rc3 preroll music
Herald: And welcome back from our studio
live, as you could see in Halle! laugh The next
talk will be Natalie Kilber. She will talk
about tales from the quantum industry.
Natalie works since many, many years on
quantum computers to make them real and
useful.
Natalie: Hi, I'm Natalie, and I've been
talking about the progress, the prospects
and the poppycock, the nonsense of quantum
technology, or you could also say tales
from the quantum industry. A little bit
about me. I'm a prehistoric creature that
has been there since the field emerged.
I'm the masses of stories for you today
and you might ask yourselves, but why are
you so gung ho about quantum computing,
buddy? Well, if you look at the Moore's
laws trends, then you know, already that
since 2000, the clock speeds have been
kind of stagnating and we're going smaller
and smaller, and IBM is going to fabricate
a chip of about two nanometers in 2023.
The problem here is the smaller you go, if
you go smaller than one nanometer towards
sub nano meter scales, then you go into
the quantum regime. And if you have a
single electron transistor, you already
have a quantum dot. That means that's a
qubit, that's part of a quantum computer.
But it's not reliable for classical
computing, for any classical computation.
So. Well, there we have it. We already are
at a quantum regime if we want to go into
the future. So why do I want quantum
computers? I'm gung-ho about speed. I'm
gung-ho about premium power. I want more
juice. So first, look at your PC, now back
to me now back in your PC. Sadly, it isn't
an eight Core i9, or maybe it is so yet.
Are you happy about your wiring? Well, I
don't know why, but this fellow's also
happy as a muffin about his wiring. This
is a quantum computer in Google's lab, and
you can see the wiring is not trivial for
this. And this is just one little chip. So
a quantum computer is an accelerator, and
you need a co-host CPU to to do any sort
of meaningful computation with it. And if
you look at the wiring here, there's a
different type of quantum computer. You
see an optical table with optical
components on it. I think this one is the
QuEra startup, and this guy's not that
happy about his wiring. You can see why.
There's lots of other examples that look a
bit difficult. And here specifically, you
see a lot of controls that are sending
signals into the quantum computer. And
this one again, is QuEra with a bit better
wiring. This is specifically a trapped ion
quantum computer, so they use trapped ion.
Quantum computers don't come in one
flavor. We have different flavors with
different bases of fundamental technology
that we use with different types of
components. So in trapped ions that use
photonic components, there are photonic
computers in itself. And for example, this
one is a huge cryogenic fridge. So you go
up to mini sub Kelvin stages right at the
bottom and the first picture you see in it
without his clothes, without the enclosing
the enclave. And then again at the top, the
massive wiring for just one chip. Then you
have other examples like, for example, AQT
Alpine quantum technologies over there
in Austria, and they know how to stuff
their cables really well and well. You
might wonder why am I not talking about
quantum inhalers? And well, if we define
them, if we define quantum computers, so
legs then a solid is a quantum computer
too, you know, why a solid any sort of
type of plant uses quantum phenomena as
well. So because of photosynthesis, the
light tries to travel as fast as possible
to the side, and they do it through
quantum tunneling and it solves it pretty
fast. Yeah. So no quantum inhalers. Then
look back at your PC, can it stand five
gigahertz? You think you're unhappy about
your clock speed? I think I won the
Complaining game a quantum computer can do
no more than 100 kilohertz, and that's
twice the speed of the ENIAC back in the
day. But then again, don't be so harsh in
your setup or on the quantum computers. We
still tinker with them with the
capabilities we have. You've seen in the
pictures before, there's a lot of wiring
there, components that are quite big, that
haven't been invented yet. So. The
bottleneck component of a quantum computer
of any set up, the slowest component is
your bottleneck clock speed or a
bottleneck in your clock cycle. And that's
why they're so slow in quantum computers,
they have digital signaling processes.
That means you have to convert digital
signals to analog and analog signals to
digital signals again. And we have that
everywhere in our phones and our cameras.
Imagine just sound that is analog, that
has to be converted into digital signals
or, you know, literally light photons. If
you take a photo in two digital signals,
that's a analog to digital interface that
we have. And here, because we're like
shooting microwave pulses and, for
example, superconducting computers and
qubits, that's kind of difficult to do.
So, yeah, you might say, but they are
parallel and they do everything a little
bit different. Yeah. For algorithms, when
you have such slow clock speed rates, if
your time to solution outlives you, that's
a problem. If you don't live to see your
solution, that's too slow and. You might
want to listen to your computer, listen.
So 30 dezibels or 40 decibels? This is
what economists. You can hear this.
Yeah, that's the
Woman voices speaks "Welcome to sound of IBM.
truth of that sound, but you needed it.
That's so annoying. One tempers at the.
You get into this by this. Then you have
this wonderful. The nightmare is now. It's
a quiet place. But one thing is always the
same. We always talk about size and with
size, I mean, we talk about qubits. You
might read in Wired or Spiegel or wherever
you want and hype articles or just
articles talking about the advancements in
quantum computing, how many qubits they
they could instantiate on a chip. IBM
released about 127 qubits in QuEra bit about 200
years. The trapped ion one and IBM was the
superconducting one of these cylinders and
the cryogenic fridges. But here you have
to discern a lot of physical qubits is
good, but a logical qubit is what you need
for computation. We have a high error rate
for just one physical qubit because of the
noise. Because of temperature. All types
of noise. All types of environmental
factors that you can't eliminate yet,
because this is this is quite fundamental
research how you can control these things
and how you can adjust the parameters so
noise stays low. And then again, we have
these types of signals and in our normal
classical devices where we need parody
checks, where we need error correction and
so do quantum computers. Error correction
was a huge field that needs to be
advanced, and we use things that are
called surface codes and these are error
correcting to get one logical qubit. So we
have reliable computation. We need a lot
of physical qubits. So you could say
there's a lot of overhead for those error
correcting code and parody checks. So if
you hear about those, many Qubits have
been have been accomplished by a company.
It's usually physical qubits, but then
another factor of 20, that's just one
logical qubit that you can use. Yeah,
that's difficult. And there's a famous
physicist that said, Well, he's still
alive, so it's actually on Twitter. And he
said, Well, Qubits are like children's
better to have a few high quality ones
than a bunch of noisy ones. Yes, I agree.
And John John Prescott has been at
Microsoft before, and now he's at 8WRS.
But at Microsoft, we witnessed
Maironascandal. Well, we thought we can
have a topological qubit that has no
noise. That means if we have one qubit, we
don't need these many physical qubits to
have one logical one because it is a
topological one with no with entrenched
error correction. One could say by the
physical nature. So you also run a blood
cells. Fittingly, it was called the
elusive marihuana particle because yes,
we've been waiting for 10 years for this
sort of maiorana qubit. But there was this
scandal. The big maiorana qubits
wasn't the big one. After all, they had to
retract the paper that said they found
one. So we're still looking for it. Yeah.
But then again, it is better to have a few
noisy physical qubits than none at all.
So, yes, quantum computing is full of
challenges. You've seen the wiring.
Getting so many wires into one of those
cryogenic fridges is very difficult. So we
have to find new ways to get see my those
those control those little controllers
into that fridge. So we have to reduce the
wiring, for example. And that's not a
trivial task because you get a lot of
resistance when you go colder for four
cables, for example. We're advancing
microwave technologies with quantum
computers. And one thing that kind of
worries me the most is that we don't have
quantum memory yet. So cue run from random
access memory because at the moment, a
quantum computer is just an accelerator.
So it's a read only memory. So everything
that is on the chip or on the qubits, on
the setup that is read out like that you
can store them or you can do any more
meaningful computation. So that's a huge
bottleneck. Another thing is the ethical
dimension we have to use in
superconducting quantum computers. A lot
of helium and helium has supply
bottlenecks, with just two companies Qatar
Gas and then one Northern Texas company
that supplies helium. That is not really
the problem, though, because we need even
something else. We need three helium,
which is an isotope, and that you get by
as a nuclear byproduct for tritium. That's
not something I want to I'm going to count
on, especially because these are limited
resources. And sometimes the components in
quantum computers themselves. They're also
rare earth metal. Those are also limited
resources. And then people keep talking
about democratizing quantum computers. Yet
you have other problems there first. Not
everyone needs access to something that
doesn't, doesn't solve a lot of things
yet. And to be honest with the security
controls in place, it's kind of an open
system already. But yeah, when we look at
quantum, we have to think about the
references. Which specs do you have to
look for? And the magic here is common
sense. I've shown you compared to what,
you know, the components that you know.
And again, the magic is common sense. And
quantum computers are very specific that
quantum technologies, the component of a
quantum computer, the sensors, single
electron sensors that we did. We use an
MRI that we're using spectroscopy for
microscopes and yada yada, even more
things and quantum communication types. So
semiconductors or or something else,
semiconductor components or just our
infrastructure and communication. They can
be part of the quantum technologies as
well. But you have to be also careful.
Everything is quantum now. It's it's quite
the hype. So finance is doing somehow
quantum. I don't know what other companies
think. Well, the buzzwords cyber isn't
enough and used two buzzwords: quantum and
cyber. I'm very curious what they do. Then
there's quantum transportation. I'm lost
here. I don't know what they do. I don't
want to know. And here, I mean, I'm sure
that is pain free. Yeah. You can also have
to be in and actually I really wanted to
find this in April 20 20. Yeah, so quantum
computing is claimed to solve a lot of
today's problems. Some companies claim
they're battling climate change, that
transforming the pharma industry to
transform the finance industry into the
break all encryption in the future, as we
know. So quantum computers will break the
internet in the future. Yet again, looking
at the reasons estimations, not including
the clock speeds or the actual
performance, that's difficult to claim.
But then again, looking at the references
and the facts and the specs, not all claim
these weird things, but the reference
facts like see lobs from IBM. These are
advancements that are meaningful. But then
again, we have this flood of references of
qubids of we're advancing this and that
complexity theory claims. But how are you
going to test these complexity theory
claims? Well, because we don't have the
Quibits, are simulated on a fantasy
machine. And if anyone like this old chap
here had to deal with theoretical
complexity resource estimates a.k.a.
fantasy language. Well, welcome to
imagination land this town. This town is
not a nice place for little fillies all
alone. There are lots of twists and
corners that could lead to the unknown.
Let me guide your way, and I'll be sure to
help you through. You could really use a
friend of the , and luckily, I've picked
my three favorite corners for you. Well,
quantum applications were applicability is
optional. So come on, let's start with the
very well-known topic of optimization and
the beginning of talks. I want premium
power. I want maximum juice. So VSLI
design, what is a VSLI? It's very large
scale integration and it means you need to
partition these little chips that we have.
And the first chip that we had back in the
day was an integrated circuit to help
people hear better. So it was a hearing
aid that Jack Kilby in 1958 designed and
this thoughtful design thought were the
basis of it is the basis for our
technology everywhere, and it's not a
trivial task to design these chips. So you
don't have a lot of waste and we can pack
more and more components on these little
chips. So integrated circuits. If you
don't know anything, IoT is that. Another
problem? The mathematical basis for this
problem is the same for network design or
less waste and manufacturing like stenting
or lasering, even flight scheduling
between cities has this mathematical
problem. Some might know it as bean
packing, max card or multi card problems.
You either seek to minimize or to maximize
an objective. So those commentary problems
are really one of the hardest one to
solve. And I like to call them
combinatorial black magic. These levels of
hard to solve are classes in themselves,
and this is actually a real graph. This is
a Peterson graph, and you can tell it's
it's black magic you might think. This is
not that hard, but I'll show you a
benchmark of max card problems. This one's
NP hard NP complete. This is one of these
fantasy language classes. It just means
that no polynomial time algorithms for max
card in general graphs are known. That
means, again, your time to solution
outlives you, and it's a problem if you
need to wait until your solution comes and
you die before. Or maybe it needs be a
couple of hundred years. I don't know how
long you live, but some say it's almost as
hard as beating cut of meat and dark
souls. But yeah, you don't live to see it.
That's the that's the drawback of this. So
yeah, you might think optimization, it's
black magic, it sounds weird, but you have
heard these terms before. I will
specifically be gung ho and talk about
nature inspired once the physics inspired
algorithms. But, you know, neural networks
you probably know to boost surge linear
programing, mixed integer problem
programing with a branch and cut, you can
see the max cut promise that's in the in
the brown part. And then, of course, other
nature based methods like bad surge,
genetic algorithms of small methods where
it becomes quantum is. And that's that's
what I like about the nature inspired part
the nature inspired optimization
algorithms. For example, they minimize the
Hamiltonian of an icing model. So whatever
mathematical but mathematical basis you
have, you minimize and maximize your
objective. Hamiltonians are something you
use in quantum computing and the icing
model I can explain later free up a little
bit more time. So what we need here to
with classical and quantum computers is
benchmark, so we can compare apples to
apples because classical computer and
quantum computers is more like apples and
bananas. So we need a common ground. And
if you want standardized benchmarks for
such problems, you can Google Chuck SHOOK.
It's a it's an open source benchmark suit,
and you probably see it in the slide. Good
old Professor Katzgrabor. He has written
this benchmark suit and he's gung ho about
cats, so please spare him of cat content.
So, yeah, I told you we'll get into the
max cut benchmarks. This is from a paper
of Cambridge, I think Cambridge quantum
computing and these little circles, these
little dots, steel nodes. And you can see
they have done they've done it on a
quantum computer for ten nodes. And it's
very complicated. Yeah. And the problem
here is when they went to 13 or twenty
three qubits, logic of qubits, they had to
simulate it. They had to put it on a
fantasy machine and classical hardware.
And yeah, that's that's also one algorithm
they used. Vicki variational quantum ion
solver and Qrolla, both of these are
approximate algorithms you can think of
very noisy, annoying quantum computers
that don't spit out results. But if you
run it 100 times, the majority of it will
be towards the correct regime. And yeah,
that's that's how you go about it. And
this is a relatively new paper, and I have
to say these resource estimations, these
are amazing results, and I'm not worried
about the algorithmic advances in quantum
computing because we have smart people and
I want more smart people. So if you want
to, you should get into it. So, yeah,
that's that's not what I'm worried about
yet. I don't want to solve something for
ten qubits or sorry, ten nodes on a
quantum computer, yet we can solve
something bigger. So this is from another
paper from a nature inspired, physics
inspired algorithm. Some already call it
quantum inspired. These are 100 nodes, but
at the lowest, you can see the physics
inspired GNN and Pi G, and they managed to
do it with a ten thousand nodes. So on
classical hardware, the quantum the
quantum algorithm put on classical
hardware to overcome the cube hardware
limitations by treating these physics
algorithms as optimizes. So from a
business perspective, if I want to have
maximum power and maximum Dru's, I would
use classical computers and use heuristics
from quantum and classical until the
quantum computers are ready. So, yeah,
neuro, I'm sorry. Nature inspired
optimization with quantum algorithms.
That's like putting neural networks on
steroids. Quite like that. This is the
paper for it. But yes, we've been far deep
into one corner. So I'll drag you back
here and I'll show you another one. Some
companies claim we were solving climate
change with it. We're transforming pharma.
And yeah, this comes from from ideas of
physicists. What I said. Well, the nature
is quantum mechanical. We might as well
need quantum phenomena to simulate what is
right. But yes, it's not that easy. This
physicists played bongos and strip clubs.
He's a real hero. Many of the physicists
he's known for that are talking about
chemistry. Here's ammonia. You don't think
this is difficult, but ammonia is used for
a lot of things in the world who use it as
a base, if there's something acidic, you
use it as a fertilizer, you use it in a
lot of things in chemistry and even raw
latex is has been transported with it or
anything that has an acidic nature. You
get it by a very difficult process. Well,
it's not a difficult but energy
expenditure high one. So you need high
temperatures and high energies to put it
into the harbor Bosch process, and it
accounts for two percent of the global
energy expenditure. It's a very famous
problem that quantum physicists wanted to
solve because it's really useful stuff
ammonia. And if we can cut two percent of
the global energy expenditure, that's a
good thing. It's not trivial, though,
Richard said it. It's not an easy thing to
do here. You can see just the active side
of an enzyme where you can produce ammonia
without high temperature and high energy.
Bacteria can do it by room temperature,
ambient temperatures. There's algae.
That's all types of bacteria that can do
it, and the active side is called from
FeMoco. You can see the resource estimates
for half of the sides, for the for the
energy to simulate, to see how this works,
because bacteria can do it. We don't know
how they do it. That's why we use so much
energy in temperature. The enzyme and the
material looks like this. And then again,
look back at the computer for both parts.
We need over 2000 logical qubits. Now,
think back, physical qubits are by a
factor of 20 or 100 more. So we're not
here yet. Then again, classical computers
can simulate it either, and we will
probably simulated that on quantum, but
we're not there yet. And to put it into
perspective, to the far right the orange
little molecules to form local bits in the
whole enzyme. And you might wonder what is
the THC cost while that's tens or hyper
contraction, so you algorithmic
advancements, I'm not so worried about.
We're pushing, we're pushing the frontiers
there. So yeah, but but the imagination
land, the most powerful magic is common
sense, and you should read it. So what do
you think? Do you want to use a quantum
computer or intermediate steps to find out
what we need? Well, what people do these
days is they're bit smarter and they do
simulated. They do use some digital parts,
but it's mostly haptic. Haptic means they
simulate a little bit and they tested in a
lab and got it tested in the lab. They can
funnel down what they need to simulate.
The paper I'm talking about for the
smokable and theological cubits is a very
recent one, so it's just a couple of days
it's been published and I think this is a
preprint even. And if you want to know
anything about resource estimates and
quantum computing for chemistry,
specifically Nathan Vibha and Ryan
Burbuja, a good place to look for. Then we
are still a quantum applications for
applicability is optional and it has been
true so far, hasn't it? Let's move to a
corner that hits closer to home,
cybersecurity. We have to be specific
here. I know a lot of companies claim
there won't be any type of encryption as
we know of in the future, because quantum
computers will break it off for once a
year to fifty sixty five fifty six years.
As bad as 256 bit mode can be broken by
quantum computers and symmetric key size
symmetric encryption methods are known to
be quantum secure the specific key size.
So not really. What people usually think
of as asymmetric encryption. So, yeah,
these are some resource estimates to look
out for. This is a Microsoft paper not too
long ago, and they said through a punch
line, it is easier to break elliptic curve
encryption than RSA. Then Google, not too
long ago, came up with two million noisy
qubits or physical qubits to break RSA
2048 bit in eight hours. And then also the
news paper saying that factoring a 2048
bit RSA integer can be done in one hundred
and seventy seven days with about a little
bit more than 13000 qubits, but with a
multimodal memory that does not exist yet.
These are incredible results over the
years in resource estimation numbers. Yet
again, let's put it into perspective. So
2012 he said, it's a billion in this year.
2021 isn't over yet. This year, Google
came up with 20 million noisy qubits and
then Gaussian came up with a little bit of
thousand or more, but let alone any
workable implementation of curium as a
purely theoretical nature as of now. So
we're still in imagination land when it
comes to breaking the internet as we know
it. It's time to leave Fantasyland, or you
might say, hey, but we did factor
relatively high numbers back there in
2013. You've heard this in the news. Well,
yes, we did. But if you know the base
beforehand, so if you know that with
thirty five, the number thirty five, you
can divide by five or seven if you know
one base, that's a really easy thing to do
and you can do that classically as well.
So IBM had to counter published that they
were oversimplifying quantum factoring,
and the algorithm you use for it is
Schwar's algorithm. It's one of the
purebreds quantum algorithms out there.
And then again, another one pretending to
fact the large numbers and quantum
computers. So no, we haven't been able to
break it so far. Another one in 2019, and
this is in very, very interesting one
because IBM goes close to these problems
and says, yeah, well, I want to test it. I
want to simulate it. A sorry, not
simulated. I want to test it literally in
quantum hardware. And they did so, but
they failed to factor just the number 35.
So I think we're safe for some time. You
have to think of quantum computers not as
a quantum threat, but more as a quantum
advantage. If someone knows how to steer
encrypted data and store it about 20 years
to decrypt it, you know, get it now and
decrypted 20 years later and stored
somewhere, they probably know where to get
it unencrypted as well. They're more low
hanging fruit for them, and I don't think
they will wait until the quantum computer
comes into fruition to do these sort of
things. So let's put the quantum thread
into perspective. Quantum computers are
logical extensions of Moore's law strand,
and quantum computers are tailor made for
simulating the behavior of quantum systems
like molecules or materials, and whether
they lead to breakthroughs in cryptography
or optimization problems. That is less
clear yet, but we're we're pushing the
boundaries. If anything, components of
quantum computers are pushing the
boundaries for us literally now, if we
have better seeds like quantum random
number generators for short Q, R and GS,
that is very useful. We need seeds that
are truly random. For example, in places
where we can't use true random number
generators that use entropy to generate
the random numbers because in a data
center, you don't want a lot of entropy,
so you don't want temperature diversity,
you want it to be cold and stay cold, or
sometimes you don't have the possibility
of having this anywhere where it's just
not there. So we do make things smaller
with it as well. You've seen the wiring,
so we have to design microwave technology
or any type of cabling, any types of
chips, um, pre processes that can go into
smaller and smaller spaces. So yes, we do
need quantum computers and the research
around it. We don't need it in business
settings just yet because they're not
ready. This is still very much fundamental
research, and we should note that so
mathematical concepts are more useful to
find. Also new ciphers when we're talking
about cyber security. And I'm not talking
specifically about peak. You see, there
are other mathematical mathematical
concepts for asymmetric and symmetric
encryption that can be that can be used.
But for now, let's leave imagination land,
and let's think about how quantum
computers interface with the world. Well,
I've shown you before that quantum
computers sometimes have a crude and
fridge, so if you look at the cylinder,
you see the the enclosure of it. So this
specific example, I use a superconducting
computer for now, I've told I've told you
before we need a host CPU and then a
control system. Lots of peripherals and
wiring to get into the cryogenic stage and
the enclosure. And there we usually have
an analog to digital digital interface.
And at the bottom where it's the cold is
the qbu. So you can think of it as, yeah,
a huge system. So this is an example of
Google's setup. And I think the key
concept that needs to be highlighted here
is the quantum computers are merely core
processes. And as such, they depend on
traditional compute environments to host a
quantum processing unit, a cube you
require as an analog to digital interface
to to convert those signals back and forth
and in turn, the application logic in the
host CPU. You may connect to a network
may. Some people think if I have it in the
lab and it's not connected to anything,
there's must be air gapped. But then
again, you know how loud these devices
are. So you kind of want RTP so people
don't become death and we've corona, you
kind of want people to work from home as
well, so they won't be arrogant. For the
foreseeable future, I guess we're for the
next year at least. So the issue of cyber
security and mass and quantum computing
resources that is rarely discussed, these
systems are and they will be hybrid
systems for the foreseeable future with
those CPU hosts with cloud based or
managed APIs. And we need reliable
services and secure services and
architectures as this arises. So
subsequently, the critical applications
and data these systems will handle and
store if it's the knowledge and the
algorithms, how to how to simulate for
Mocko we can produce the ammonia with less
energy expenditure if we design new
batteries. These are probably patents, so
we want to secure the data behind it and
those algorithms. So this means that all
classical security best practices hold for
quantum computers. So this example, the QC
lab at Google, sees enterprise system
constituted of a mix of Windows, macOS,
Linux, maybe Azure, Adi, SAS network,
containers, whatever platforms. And
they're part of these industrial control
systems and programable logic controllers,
pulses or discrete process control
systems. You know, anything in ICS, Escada
that is rarely air gapped or physically
means physically separated from any
network. So we need API hardening. I see
our security is not a big topic in quantum
computing yet because it's still just a
system on the internet, and it's not quite
ripe yet. People sell it and companies put
sensible data on there. So if this is back
in the day got infected with the MIMO worm
that was considered air gapped. No, I CS
system is truly, really arrogant anymore.
So before we offer quantum computing as
breakthrough accelerators, we need to make
them safe to use. So if you want to join
me, let's protect quantum computers from
getting pond. Thank you for listening to
me. That's talk.
Herald: Thank you so much. Um, we have
some time for questions. So, uh. Audience,
dear audience, please ask some questions.
The hashtags are on Mastodon and Twitter
hashtag RC3Chaos Zone, and the I.R.C. room
is the Channel RC three Dash Chaos Zone.
All right, and I will watch the questions.
All right. We have some questions already.
Herald: What do you think about rolling
out so-called post-quantum cryptography
now?
Natalie: Oh yeah. Post quantum crypto, I
know it's been. It's been a useful concept
promised and they have a never specific
problem in mind. And this is for the
national security and probably the
government, for infrastructure and in the
U.S. specifically. But they're thinking
of, along lived systems, the pig. You see,
you have the problem. It's highly
computationally intensive. So a lot of
infrastructure can't, can't cope with it.
So we need to deploy other infrastructure.
And if you're worried about your data,
you're in the intelligence behind your
data being stolen and then, you know, say,
for 20 years. Not many companies have
secrets that you can store for that intel
isn't that specific data that data steal
and store that is useful. So if you have
data, doesn't intelligence lie for over 20
years yet? It's useful if it's a
government side of, you know, it's a
nuclear bomb placed side or something very
critical. Yes, you have to think about it
now, and we do need time to implement the
infrastructure. And I mean, the hits close
to home. We've heard about crypto agility
to think that we would like to have, but
it's it's not the reality. We just have
legacy systems. We have to keep them
running. And especially if it's critical
infrastructure, you can just turn it off,
build something new and turn it all and it
has to work throughout. So you see is
useful for some problems, but not for all.
It's not a one fits all glove.
Herald: All right. All right, thank you.
The next question is, you talked about the
current number of qubits and how no
practical problem, a lack of the difficult
problems that the people are hopeful for
quantum computers to solve. The technology
isn't there yet due to the low number of
qubits. Would it make sense to serialize
the problems and run them on low qubit
count quantum computers? Does that work?
Natalie: I think I might not understand
the question fully, but I assume you mean
you package these little programs and I've
shown you the algorithm, the THC, the
tensor hyper contraction algorithm that
the chemical guys have used where we do
these sort of things. But then again, one
qubit you can think of roughly as one
transistor and you just need a couple more
than five or 10 to do meaningful
computations, as you've seen. That is a
very good question that we do package
these problems into smaller bits. And if
you go back into the slides or look into
the the the paper of Nathan Vibha and
Rayen Babbush around because you see that
you need still about more than two
thousand logical qubits, so you're spot
on. This is the direction that they wanted
to go and we have to go and there to try
to. Unfortunately, we still need more than
a couple of hundred.
Herald: So are there any current quantum
computers that are programable to do
something useful?
Natalie: I mean, it depends really useful.
It's very educational to use them. If you
want to have a have a workforce in 10
years that knows how to use them, you need
to do. You need to have, you know,
postdocs or master students who know how
to program these things. We need to know
how to write better compilers. What are
the what are the bottlenecks, how we can
swap gates, quantum gates? Some of these
are operations on a quantum computers. So
how we can swap these things and there
that's a useful thing for them to do in
any stage are workable quantum computer.
Just a few qubits is still needed to
advance the field and to advance the
workforce. So for me, it is still useful.
Herald: All right. Yea, it makes sense.
What do you see as candidates for earliest
productive uses of quantum computers?
Natalie: Oh, so you mean the question of
the killer application for quantum
computers? That's a difficult one. So for
cryptography or for optimization of I've
said it before, it's less clear. But for
chemistry, once we hit those 20000 or more
logical qubits, we'll see advancements and
catalysts. You see it from local molecules
to active side for the nitrogenous to to
get ammonia at room temperature. And
that's where I see the advancements for
four small catalysts for get alloys and
metals to find better storage batteries.
There's there's still a field out there
that we have that we couldn't simulate on
classical because it's quite intractable.
But we're pushing the field and I think
chemistry could be one of the first ones
that's just not there yet.
Herald: All right. Do you also think
that'll be the earliest one's chemistry
applications?
Natalie: Small molecules for catalysts?
Yes, they could be. I mean, the smarter
people than me out there might have better
ideas. Maybe design a completely new
battery storage or I mean, ammonia is
being used in fuel cells as well for
storage. Maybe they'll simulate how to get
ammonia, cheaper energy expenditure wise
and then use it to store, have better
storage and fuel cells yet. I mean, there
are some quantum computing services out
there that are kind of interesting depends
what you're looking for. Yes. In
Cambridge, quantum computing offers a
three qubit encryption suite if you want
to do QCD. I mean, it's a fun toy game.
I'm not sure if it's very business
relevant, but if you want to look at your
current infrastructure could hold it.
That's an interesting one. Quantum
communication components, especially in
that part of the quantum tech world, is
more advanced and more ripe. So a lot of
devices in quantum communication you can
use now already. So it's just about your
risk appetite. Do you want to, well, spend
a lot of money on it? Do you want to
invest into it and try it out? There are
some test beds in Berlin and Paris where
they're trying out QKD networks yet.
You know, this is telecom. This is not
quantum computing, but it would be the
backbone if we want to have a quantum
internet where then again, quantum
computers are useful. So everything is
useful because it's it's an intermediate
step towards something you would like to
have. But most of the things in quantum
computers, they don't fit classical
solutions yet.
Question: All right. You talked about the
attack vectors on quantum computers and
dramatizing this a little bit. And what is
the worst case of the quantum computer
getting on?
Natalie: I mean, worst case is some
company has their sensible business data
on it, and they harvest that. I mean,
because they're not, you know, they're not
critical components as of yet. And there
are a lot of down times because they have
to recalibrate them, you know, get them
off the grid, see if the fridge works or
do some sort of maintenance. You don't
have to use usually SLS with them yet, but
think about all these companies that don't
know what they're doing, and they might
have, you know, the critical data up there
in the cloud pushing it there. And if the
API isn't, isn't hard and if it's, you
know, open access for everything, they may
just have low hanging fruit to pick out
their.
Herald: Thank you so much, Nacho. This was
tales from the quantum industry. Bye
Nacho. Thank you. Thank you. All right.
Our next talk will be at 17:30. What is
Algarve? It's about a community that live
codes music and celebrates the artifacts
and the algorithms that they use.
RC3 postroll music
Subtitles created by c3subtitles.de
in the year 2022. Join, and help us!