CQC2T: Centre for Quantum Computation & Communication Technology

00:06

my name's Hyman Devitt I'm a founder of

00:10

a quantum consultancy company called

00:11

h-bar I'm also the chief science officer

00:13

of a new startup called Turing and I'm

00:17

delighted to be asked to chair this

00:19

panel on the quantum applications stack

00:21

and how it might influence development

00:24

within the private sector and

00:25

revolutionized the world so first of all

00:28

we do have representatives from both the

00:31

startup community more established

00:33

corporate representatives and from the

00:36

academic community and I thought

00:37

initially I just run down the panel and

00:39

let everyone do a minute or so

00:41

introducing themselves and what their

00:44

primary focus is so I suppose we'll

00:46

start with Martin down the end from

00:47

Microsoft thanks Simon a Martin role or

00:50

with Microsoft Research let me just

00:52

begin by saying it's a very exciting

00:54

time to be working in quantum computing

00:56

you see a lot of emerging devices we've

00:58

just heard about a Google effort and

01:00

several other kind of companies and

01:02

other outfits that try to build a

01:04

quantum computer

01:04

some of them allow you to actually play

01:06

with it I personally work on on the

01:08

algorithms side I try to find problems

01:10

where it can wait we can have it and a

01:12

tremendous speed-up over what you can do

01:14

classically of course historically the

01:17

first such problem was factoring which

01:18

has implications in security and privacy

01:20

but we're going to move beyond that and

01:23

find trying to find problems that are

01:24

more of an immediate business value you

01:27

heard by in Alan's talk about chemistry

01:29

we have big chemistry effort going on at

01:32

redmond we have matthias Troyer just

01:34

recently joined our group he's one of

01:35

the leading a classical experts on the

01:37

topic we have material science

01:39

applications we are looking into machine

01:41

learning applications that's kind of a

01:43

new area and the applications are not as

01:46

clearly defined on my like personally

01:49

I'm interested in costing algorithms so

01:51

you want to kind of at this point it's

01:53

everything is unclear what's the final

01:55

gate said what we can actually support

01:57

but you can still even at this point

01:59

make some statements about how big the

02:01

machine needs to be roughly to run some

02:03

applications I mean I'm interested in

02:06

that try to build the software that

02:08

allows us to do it and we kind of cost

02:10

out the number of gates the number of

02:12

queue

02:12

it's that that's kind of part of the

02:14

Microsoft approach to what's going on

02:16

computing great thanks mom Matt from

02:19

juicy we're sure my name is Matt Johnson

02:20

I'm the CEO of QC where we are a quantum

02:24

computing software company based in Palo

02:27

Alto California I guess the the main

02:30

points to think about with us is first

02:31

of all we're focused on applications and

02:34

the application sets we care about our

02:36

quantum machine learning finance and

02:39

search and search for us in the broadest

02:42

sense of the word means finding a needle

02:44

in the haystack for graph based or

02:47

network based problems another

02:50

characteristic of our firm is that we're

02:52

specifically focused at enterprise

02:54

problems and making sure that the

02:56

software we're developing is reachable

02:59

and usable for novice enterprise users

03:01

so the composition of our team reflects

03:04

that we're computer scientists quantum

03:07

algorithm people and and physicists so

03:12

we work together as a team to build the

03:14

technology stack that sits on top of the

03:16

bare metal or the API that the hardware

03:18

vendor will expose up through the

03:21

application side all right thanks man

03:23

Michael from Q branch hi good afternoon

03:26

my name is Michael Brett I'm the CEO of

03:28

a company called Q branch we're

03:31

headquartered in Washington DC and have

03:33

a development team based in Australia

03:36

where a data analytics company the bulk

03:39

of our business is focused in applying

03:41

machine learning and data science to

03:43

challenging enterprise problems but we

03:46

see quantum computing as a long-term

03:48

strategic differentiator that we want to

03:50

be involved in so we've been investing

03:53

some of our equity and some of our our

03:55

partnerships into understanding the

03:57

quantum computing landscape that sort of

04:00

applications that might be enabled by

04:02

quantum computing that would have an

04:03

impact on our existing business and

04:05

hopefully those two business areas over

04:07

time will merge together and we'll

04:09

introduce our our classical analytics

04:11

customers to quantum and our quantum

04:13

customers to the analytics work that we

04:15

do on the classical side it's a pleasure

04:18

to be here thanks for the invite great

04:20

things Michael and Gabriel from

04:21

Volkswagen hi everybody i'm gabriela a

04:24

composer

04:25

from the Volkswagen data lab here in

04:27

Munich forks wagon meaning one of those

04:29

companies that we heard is going to be

04:31

disrupted by new technologies so the

04:36

data lab here for which I work for is

04:38

basically a center of excellence for

04:40

everything related to data analytics of

04:42

dense analytics within the company

04:44

Volkswagen that means we work with all

04:46

the group brands that are owned under

04:49

the umbrella companies of the Volkswagen

04:51

Group our idea is to bring together

04:53

business experts data analysts

04:56

technology experts and try to build data

04:58

driven innovation for the company as

05:00

part of this effort we focus on

05:02

different areas like again big data

05:05

analytics machine learning deep learning

05:06

but we also have some research topics

05:09

like that go into the direction for

05:11

example of natural language processing

05:13

and chat bottle iPad one of the research

05:16

topics we are working on is also quantum

05:18

computing to try to prepare us for the

05:20

new revolutions coming and the new

05:23

products that we will be able to create

05:25

with that so that's why I'm here and

05:27

thanks for the invitation great thank

05:29

you and finally Michael Bremner from UTS

05:31

in Sydney and the CQC T now that's right

05:35

so as you've just said I work for two

05:38

research centres both based in Australia

05:41

I work at the University of Technology

05:43

Sydney --zz Center for quantum software

05:44

and information which is a Research

05:47

Center which is focused on the

05:49

theoretical aspects of quantum computing

05:51

so that goes everywhere from blue sky

05:53

research into the mathematics underlying

05:55

quantum algorithms all the way through

05:57

to working with teams for

05:59

experimentalists on developing

06:02

benchmarking and applications for their

06:03

devices I do that in part with the

06:06

Center for quantum computer technology

06:08

the quantum computing and communication

06:11

technology which is an Australian

06:13

government funded research center great

06:17

so what I thought I'd do is try to

06:19

tailor off from what Alan did but try

06:22

and give it a little more details so

06:25

first of all I'd like to get Mick and

06:27

maybe Martin might want to jump in on

06:28

this to sort of give us an outline of

06:31

sort of the the quantum algorithm and

06:33

the quantum application landscape

06:34

because we've talked with hope from Alan

06:36

there's two very important regime

06:39

here there are things that non error

06:41

corrected quantum algorithms and what

06:42

can be done there all the way up until

06:45

fully quantum error corrected Universal

06:46

algorithms which as was mentioned is

06:49

going to take some time so make if you

06:52

want to talk about sort of this area a

06:54

little bit and then Martin might want to

06:55

jump into some of the work that

06:56

Microsoft's done to try and find few

06:59

qubit applications and try to get around

07:02

not requiring error correction okay so

07:06

that's gonna give me the whole history

07:07

of quantum computing in thirty seconds

07:09

but all right so I guess it's important

07:13

to understand the quantum algorithms

07:16

development has come has been going on

07:18

now for since the early 90s

07:21

and to begin with it was really a

07:24

mathematical topic it's a topic of like

07:26

studying these mathematical structures

07:28

and it was done very hypothetically yeah

07:33

imagine if you had this thing called a

07:34

quantum computer which maybe one day

07:36

someone develops and people focus on

07:39

developing algorithms and for universal

07:40

quantum computers of work you know

07:42

ideally with error correction and

07:43

everything like that in recent years

07:46

this is really shifted and I think

07:47

Marcus also played a big role in this as

07:49

has Google and a few others too as

07:52

Hardware has got like much better to the

07:56

point where we're now actually you know

07:57

talking about building really

07:59

interesting devices we've focused a bit

08:03

more on near-term applications because

08:05

now we can actually see the thing we

08:06

need to develop forward we can talk

08:07

about what is the fidelity what is the

08:09

architecture what's it going to look

08:10

like we're now developing applications

08:13

much more specifically suited to these

08:14

sort of tasks now as was alluded to by

08:18

Allen like we're approaching this regime

08:21

which people call sort of those raging

08:23

with quantum computational supremacy

08:24

which is its it can be a sort of vaguely

08:28

defined term which means we're heading

08:31

towards building devices which in

08:34

principle could have the capability to

08:36

for very specific applications I'll

08:39

perform every classical quantum computer

08:42

sorry classical computer to begin with

08:45

these devices these light applications

08:47

are very

08:48

we were put it and the way that they've

08:49

been developed has been to you know find

08:53

some mathematical leverage which you

08:55

know lies down in this you know the deep

08:57

sort of structure of some problem and

08:59

then build that up to talk about an

09:04

actual sort of application or a

09:05

benchmarking tasks that a device can do

09:07

which is quite different from the sort

09:10

of traditional way of doing development

09:12

for enterprise in an industry very much

09:16

agree with those points at Microsoft you

09:18

of course observing very closely the

09:20

developments around quantum

09:22

computational supremacy which means kind

09:25

of you tried to separate the classical

09:27

world from the quantum world and find a

09:29

problem that you cannot solve on a

09:30

classical machine period so the trouble

09:33

with it in a sense is that none of these

09:35

problems are really seem really useful

09:36

was we of course want to set our sights

09:38

to problems that we care about the

09:41

hardest problems people care about and

09:43

we cannot solve them in any reasonable

09:45

amount of time even if we cover the

09:47

entire planet with computers we could

09:49

not solve them and one area where we

09:51

found such problems are in computational

09:53

chemistry so understanding chemical

09:55

reactions it's very important we heard

09:57

in in Alan's keynote about the problem

10:00

of nitrogen fixation for instance it's

10:02

done currently like harvesting nitrogen

10:05

from the from the air it's done

10:06

currently with a very inefficient

10:08

efficient process called the

10:09

abba-baba-baba process which is like

10:13

happens under high pressures high

10:15

temperatures and you need catalysts for

10:17

this and has a very relatively lowly low

10:20

yield so a quantum computer could help

10:23

us it has been shown on on a piece of

10:25

paper that it could could help us kind

10:27

of innovate there and find better

10:30

catalysts that help us having higher

10:31

yield for the process of course now it

10:34

kind of you have to ask several

10:36

questions first of all how large has a

10:38

quantum computer does it have to be to

10:40

actually implement this like how many

10:42

qubits how many operations will you have

10:44

to do and kind of can you really solve

10:47

the optimization problem of finding a

10:48

catalyst and then implementing all these

10:50

things are kind of we have done some

10:53

baby steps towards that we cost that out

10:55

initially that you will need around 100

10:57

logical qubits

10:59

to understand the process of nitrogen

11:01

fixation the way nature does it using a

11:04

molecule called nitrogen ease and nobody

11:07

really understands how nature does it

11:08

but having a quantum computer we could

11:10

at least kind of have a chance of

11:13

understanding it and there it's maybe

11:15

interesting for the late for the lay

11:17

persons to know like one of the key

11:19

applications is just simulating physics

11:21

so we try to build this machine this

11:23

incredibly contrived machine and one of

11:26

the key application is just simulating

11:27

other physical systems and that's one of

11:29

the killer applications presumably is

11:31

just to do that

11:33

if we have such a machine we can apply

11:36

to problems and material science maybe

11:38

find better ways to construct

11:39

superconducting materials that would

11:41

super conduct at higher temperatures

11:43

that would be a really really big deal

11:45

breaker we could we could hopefully have

11:48

kind of the catalysts I mentioned in

11:50

machine learning would we could

11:52

hopefully use a quantum computer to

11:54

train models in a better way there early

11:57

indications that that might be possible

11:59

when you look at deep networks you can

12:01

model it by a Boltzmann machines and

12:03

it's known that a quantum computer can

12:04

help train Boltzmann machines in a

12:06

better way so better is kind of tricky

12:08

to define in that context it means

12:10

specifically that a method that it's

12:12

classically used based on stochastic

12:14

gradients can be improved on by having a

12:17

better quality gradient in a like if you

12:19

have a computer you can have the kind of

12:21

you can point in the right direction

12:22

make a step so in the in the gradient so

12:24

a few very domain-specific applications

12:26

that's kind I think one of the keywords

12:28

that you said in there is when quantum

12:30

chemistry you're talking about a hundred

12:31

logical qubits error correction is still

12:34

going to be required at this level so I

12:36

mean if I can talk to Gabriel about

12:38

certainly the interest of these bigger

12:40

companies such as Volkswagen I mean

12:41

where do they come in on this in regards

12:44

to what they're looking for are they

12:45

hoping that there are going to be

12:47

applications out there at this sort of

12:49

50 qubit to a hundred qubit level not

12:52

logical qubit qubit level well I think

12:56

we are taking the proactive approach

12:59

that Alan mentioned before so we are

13:01

trying to see what's already out there

13:03

and how we could use it for realistic

13:05

problems and we had an experience just

13:08

this year a few months ago we were able

13:12

to try

13:12

our first let's a quantum application I

13:16

didn't have I don't know if I qualify as

13:19

a quantum developer so I didn't have the

13:20

gut to stand up before but basically

13:22

that's what we did we had a partnership

13:25

together with our colleagues at d-wave

13:27

and we tried to set test their systems

13:28

so not at 100 cubits scale but a

13:30

thousand to 2,000 and what we did was

13:34

with a small team locked in a room for

13:37

two months and the Hut without pizza -

13:41

we tried to build something useful out

13:43

of it so we tried to to ask ourselves ok

13:46

so we have a real quantum computer here

13:48

that is available this is the top of

13:49

tech the technology that we have today

13:51

so can we use it to solve a real problem

13:54

that matters for our industries

13:56

something along the lines that was

13:57

discussed before and the things that the

13:59

thing that we did was we created a small

14:02

application that ran on the d-wave and

14:04

that could optimize the traffic flow so

14:07

optimal routing for many cars on the

14:10

street we had some very interesting

14:13

results that represented at cbiit this

14:15

year we are still working on that but

14:17

the outcome of this exercise was that

14:19

yes there is something that we can

14:20

already do there is something that

14:22

already works with the current quantum

14:24

technology of course the limiter is a is

14:27

the scale at which we can push this

14:29

problem we are now able sorry to let say

14:33

perform this optimization in a few

14:35

seconds so from 5 to 15 for a few

14:39

hundred cars on the street and we can

14:41

select between multiple possible routes

14:44

the optimal route for each and every car

14:47

based on the behavior of all the other

14:49

cars so yeah so this is for example one

14:52

of the possibilities of some new let's

14:55

say applications that are not common in

14:57

our industry yet that could be

14:59

differentiators for our company we are

15:02

still of course exploring other possible

15:04

applications that are related to

15:06

optimization in many other aspects of

15:09

the things that we do I mean in the case

15:11

of how how Volkswagen look at this this

15:13

problem itself is it do you reach out to

15:16

the quantum companies and say hey we're

15:17

interested what do you suggest or do you

15:20

have your own in-house teams who have

15:22

some understanding of some expertise

15:24

just

15:25

then go this might work for us and so

15:28

we'll reach it right so of course we are

15:31

not able to do this all ourselves we

15:34

have worked in a close partnership with

15:36

d-wave to perform this exercise we are

15:39

also gaining some some education

15:41

ourselves I'm a physicist as background

15:43

so I sort of understand some of the

15:47

technical problems not all of course but

15:48

and my colleagues too are trying to

15:51

learn more but let's say we are we are

15:54

going through two different paths one is

15:57

to be proactive again with respect to

15:59

the evolution of the technology the

16:01

other one is trying to build within the

16:02

company already the competences that

16:04

will be needed

16:06

once quantum computing becomes really

16:09

commercially available and more spread

16:12

as it is right now yeah that's something

16:14

that I've noticed quite a bit is that a

16:16

lot of these companies are looking to

16:17

develop competency within this area so

16:19

Matt and Michael would feel free to sort

16:22

of jump in on each other you guys your

16:25

company's sort of work in this software

16:26

space it's sort of these these non error

16:28

corrected levels to try and consult with

16:32

people or develop these applications

16:34

with with various stakeholders so how do

16:37

you guys tackle I mean you both work

16:39

with financial sectors a little bit as

16:43

well so how do you approach that again

16:46

do they come to you or do you go to them

16:48

yeah so building on like on the

16:51

experience from Volkswagen that that is

16:54

the style of project that we typically

16:56

get engaged in where a company sees

16:59

quantum computing on the horizon that's

17:02

coming up towards them they're picking

17:03

up signals like Google are putting out

17:05

today that hardware will soon be ready

17:07

in a tractable timeframe to start to

17:10

think about this technology and so

17:12

organizations are seeking to understand

17:15

what the impact is going to be on their

17:16

business and one of the troubles with

17:18

quantum computing is that the specific

17:20

application in that business is

17:22

non-intuitive and non-obvious in an

17:26

initial meeting so you you need to spend

17:28

some time with that company to

17:29

understand their particular processes

17:32

where they're spending a lot of

17:33

computational power already or what are

17:35

the problems that they're

17:37

avoiding because it's too

17:38

computationally intensive and how

17:40

quantum computing might be able to have

17:42

some impact there I think it's important

17:44

to consider the importance of emulators

17:47

of quantum computers so these are

17:49

systems that run on classical machines

17:52

but emulate the behavior of a quantum

17:54

computer at least to some level of

17:56

fidelity into some number of qubits and

17:58

that provides a inexpensive accessible

18:01

way for an enterprise to start to

18:04

develop those capabilities and skills

18:05

and build the awareness of quantum

18:07

computing and demystify ER in in a

18:10

practical way so getting hands-on with

18:12

an emulator is something that can be

18:15

done ahead of the hardware development

18:17

and then once you've got applications in

18:19

mind then test those on the hardware as

18:21

and when it becomes available so I agree

18:25

with what Michael said I would add a

18:27

couple of points and I think this

18:29

applies to any member of the quantum

18:31

computing software ecosystem the most

18:34

important thing to do is to be working

18:38

very closely with enterprises regardless

18:40

of the sector they're in and and there's

18:43

this amount of creativity that's even

18:46

more important than technical knowledge

18:47

of the hardware topology or any other

18:50

aspect of quantum computing and it's

18:53

that ability to map the problem the

18:55

enterprise has and and to have a quantum

18:58

representation of that to enable it to

19:00

be run on a quantum processor and that

19:02

sounds like a very technical thing to do

19:04

and it is but it's actually it requires

19:07

an equal measure of creativity and so

19:10

when we I guess tactically think is a

19:11

company where we want to focus our time

19:14

clearly there's a better scientific

19:17

argument to say let's focus on quantum

19:20

chemistry quantum quantum materials

19:23

quantum quantum physics because those

19:25

datasets are already in a quantum state

19:27

and they can run kind of natively on a

19:29

quantum processor but we're actually

19:31

taking a bat a very clear bet that the

19:34

hardware vendors the ones that are in

19:36

this room so that's IBM Google d-wave

19:39

Righetti we're taking a bet that the

19:41

engineers there are going to be able to

19:43

pull forward useable early generation

19:46

quantum processors in a fairly short

19:48

period of time

19:49

so we are not just doing that kind of

19:52

application set but finance working with

19:56

the Shah one of our investors working

19:59

with Airbus on engineering design and

20:02

test problems though those are areas

20:04

that may be somewhat under addressed

20:06

right now but we're taking the bet that

20:08

quantum processors are going to be more

20:10

powerful sooner than the sort of the

20:14

conventional wisdom suggests they are

20:15

and so that's really where we want to

20:17

put our footprint down so I mean how do

20:19

you I mean a few qubit quantum

20:22

algorithms that have a marketable impact

20:24

something that that is demonstrably

20:25

better than what we can do classically

20:27

has an application that can be marketed

20:29

or commercialized I mean this has been

20:31

the holy grail of the academic community

20:33

to a certain extent for the last twenty

20:34

years and still really doesn't exist yet

20:37

so Mike what you talked about emulators

20:39

now I mean emulators only take you so

20:42

far

20:43

arguably you can do a lot of it on pen

20:45

and paper once you get to twenty thirty

20:47

cubits and then when you're pushing past

20:49

thirty cubits you find your classical

20:50

computer can't keep up so within what

20:53

you see wearing q-branch are doing I

20:55

mean how much of it is app development

20:57

at this sort of non error corrected

21:00

level so we're going to have a really

21:02

tough time as a community if we can't

21:05

get quantum programming into the hands

21:08

of classical computer scientists and

21:10

software engineers and to make it normal

21:12

for graduate from computer science from

21:14

Stanford who's never thought about a

21:16

quantum computer before to make it

21:18

normal for that computer scientist to

21:20

pick up a library and start to use it

21:22

and think about it as a classical

21:24

problem so that means building out the

21:26

libraries and the toolkits and the

21:28

emulators that exist within a normal

21:30

development environment that can run on

21:32

premises at their location whether it's

21:34

at a financial institution or university

21:37

or wherever to get access to the data

21:38

sets that they have and so for for our

21:42

company at least the way we think about

21:43

our role here is to do kind of a

21:46

top-down work so to look at the

21:48

applications that might be relevant and

21:50

to drill down words and building out

21:53

those libraries and compilers and

21:54

toolkits that go into the hands of

21:56

regular developers well the hardware

21:58

teams and the people close to them do

21:59

the bottom-up build the bare metal the

22:02

all systems and the the abstraction

22:04

layers on top of that and so building

22:08

out that workforce that has some

22:10

understanding of quantum computing and

22:12

can use those libraries that is one of

22:14

the enablers to making sure we can find

22:16

the applications down the track and just

22:19

really quickly to add to that you

22:21

brought up a very interesting point you

22:22

talked about for circuit model machines

22:24

at least let's talk about adiabatic or

22:26

annealing processors in a second you

22:29

talked about the lack of development so

22:32

to speak it's an odd thing to do so as

22:35

Michael Wright and and certainly there

22:38

are you know six or seven quantum

22:40

algorithms that have provable end to end

22:43

speed-up potential the issue is those

22:46

algorithms were developed in the

22:48

scientific and research community and

22:50

they really weren't expanded or put to

22:52

put to the test with real data sets so I

22:54

think the interesting challenge will be

22:55

to take these algo primitives that are

22:57

out there and to mash them up against

23:00

enterprise problems and to force the

23:02

algo community the practical algo

23:05

community to try to shoehorn those real

23:08

problems onto those algo primitives I

23:10

mean I mean those algorithms that do

23:12

have provable speed-up unfortunately do

23:14

require extensive error correction to

23:16

implement sorry you want to jump in yeah

23:23

I'm just going to jump in and say I mean

23:27

it's really important to understand that

23:30

quantum computing speed-up is it's not

23:33

just a matter of which I think everyone

23:35

here agrees with but it it's important

23:37

people understand it's not just a matter

23:38

of something going faster so it's really

23:41

hard to map problems into a subroutine

23:45

which is you know will deliver you a

23:48

speed-up and and it really deeply

23:50

depends on in some ways the architecture

23:52

the way in which especially we're

23:55

talking about near-term devices like not

23:57

everything's going to work perfectly and

23:59

sometimes those imperfections can

24:03

swallow your speed up completely or it

24:05

turns out that you know a classical

24:07

computer can do just as good a job this

24:09

is a thing which I think everyone the

24:11

stage here is

24:12

is kind of focused on trying to get to

24:14

the bottom of from different

24:15

perspectives I think it's interesting

24:17

that people looking at sort of

24:19

enterprise development and how do we map

24:21

those problems onto the existing

24:24

subroutines and from an academic

24:26

perspective there's really a little work

24:28

on finding those points to differ a

24:30

differentiation between classical and

24:32

quantum computers and trying to leverage

24:34

that out I think that's yeah it's a bit

24:38

of a summary of what I think is going on

24:40

I mean it is the interesting approach of

24:42

Microsoft is the Marten you want to

24:44

gonna have to say it real quick because

24:45

we're getting short on time in kind of

24:48

this annyoing so the this method that

24:50

people might have heard of called

24:51

topological quantum computing there's a

24:53

bit of a distinction there are

24:54

topological quantum error correction

24:56

codes which work on qubits

24:58

but what Microsoft's approach is is

24:59

these systems with intrinsic topological

25:02

protection and it's a bit of a gamble

25:04

because it would in one sense

25:05

mitigate the error correction

25:07

requirements but it's also based on

25:09

particles that don't really exist yet

25:12

right

25:13

yeah okay let me try to set it in a few

25:15

remaining minutes so for the problem of

25:19

decoherence has already been mentioned

25:20

it's a very tough problem that the

25:21

corner computer has to deal with

25:23

intuitively it's like kind of like a

25:25

negative interest rate like if you put

25:27

money in the bank and after a year you

25:29

have instead of hundred dollars you have

25:30

99 dollars and then it keeps compounding

25:32

the negative interest rate right and you

25:34

can see that quickly that that kills the

25:36

computer if you have a number of steps

25:37

you want to do it just drops

25:39

exponentially the amount of signal you

25:41

get from the machine and it makes it

25:43

hard to scale up a quantum computers as

25:45

simple as that so there several ways out

25:47

of that we could just go bare and work

25:49

with the pair qubits and hope that our

25:51

circuits never have to be very deep and

25:53

that might work with the great proposals

25:56

on quantum variation eigen solvers that

25:58

actually might give you a good good

26:00

answers the other approach is to add

26:01

error correction which comes with

26:03

overhead so you have something like

26:04

classically of redundancy triple

26:06

redundancy yourself stuff like that we

26:08

have that in the quantum world - but our

26:10

factors are larger they're not three

26:11

there may be seven and the most

26:13

optimistic case may be 10 to the 4 in a

26:15

more pessimistic cave so Microsoft has

26:17

set it aside on a particular approach

26:20

when we try to build a sort of a

26:22

self-correcting memory so classically

26:24

you know like hard disks

26:25

for instance they're pretty resilient

26:27

right if one of the magnetic magnetic

26:29

spins flips it flips back because the

26:31

neighboring spins just force it back so

26:34

there are similar things in the

26:36

topological insulators so structures

26:38

that have topological materials

26:40

semiconductors topological materials in

26:42

fractional quantum Hall effect

26:45

they have been observed they show

26:47

extreme resilience against actions that

26:51

the environment might do to your system

26:52

so in a sense it's very protected

26:55

against the environment on the flip side

26:58

it actually doesn't want to want you to

27:00

talk to the system anymore so the only

27:02

thing you can really do is some

27:03

topological moves you can braid it

27:05

around each other and hopefully by doing

27:07

so you can do operations so to answer

27:10

the question we have set our side on

27:12

that particular approach to build a

27:13

topological quantum computer the

27:15

experiments running right now in

27:17

Copenhagen and Delft the two

27:18

experimental groups that are now part of

27:20

Microsoft they try hard to build a small

27:22

computer at this point once we go beyond

27:25

a two qubit gate which will be a crucial

27:27

step in that roadmap there there seems

27:30

to be no limit like for a physics point

27:32

of view there's nothing that on the

27:34

theory side prevents us to build a

27:36

large-scale computer that needs only

27:37

very little error correction and then

27:40

when we when we scale up that machine we

27:42

could reach practically quantum computer

27:44

quite quite quickly and have to worry

27:46

about actually compiling into that that

27:48

architecture so that's kind of

27:50

engineering challenges and it's chic and

27:52

very for Microsoft Challenge sure so I

27:55

suppose that's what we really have time

27:57

for on this so I've got a couple of

27:59

minutes for questions if people want to

28:00

throw their hands up before somebody

28:03

cuts me off so yeah I've heard it said a

28:09

number of times that the basis for

28:11

Microsoft's quantum computers these

28:13

quasiparticles haven't been proven to

28:16

exist and yet I've also heard it be said

28:18

that there is no explanation other than

28:20

these particles exist for why these

28:23

machines are doing what's happening

28:25

right now so I was wondering if we can

28:27

sort of officially get to the resolution

28:30

of this question right here today is it

28:33

true that the Microsoft machine

28:35

essentially utilizes an effect that

28:37

there is no

28:38

explanation other than you have these

28:40

quality particles he seems like it it

28:42

was first observed in 2012 by Leo

28:44

københavn and in Delft back then there

28:46

was a big debate and there were several

28:48

people the papers written that had stuff

28:49

like smoking gun evidence yes or no in

28:52

the title I think by now there's so many

28:54

signatures of this these so-called

28:56

Mariana the remotes that have been

28:57

observed that people kind of settle on

28:59

yes this is a zero mode this can be used

29:02

to encode a qubit the questions are now

29:04

of a different nature the questions are

29:05

like can we actually do it in a

29:07

meaningful way so that we can operate on

29:09

it can we actually couple several qubits

29:12

together so that's kind of where we move

29:13

to at this stage and there is a white

29:15

paper out there's on the archive that

29:16

lays out a particular architecture how

29:19

we can scale but kind of on the

29:21

engineering side has several really

29:23

tough problems that that we have to

29:25

solve but I think I think it's pretty

29:27

much settled that we have observed in my

29:29

Arana zero modes yes I should do what

29:31

Allan did and get all the condensed

29:33

matter theorists and experimentalists in

29:34

the room to stand up because there's no

29:36

condensed matter theory yes per

29:38

Mentalist on this panel take it with

29:39

some rocks yes rain of salt any other

29:42

questions you spoke about this idea of

29:49

classical emulators of quantum computers

29:53

so could you tell what is a state of the

29:55

art how many qubits can be emulated

29:57

presently with some computers

30:03

thanks it depends on the computation so

30:09

for yeah I mean it depends on the depth

30:12

it depends on the layout and the actual

30:14

specific thing that you're trying to do

30:17

for the quantum supremacy stuff which

30:21

Alan alluded to earlier which I've

30:25

worked on with Google I know that the

30:27

best classical simulators can

30:30

specifically for those classes of

30:32

problems get to around 45 ish qubits and

30:38

there are real problems and going too

30:41

much further than that because of RAM

30:43

limitations and other things but for

30:47

different classes of problems you can go

30:49

a lot further like you get it you can

30:50

see

30:51

a little more cubits but they're the

30:52

nature paper from IBM about factoring an

30:55

arbitrarily large number with two cubits

30:59

quick advertisement for q-branch

31:02

we have developed an emulator that looks

31:06

at a particular hardware system that has

31:08

been deployed on premises at the

31:10

Commonwealth Bank of Australia the

31:13

purpose for that is to get the emulator

31:15

into the hands of their developers

31:16

inside the bank so that they can start

31:18

to just broaden their awareness and

31:20

understanding of how quantum computers

31:22

work and what the applications might be

31:23

that particular emulator does at least

31:26

30 cubits but again it's very dependent

31:28

on the problem and will continue to push

31:30

that state-of-the-art as we go forward

31:32

okay I've got Jason standing up can I

31:38

can I have one more quick question for

31:48

Matt because you brought up that you

31:49

guys are looking at potential financial

31:51

applications is that just because DE

31:53

Shaw is an investor or I mean what do

31:54

you guys see as you know are the big

31:56

algorithmic capital allocators investing

31:59

in this or yeah the big reason and we

32:02

can say this on a comparative basis

32:06

algorithmic traders principal traders

32:09

have the ability to be really really

32:11

supple in terms of kind of being a

32:15

solution looking for a problem whereas

32:17

in most cases you need to identify a

32:21

problem and find a solution to it so

32:23

again there's this creativity thing that

32:25

that quantitative traders look for to

32:28

try to get an edge and you can look at

32:32

problems both that would be run

32:34

efficiently on Anil errs and circuit

32:36

model machines that could be translated

32:38

into trading strategies and so I do

32:42

actually think that there's an argument

32:44

to be made forget about the underlying

32:46

math and whether you can mathematically

32:49

argue that there's a particular

32:50

advantage to a particular industry for

32:53

quantum computing but but I think that

32:56

the sort of this psychology of

32:58

principled traders and their their sort

33:00

of creativity and and flexibility

33:03

some like an obvious an obvious place to

33:07

explore implementing this technology

33:10

yeah thanks well thanks again and please

33:13

round up up holes for a la palace

33:15

[Applause]

English (auto-generated)