quantum. We're talking about very small particles at
the quantum level, right? So when I started to like understand
quantum theory and things like that, you know, a lot of the a lot
of the scientists and I'm sure we'll touch on that a minute. But
lots of the scientists came up with lots of thought uh problems
around this one being Schroinger as cat which we look at in in a
little while. And I think that's where sometimes people start to um
get a bit confused because you start to think about quantum in
terms of you know our current environment you know our size can we
be in two places at the same time and all that kind of stuff. But
from the outset quantum is very much talking about things that are
very very small particles that are real quantum level, right?
Yeah, absolutely. And you're right. This kind of this is that leap
into things that just don't marry up, don't make sense in the world
we live in today. To your point, when we get into those topics of,
you know, of superp position, for example, we'll talk about this
idea that, you know, that two things, a thing can be two things at
the same time. And it doesn't make sense. We can arguably say that,
you know, a piece of paper can be blank and then filled in and we
get that logically that makes sense to us that a piece of paper can
change its state but something acted upon it and it's never in both
those states at the same time and this is where quantum starts to
go
you know I don't make sense of that so let's break it down I think
it's important to kind of step through the bits because as you say
quantum is
quantum starts with quantum mechanics if we think about this in
terms of a how did humans start to do this it's observable world we
can we can observe quantum mechanics at play so you know when we
think of photons electrons. We look at the way light works. Um,
very early on, people like Michael Faraday, who we all know now
today when we think about the construct of the Faraday cage, an
early pioneer of the idea of quantum mechanics, the idea that there
was something happening in the universe at this sub atomic non kind
of non-physical sense. It was happening at a level below what we
perceived to be the physical world.
Yeah.
So, hundreds of years, well, Michael Faraday, I guess late 1800s,
so 100 plus years ago, these ideas were there, but nobody really
understood what this was or how it was happening.
Then you've got your friend Schroinger uh who a friend of Einstein
and you know these people the in this early 20th century 1920s
1930s sharing their stories um you know with Schroinger's cat as
I'm as I'm sure you're familiar with the story is one of those
great stories that we teach kids today and we learn about the idea
of of quantum from that story. What tell me tell me your version of
the story Dan.
Well it'll probably be very basic and we love lots of of physicists
writing in but but essentially um the way I've explained that
before is is you know Einstein and Schrodinger and lots of those
scientists of the day used to you know when somebody comes up with
a mathematical formula they try to create um realw world scenarios
to kind of explain that or push those things to the boundaries and
Schrodinger's um explanation of having multiple states and at
quantum level was when he he essentially had a a cat, you put it in
a box with a vial of poison and you close the box and then the
thought experiment was that that cat was alive and dead at the same
time and only when you opened the box and observe the cat would
then it take one of those states. Um so that was his kind of
thought um experiment to try to explain some of the quantum
theories that were happening and I suppose that also came from
uh some of the initial work with particles and light and the wave
particle duality elements where where we talking about um uh the
light traveling. I think it was really early into this where they
managed to transmit a particle of light through a piece of gold
foil and it landed the other side in two places at the same time
and it was very very um uh small but then they tried to extrapolate
that into a thought experiment which then pushed into Schrodinger
sea of this cat to make it feel more real world but when you talk
about it like you said at the beginning when you talk about
something real world like a cat being in two states it's very hard
for you to think well it can be but obviously that cat is made of
um billions and billions of particles that that you know you you it
becomes very complicated.
Yeah, it's I think it's a good it's a great example by the way and
great great uh great description in that it's a way that we try and
rationalize a construct that doesn't make sense to us in the world.
So by the idea of a cat in a box, we can all get our head around
the idea of a cat in a box. Everybody understands that principle
poison, you get that principle. But then when you say the cat's
both alive and dead, that's when you stop and you go, "Well, hold
on a second. No, because that doesn't make sense." So you it's a
it's a great way of thinking about it and and to your point, you
know, what what sort of happened, you know, this is say back in
1920s, 1930s where Shreddinger and Einstein were postulating on
these big ideas of this is how the universe is operating. But it
then was really nothing much happened. There's lots of things
obviously going on but it wasn't until the 70s I think and as
computing became more and more of a uh not mainstream but computer
became more realistic but we started to see this sort of evolution
of theory and discovery of quantum we had the technology to at
least build out theorems around well how does quantum work because
you got to understand at this point troing as cat's idea was
logical in sound principles of how quantum mechanics works but
nobody understood how it would do it they just
observed it as a thing that happened in universe.
Yes.
So you've got that 70s where people are sort of theorizing on it
and discovering you know the kind of the mechanics of quantum. The
80s are where we start to see a lot more of the principles and
mechanisms to measure quantum. And this is we'll come back to this
because this is one of those big fundamental challenges actually in
quantum computing is not the building of quantum cubits but
measurement of those cubits in order to derive some real measurable
um manageable outcome
without changing the state. Right? So that's the shorting thing. As
soon as you open the box then you change the state in fact and not
only change the state you collapse the state and it's a very again
it's the subtle difference is that you haven't just changed it
you've actually the state is gone and it's no longer a measurable
state
so you have these things called quantum gates where you're
you're no longer measuring what's actually state you're also
measure you're kind of measuring the what is the what is the
condition of the state if that makes sense and that's where I start
to go a bit huh but that's what you're trying to do you're not
measuring exactly what it is but you're measuring everything around
it to determine what the actual state of the thing is without
touching the thing to measure it so you that in the 80s kind of
getting these ideas of well it's a thing there we know it let's
figure out how to measure it then the 90s was a lot more
experimentation you know we're starting to build these algorithms
to to actually calculate on quantum levels because that's the other
thing to think about is that you know principle you know you
anything you can do on a classical computer you can theory do on a
quantum computer um and there's this construct of quantum supremacy
where you can actually do something on a quantum computer that you
can't do in a traditional economical way on a classical computer
But we had this sort of thought process that we need to build
mechanisms of asking the computer to do things that align to the
way quantum mathematics and quantum works. But it wasn't really
until the early well 2011 a Canadian company called D-Wave built
what is considered to be the first commercial computer 20 cubits in
this computer. Uh and we'll talk about cubits in a bit. Um and it
was built out as a computer. It was quantum. It was it had all the
characteristics of quantum but it was actually later proven to be
whilst it was a quantum computer. It didn't actually do anything
faster or better than a classical computer. It just used quantum
bits for its calculative model. So it's it's a first step, but
that's where it is. So if you you know when you kind of put it in
that sense, it was really only in the last 10 years have we kind of
really accelerated into this point where actually we're building
quantum computers. And there's what you might even call now a
general market for quantum computers. People are building them.
Companies are buil built around just building quantum computers
and that's the thing for me you know when you when you think about
when when I have that conversation around Schrodinger and the that
kind of entire premise about the the quantum bits and things being
you know in in multiple states then you know actually the thing
that kills people off when they say that can't be true say well
we're actually building these things now it actually does work at a
quantum level right
yeah yeah it does and it and they're happening and it's it was
funny I was when I was doing some research for for our had on this
one. I found out that actually in just recently 2020 uh University
of New South Wales here in Australia have developed what they call
a hot cubit. Now this will come into context when we talk more
about cubits but the idea of a hot cubit is a cubit that can exist
at 1.5 Kelvin. So for all of you all of you weather buffs out there
know that 1.5 Kelvin is just above absolute zero which is
equivalent to aboutus 273 degrees centigrade. Now the coldest place
on Earth Earth ever measured ever is 184 Kelvin and this thing is
running at 1.5 and they call it hot cubit. So it gives you a sense
of some of those physical challenges that the cubit itself gives us
when we try and create it.
Definitely. So so when we look at some of the architectures in a
minute because I know I'm I'm really interested in that and that's
where my knowledge starts to go out the window very quickly um with
the way that different companies are are kind of tackling this. But
let's go back to the point then that you mentioned about the cubits
then. So What what is a what is a cubit?
Right. Okay. And again to all the quantum physicists who listening
to this podcast, feel free to write in afterwards and tell me that
you completely missed the point. But I think what it's because
let's talk about it in the context of things that we can get our
head around because this is a domain that is really very specific,
very mathematical, very deeply versed in some of that stuff. But
there are some basic principles in it. So the cubit is the
functional equivalent to a bit in a binary computer. It's the
functional equivalent to that measurable state of a of a binary
particle. In this case, a binary quantum particle that we want to
measure it. It has two states. It is you know it can be read as off
or on. But the unique part of a cubit is of course that it is in
multiple states at the exact same time. Yeah.
So the the way that I I read this the other day and I think this is
a really good way to think about it because we all think about
flicking a coin. When you flick a coin in spinning in the air and
when it lands it will be with an absolute 100% certainty guaranteed
one of two states. There is a 50-50 chance it would be heads or
tails and any other sense of probability probability is not true.
It will probably be
50% yes, 50% no, 50% heads, 50% tails every single time you spin
it.
But until it lands, you don't know.
That's you that coin. Now, if you could watch that coin as it's
spinning, if you could observe it as it's spinning and see its
states, what you would actually see is two states, heads and tails
in constantly being displayed to you. Of course, I mean, we're
talking about, you know, an impossible coin spinning at an
impossible rate with an impossible ability to see it to observe it.
Of course, none of which we can physically do because that's the
separation from the physical world, the quantum world. But that's
the point is that that that's where so a cubit is is a bit that has
that is in multiple states in a continuum
and as long as you don't read it and we'll get to measurement, as
long as you don't try and measure what that state is, it is always
in
any a wave state. And I think you talked about this Earl on this
idea of a wave state. It's in every single state possible, all the
in between states and not just ones and zeros. Of course, that's
the beauty of of quantum bits is they are in variable states.
Wow. Yeah. So, that's that's a great analogy. That's a really
I like the spinning point one. I think it kind of gives you a
visceral sense of what that might be.
But what it does bring up and we talked about this earlier on and
this is just Schroing's cat story is kind of the two major things
that define quantum computing and make it make it actually
possible. These are the mechanics of quantums. that make computing
possible which is superp position. This idea that it is in multiple
states at a singular time and then entanglement and the idea that a
single bit or a particle or a quantum bit can be inextricably
entangled with another bit such that its state and the other one's
state are not are intertwined are uh it's not again this is a funny
one it's we often think about this entanglement as being well one
state is going to change the state of the other. So when you look
at it and measure it, it changes state and the other one changes
with it. It's not so much that. It's more that
and again I say these words without fully understanding the words
I'm saying, but it's that you can't describe the state of one
without actually describing the state of the other. They are one in
themselves.
They're not two different particles. They are the same particle
and then then you kind of go but that's the that's this idea. So
entanglement and superp position are the two mechanisms that uh
kind of enable quantum computing.
Got you. Okay. So, so we've got we we now know what a cubit is.
We've got an idea the superp position. So, correct me if I'm wrong
here. That's that's the different states at the same time. And then
entanglement means the way that they interact with each other. Um
not necessarily about impacting each other, but one can't be
described without impacting the other. I think that's what you you
mentioned. So, there there's those kind of elements to this which
is quite interesting. So,
um and then you also said that when you measure it it collapses. Am
I right so far? I'm trying to trying to think this back. Yeah.
So that that's and that's the measurements the key thing because
that's actually what kind of drives the ability to do quantum
computing. So if you think about super superp position so the
Schroing is cat story as we said cat is in the box
and there is a set of scenarios that define whether the cat could
live or die the poison and the vial I think in the in the boxing
cat that that but without any external influence the cat is neither
dead nor alive because we don't know if it's touched the poison. We
don't know if anything's It's just in this constantly observable
state, but we can't observe it. We can't look at it because I think
the theory is as soon as you open the box, one of two states, it
becomes the the permanent state as opposed to the the state. And
there was another way I read kind of that defines this model, which
is if you had a box of colored balls, you know, that you couldn't
see inside the box, but it was full of colored balls
and you put your hand in to grab a ball and bring one out and you
brought out a ball and you saw a red ball, you know, with a with an
absolute degree of certainty that there was a red ball in the box
and there is no longer a red ball in the box or sorry you know
there was a red ball in the box and could still be red balls in the
box but you'd have to keep checking them to check them out.
Yeah.
So and you can measure that probability of it being read again by
kind of the number of balls in the box versus the number of
different types of color balls versus the balls you've taken out
and what's left in the box is a probability state based on what
you've taken out.
Whereas a a quantum bit has a single ball in the box that is
changing through every single color in the rainbow constantly and
when you take it out it'll be one particular color but it doesn't
mean that it wasn't any of the other colors and it doesn't mean
that there are no longer any of the other colors in there but the
what it points to is because you've taken out and measured it there
is no longer any state in the ball box anymore
you've taken the so you've collapsed the state of the ball the ball
can no longer be every other state it is now fixed one state
and this is the paradox of quantum because if you think about this
back now in the binary computing world where the whole purpose of
binary computing was we set a transistor to a state and then we
read that state and we know with absolute certainty that that is a
one or a zero always. That's the measurement that lets us do
computing. You know, we we read that state, we change a data point
in the computer and something happens. In quantum computing, we
need to be able to measure that state but without changing its
state. And so that's where you know the principle of quantum means
that because a binary object has one state only but a quantum state
binary a quantum quantum object has multiple states. It essentially
doubles the potential for that computational power.
Yeah.
If you think about it, one cubit can perform what two bits could
do.
Two cubits can perform what four bits could do.
Yeah.
Four cubits can perform what 16 bits could. And and the numbers
grow exponentially. So you only need a hand cubits to suddenly be
computing at a billion bit level that we could never do in
computing today given that we're largely
That's a good analogy. Yeah, absolutely. So okay, so we've got the
theory behind um quantum but Now this is the hard bit, right? How
do we actually build a quantum computer? What are the architectures
that that we can use?
Yes, absolutely. The hard bit. Well,
the hard bit is getting your head around the idea that this
actually exists as a thing in the world. But yes,
then you want to build it. So um and so here's the interesting
thing. I mean cubits themselves are um again I I use the word I
keep coming across the word are grown because they are essentially
a figment of nature. They are a naturally occurring object at a
level subatomic particle level that we don't perceive as in our
human sense but they just grow. They are things that are out there.
You can create an artic a particle you can create a cubit. So we
can make
completely that's completely blown my mind already. Now
doesn't make sense but it does you know if you think that
everything you're looking at wherever you're listening or recording
this podcast from is made up of particles. Quantum atomic particles
make up everything in the plan in the world.
So there's kinds of then when we get into the world of cubits from
a computational perspective. Cubits that we can measure, cubits
that we can or not measure, sorry, but cubits that we can gate to
measure in some way. Uh cubits that are stable enough for us to
measure at a computational level. There's about seven different
types of cubits out there. Seven different ways in which we can
construct a quantum bit. Um and there's varying levels. And the
reasons why these different ones exist is because some of them are
much easier to make, but they are extremely fragile. high levels of
error, something called decoherence, which is when it collapses.
Uh, so high levels of decoherence. So you could build a cubit, but
it's so fragile and falls apart the minute you try and do anything,
but you can't computate anything on it. So it's kind of a bit
useless.
So you've got things like superconducting loops, which are well
proven. Everybody's using them. Google, IBM, a lot of them use them
in their models, quantum computing bits today. Uh, super great,
easy to do, but they're very, very fragile. So what you end up
having to do is building an extreme complex, difficult, and
expensive environment in order to house those quantum bits. We're
talking about bits that need to live at a uh at a zero degrees
Kelvin environment. So, colder than space, colder than anything on
the planet, colder than anything we understand as terms of
temperature. And they need to be held in that. And the slightest
noise, sound, vibration, solar flare, interference, electromagnetic
from the plan, from us around it could completely decoherence that
uh uh that that that bit. So that's you got the superconducting
loops. Then at the other end and this is where Microsoft's work
largely is is in what we call topological cubits. So topological
cubits are kind of like best described of imagine strands of fiber.
And when you tie them together there's a a point at which the the
the fibers overlap, you know, in the in in the in the strands.
Okay?
And as you move those strands and that kind of represents the
fragility of them as you move them around. because there's
vibrations, um, electromagnetic interference, all those things. You
find that the bit the the tie between the two bits of string moves
up and down the string a bit, but it's actually a bit more robust
than if they just weren't connected in some way. They'd fall
apart.
So topological cubits are very resilient, but and here's the rub,
largely still theoretical. We actually haven't built any of these
things. They're not, we know they exist, but building them is a
whole different thing because of this because of the extreme nature
of them.
But the reason why we're Microsoft is is going down that path is
because the resiliency which leads to lower er error rates, lower
decoherence and ultimately a better value return on the investment
should you be able to get past that kind of theoretical principle
of of actually making them. So you need that. So that's the
different types of cubits and you need cubits to build a computer.
You know that's that's kind of an air basic step.
Yeah,
they need to be stable. So we talked about so you know there's that
as I said it's quite easy to make them because they exist in nature
everywhere and we you know there's we got amazing innovations in
the world like the you know hydrron colliders and things in the
world that are understanding our world at this subatomic particle
level
big challenge is holding it in that measurable state and today in
most quantum computing facilities around the world that challenge
is because you need to create a stable environment that sits at
around 30 millvin so very much close to zero
yeah
absolute zero and you've got to store them there and of course they
don't last forever the longest living cubits we've seen today is
about 3 hours before they go into decoherence. So anything from 3
seconds to three hours. So you build this bit
and then it collapses and you got to build another one. So it's you
can see the challenge from a research perspective to build
something that is sustainably lived to be a computer
is a challenge
and and these universities and research facilities and that are
working on these are they so that presumably there's there's uh
lots of global efforts to kind of do this because there's two
different ways you said with the superconducting loops and
topological cubits which which what countries are kind of working
on this at the minute and what kind of which companies are doing
different anything interesting in this area
yeah look there's a lot of different pieces of work I mean
obviously Microsoft but also IBM Google Facebook are are investing
in the potential for quantum computing to accelerate kind of the
contributions they make to to society and to technology you've got
governments um you've got obviously you know the large superpower
of the world see this as a great opportunity. Interestingly enough,
it was one of the things that popped out of the Edward Snowden
incident was work going on in the in the governments to use quantum
as a mechanism to uh challenge the ciphers the the right
cryptography the world because obviously that's a tremendous
unfortunately in side of it there's a weaponization approach to
that which is that you know when you de develop a quantum computer
that can essentially pull to pieces like a baby pulls toy apart uh
current cryptography models to change the world in terms of you
know privacy and security. So there's a lot of that stuff going on
um you know and I think we should get to we should talk a bit about
kind of where the positivity is in that you know what
yeah so so before we get on to that that that element then so just
to just to finish off the technical side of it so we've got these
cubits we've got it's very cold it's nearly as cold as it is in
Wales at that particular point it's freezing
I think we get to 31 mill Kelvin in in Wales. But um we've got
we've got this cubit in in a in a stable environment, but now this
is the million-dollar question, isn't it? Now we've got to read it
um somehow and get some data from that.
Yep.
How does that happen?
So that's that's yeah, that's the really that is the million-dollar
question. And you know, there have been examples of this happened
already, you know, with a D-Wave computers back in 2011 and they've
progressed that before. Google have built computers, IBM have built
computers at Microsoft built computers, but you know, we're all
doing it in sort of different ways and looking at different
approaches to what we want to deliver with that quantum computer.
But I think it's a really good way to think about this because it
worked for me when I stepped it through my own head is I understand
how a binary computer works and I think if you understand that and
measure you compare it, then you get the sense of it. So here's
here's how it works.
Today you think about computer a CPU on your computer, we know it's
made up of of millions and billions of transistors on a piece of
silicon and what you end up with is a CPU. So In order to build a
cubit that you can compute on that can hold and measure a value
from you, you need lots of you need logical you need to create a
logical cubit. Logical cubits the CPU and the physical cubits are
the transistors. So you build lots and lots of these physical
little cubits, these physical atoms. You make them stable enough
that you can actually compute on them. You build them in into a
logical computer set like we build a CPU and that gives you the
kind of a quantum CPU if you like. That gives you the heart of the
computational capability. Then, as you know from being a programmer
in the 70s, you can't just tell a computer what to do. You know,
you need to have a language to communicate with it. So, you need
what they call a low-level language, something that speaks to the
very transistors on the board or in this case speaks to the bits,
the cubits,
the the physical cubits in the logical cubit. So, we would have
used assembler in the past.
We now need to build these low-level quantum languages. So, these
are this about languages now. And these are more about quantum
models and quantum algorithms because we're not talking to it in
the way that we think about language. There's no stack to load.
There's no registers and stack to load data onto. We're measuring
the we're observing the gated measured state of a of an object that
is spinning in constant,
you know, multiple states avoiding decoherence. So, you need a
language that can talk to that. And that's what a lot of the
scientific research is on is building those languages that talk to
cubits. But then above that, in order for you and I and me mortals
to be able to communicate with a quantum computer and give it a
problem to solve. We need a language, a high level language,
something that we can program in. And so for us, for Microsoft,
that's Q sharp that we built our quantum language. Um, and there
are others out there, but that's a language that a programmer can
then build an a case, you know, build a a piece of code that asks a
quantum computer to solve a particular problem uh in order in a way
that is logical enough for a human brain to make sense of, you
know, and that's kind of comparable to the way computers evolve
from binary to assembler to C to C and C++ and all this kind of
high level to where we are today where we're you know no code
writing and imagine that world where we get to a quantum world
where you're at a no code low code ability to basically ask a
quantum computer to solve a problem like climate change you that's
amazing vision so
yeah no absolutely and and I s I like just stepping through that I
suppose it's it's about being able to uh connecting with that I
suppose quantum CPU and then w with a low-level language and then
being able to ask the questions of that via your higher level
language so that then you can bring those things out because the
thing I was thinking as you were talking there was at some point
when the data comes out into a classical computer of some sort then
you know you've broken that cycle that's where the the actual
answer would come out so it it does it does it does blow your mind
massively this.
Well, here's a good example for you to think about and it in
keeping with our traditional geeky uh tone. If you remember your
hitchhiker's guide to the galaxy,
love it.
The deep thought computer that that was built
essentially became the planet Earth that was the computer to
calculate the question.
I know where you're going with this universe and everything. If you
ask the question, you've got to know what question to ask in order
to get the answer because if the answer is 42, but you don't know
what question you're asking,
you can't get there. So, it's it does come down to what questions
can you ask a quantum computer and you know a quantum computer in
the I say never never say never they say but in quantum computer is
never really going to be designed to run Windows office and word
like we are today you know it's not going to be running teams on
quantum that's not what quantum's about what quantum is supremely
good at what the quantum computing architecture is supremely good
at is evaluating billions of possibilities at once to determine
optimize roots or to determine you know the best path or determine
the prioritization of something. So that's, you know, where where
that's where quantum and AI collide because if you think about it,
what AI is always about is, you know, we we throw data at an AI
system. We give it a million versions of somebody's of language and
speech and at the end of it, we ask a AI system to speak to us. And
in a quantum world, we want to throw all this information at the
quantum and say to it, what is the way that we what is the right
way to tackle, you know, cancer cell reproduction? We give it
billions of data and it tests all of them at once. And you know
largely the problems we have today with the big problems in the
world. How do we tackle major diseases like cancer? How do we
continue to feed the planet develop nitrogen phosphates and and the
problems that they create for us on the planet? How do we assess
the damage and impact of climate change in a realistic data driven
way? There's too much data. There's no computational capability on
the planet today to look at all of that information and make a
decision.
Well, even something simple like the the pandemic recently with the
COVID 19 that you know that that that takes time just you know
doing some a simple project like that really to get a drug to
to um absolutely uh resolve that you know yeah that you know you
that just illustrates it straight off doesn't it?
Look it is and it's that time and you know you know that's an
interesting one because that's married up with
you know rules around how we do human trials and testing and the
process we have to go through but also there is just a the
sequencing of the of the of the of the co um uh disease itself, the
molecules and understanding that you know that took time on
computers. If we could do that at light not light speed but you
know at in seconds if we could start to break apart these we
understood the break that the breakdown of DNA we start to learn so
much more about individual pieces of our environment and and it
yeah
it it becomes less about those kinds of you know you know forgive
me for the bad example but you know how do I make more money in my
business because those are not computationally complex questions
but How do we understand the impact of 50, 60, 100 years worth of
of of carbon output and carbon management? How do we understand
that computationally? We don't. Today we do it in these macro micro
levels. But when you look at the big picture, this is where quantum
comes into its own. It's problem solving at speed.
Yeah.
You know, and and we talked about this earlier on that, you know,
today a lot of the quantum compute focused on cryptography because
that's a really interesting area for us to tackle because it's very
measurable and real. You know, we know today that most cryptography
models, the RSA algorithms, the uh encryption tools, stuff we use
to encrypt our browsing and all those things are based around
integer numbers and prime numbers.
Yeah, prime numbers.
Incredibly complex for computers to back calculate.
Yes,
it's easy to figure out what they are, but it's or at least to
build a pair, but it's very hard to actually calculate that back to
brute force it.
What quantum computers can do is look at that entire data set of a
you know a 248 bit key RSA encryption key or elliptical curve key
you know 248 digits
and figure out primes for it in seconds
which is computationally impossible to do today just by cost uh and
uh and and complexity and that's quantum supremacy or at least
that's a measure of quantum supremacy and if quantum computer can
do that then it's proven that better
and it's interesting isn't it because from from that element that's
one of the big worries obviously security around quantum but but
it's it's interesting when you start thinking about say from a
Microsoft point of view you know I'm working with um schools and
things at the minute and they they doing multiffactor
authentication stuff like that it feels a bit sometimes old school
they doing doing that but actually it's a hell of a lot more secure
than actually being able to put something in you know lots of the
banks that I use now um uh you know require like RSA can uh
encryption tokens. I know the ones in the UK do where where you put
your card in, but those things can be kind of um resolved quite
quickly, but just the the old school, if you call it, you know,
give me a ring to say there's somebody at the end of a line when
you're um when you're trying to um get into a system is actually
really secure.
Look, it it is. And there's a whole domain of of industry and
individuals and governments and organizations exploring what they
call post postquantum cryptography.
What does cryptography look like in a postquantum world? Because
it's a bit like
it's, you know, again, it's a terrible analogy, but it's one of
those moments of when you build a quant the minute we build a
quantum computer that can that is stable that can be used that can
calculate general outcomes as opposed to very specific kind of, you
know, targeted work. It's a bit like the moment when you build the
atom bomb. It's a bit like the moment when you um, you know, when
when when fundamental things change in the world and what was
before can no longer be anymore in the future and you can no longer
live in a world that is that is not quantum. You suddenly live in a
world that is postquantum where suddenly things that were no longer
computationally possible are possible. But there's so much
positivity then you know we touched on this but I think that um
when we think about what quantum can do and we marry it up with
cloudscale data storage and the idea that we as an or as a global
community can start to see the value in in sharing and coll
collaborating and doing research work together. When you marry that
up with the potential for quantum computing and the ability for AI
to have a positive impact on the way we operate as a as a species,
you start to think about those big issues of how do we ensure that
there's food for the planet? How do we ensure that the planet
survives? How do we ensure that we are saving the species and the
biodiversity in the world that is important to us all in some
degree? However, we sit on the on the spectrum of that argument. we
can do these things with quantum with AI and with cloud computing.
So it's it's fantastic.
It it it isn't it and and with the way you bring that together and
it does seem you know these projects and problems are really hard
and it does it does again take another kind of mind meld to kind of
think right how can I how can I how do how do we solve climate
change through a computer um through quantum you know it's it's you
know your brain I think my brain's been stretched in all kinds of
directions is doing this podcast. It's been brilliant. Um, in terms
of just just to kind of close it out a bit, I suppose is just in
terms of where we are with this, you know, you know, you quite
close to the cutting edge of where we are from a Microsoft point of
view, you know, what what timeline are we looking at for these
things? And and the other thing to also add to that equation as
well is obviously we've we've just released a a kind of
supercomput. So there's a there's a range of kind of processing
that we can do with supercomputers as well that are going to be
more accessible. So what what are your thoughts on how does
supercomputers and quantum fit in? I know they very very different
tools, but it might be worth talking about when we talk about big
data, but then also how close are we to being able to get to a
quantum machine, do you think?
Yep. No, good way to end. And and and you're right, there's a we're
on a journey and we you know, we have, as you say, just released
the AI supercomputer story in terms of messaging and how we how
we've built an AI supercomputer in cloud for for that exactly that
purpose and and and in some ways you can kind of see the evolution
of supercomputing not as a halfway house but as a journey on our
way towards quantum
um but it's still essentially mired in the binary world however you
know whether they are you know running on x86 whether it's AS6
whether it's risk whether it's you know FPGA whatever the
architecture of the chip is you're still constrained by some of
those very physical transistor-based gating that we have
yeah
and that's the journey we're on for bit I don't know but from what
I see and read and look and learn around this area and it's
certainly an area that has a lot of investment globally billions of
dollars being invested in quantum it's not a it's not a theoretical
domain it is a real domain of research and science I would have to
if I had to put a date on it I would say we're looking within the
next 5 to 10 years we'll have practical quantum computers uh and
within the next 20 to 20 years we'll have cloud enabled cloud ready
quantum computers, computers that everybody can access. And I think
that's the journey we're on, which seems like a long way to you and
me as I'm getting older. But I think about that in 20 years time,
we're doing quantum computing.
That's incredible to think.
And it is, isn't it? And and I think one of the things that, you
know, when we always like to go theoretical, but also bring it back
down a level as well sometimes. I think the important thing is
people be listening to this thinking, well, what does it matter to
my school or my environment or my, you know, me at home as an end
user? And I saw a good analogy um recently from our build
conference where they said well think about supercomputers and
quantum you know when these things are developed and solutions are
created for them and problems are solved then those things filter
down a bit like formula 1 where you know you get a formula 1 um
team that create things like traction control and and then suddenly
over time it ends up on you know all cars because it's a kind of
useful thing so I think um even though we might be listening listen
to this podcast and thinking well quantum you know I don't need to
use or think about quantum computing and I think it's going to
impact everybody right
it is and you know who knows maybe one day we'll have quantum
computers in our pocket like we do today I mean and that's you know
I say that and you kind of go but the reality is 50 years ago
people would have laughed at us if you said you had a phone in your
pocket that was better than the you know the the eniac computers of
the 50s that's you what we carry in our pockets today is laughably
insanely powerful compared to what we thought 30 40 years.
Very true. Well, thanks again, Lee. That was really interesting and
fascinating podcast today. I hope everybody um en enjoyed that and
you you can now go and put your head in a bucket of cold water to
kind of calm down your brain because that was that was really
interesting. But it is taking us on a completely different tangent
here, but some like signs of some really exciting things to come.
Thanks, Lee.
Pleasure Dan. Look forward to the next one.