November 19, 2025 • 53 mins
Malcolm Gladwell heads to San Francisco Tech Week to talk with IBM’s new Director of Research Jay Gambetta in front of a live audience. They discuss IBM’s plans to scale quantum computing power, the groundbreaking experiments already underway, and what impact these new computers could have on chemistry, medicine, and even finance.
This is a paid advertisement from IBM. The conversations on this podcast don't necessarily represent IBM's positions, strategies or opinions.
Visit us at https://www.ibm.com/think/podcasts/smart-talks
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Hey everyone, it's Robert and Joe here. Today we've got
something a little bit different to share with you. It's
a new season of the Smart Talks with IBM podcast series.
Speaker 2 (00:09):
This season on Smart Talks with IBM, Malcolm Gladwell is back,
and this time he's taking the show on the road.
Malcolm is stepping outside the studio to explore how IBM
clients are using artificial intelligence to solve real world challenges
and transform the way they do business.
Speaker 1 (00:25):
From accelerating scientific breakthroughs to reimagining education. It's a fresh
look at innovation in action, where big ideas meet cutting
edge solutions.
Speaker 2 (00:34):
You'll hear from industry leaders, creative thinkers, and of course
Malcolm Gladwell himself as he guides you through each story.
Speaker 1 (00:41):
New episodes of Smart Talks with IBM drop every month
on the iHeartRadio app, Apple Podcasts, or wherever you get
your podcasts. Learn more at IBM dot com slash smart Talks.
This is a paid advertisement from IBM.
Speaker 3 (00:56):
Hello, this is Malcolm Gladwell and you're listening to Smart
Talks with IBM. Every year, Tech Week brings thousands of
people together to network and learn about what's emerging across
the technology ecosystem, and at this year's conference in San Francisco,
I had an amazing opportunity to sit down in front
of a live audience with Jay Gambetta. Jay has been
(01:16):
with IBM for years and was recently promoted to Director
of Research. In this job, Jay has an important mission
helping the company build the future of computing. In the
last episode of smart Talks, I began to learn about
quantum computing from IBM Chairman and CEO Arvind Krishna. But
this conversation I had with Jay went even deeper and
(01:38):
convinced me that the development of quantum isn't just a fun,
exciting new paradigm of computing, it may be one of
the most important scientific achievements of my lifetime. Jay, Good morning, morning,
Welcome to Smart Talks with IBM. Thank you special live
(01:59):
recording here for tech Week and congratulations. How long have
you been Head of Research at IBM?
Speaker 4 (02:05):
Since October one? It's October tenth today, since nine days,
nine days.
Speaker 3 (02:10):
Can you just talk a little about the position. This
is one of the most important positions in research in
the world.
Speaker 4 (02:18):
IBM research has been around for eighty years and it's
done some tremendous technology, a lot of inventions and fundamentals
for semiconductors, algorithms, AI. Yeah, I think if we look
back to where a lot of the innovation and the
technology of the world comes from, I think you can
find Ibram's footprints on it, and you can find IBM research.
(02:39):
So yeah, I'm very excited for the opportunity, but I'm
also aware that there's big shoes to fill, and I'm
looking forward to how we take IBM research forward. Obviously,
I'm going to be bringing a lot of the quantum side,
which we're going to talk about later. Beyond quantum, there's
important work that needs to happen in AI hybrid cloud,
(03:01):
and I think we're going to also enter into this
new period of mathematics where we get to use quantum
machines and also AI machines. And there's some really good
hard mathematical questions to answer.
Speaker 3 (03:13):
How many people do you have working for you?
Speaker 4 (03:15):
I've been researchers in the three thousand researchers across many
different labs around the world. Our main lab is in Yorktown,
but then we have the lab actually out on the
West coast in Armadan or SBL Now and then we
have one in Zurich, Japan, and a few others around
the world.
Speaker 3 (03:32):
Tell me a little bit before we get into quantum.
I'm just curious about your path. So you're Australian. Yep,
we were talking about earlier backstage. Your accent has become muted.
You should crank it up because it's.
Speaker 4 (03:45):
Yeah, I'm slowly losing my Australian accent. I've been in
the US since two thousand and four, so accent, you know,
to sound very Australian. Yeah, but how do you practice it?
Maybe I got to go back to Australia. Here a
more Australians say gooday, how's it going?
Speaker 2 (04:00):
Like that?
Speaker 3 (04:01):
And you you didn't grow up thinking you're going to
be a scientist one day, now.
Speaker 4 (04:05):
I grew up in a pretty normal life. My dreams
as a kid was building things, so I was either
going to be a carpenter or a mechanic. But I
had some great teachers that inspired me to go to university.
And I didn't even know, honestly what a scientist was.
And then I found myself at university doing science, particular physics,
(04:26):
and I ended up loving it.
Speaker 3 (04:27):
So you go from there to what do you do
your PhD.
Speaker 4 (04:31):
So I did my undergrad in Australia. I did it
actually in laser science, so I think I watched some
TV show in lasers seemed interesting, so I wanted to
learn about lasers. And then I realized in trying to
understand lasers there was this quantum mechanics, and so I
was like, all right, I want to actually understand this
(04:53):
quantum mechanics. So I did my equivalent of what you
and the US school masters. We call it honors in Australia,
but we do a research project. I said, I wanted
to shoot lasers into atoms and measure cross sections and
I got really into quantum physics. So then I decided,
all right, I don't understand this quantum physics. I want
to do my PhD in interpretations of quantum mechanics. So
(05:16):
I jumped in and said, all right, what is this
quantum mechanics? Why is everyone arguing on these different interpretations.
Then I finished my PhD in Australia doing that. Then
I moved over at the end of my PhD interpretations,
it's more people arguing about the equations whilst I think
it's really important. I decided if it's going to be
(05:37):
like a collapse equation versus many worlds, or a hidden
variable model, or that just quantum mechanics decoheres because we
don't see supersitions in the everyday world, because it interacts
with environment. The only way to answer that question was
to build a quantum computer. And so then I decided
at the end of my PhD, I wanted to work
(05:59):
out how to build a quant computer. And then I
left there and I went to Yale, and then at Yale,
that's where I got into superconducting cubits, which just a
few days ago one of the professors there just won
the Nobel Prize this year.
Speaker 3 (06:12):
Oh wow, Yeah, I'm very interested in tracing because your
career follows the arc of quantum computing in a certain way.
Right at the time when you asked the question, what
I really want to do is to figure out how
to build a quantum computer. Where are we in quantum
computing at that point?
Speaker 4 (06:30):
Yeah, So that would have been nineteen ninety So there
was Shaw's algorithm came out, let's say ninety five. There
was a lot of theory. And then the reason I
went to Yale is because people had started to show
that they could see quantum effects in electrical circuits. So
these macroscopic objects they were starting to behave quantum mechanical
(06:52):
There was a really significant breakthrough in nineteen ninety nine
where Yazoo Nakamura in Japan showed that a qubit could
exist in these electrical circuits. And then I found out
the group at Yale were really trying to take these
electrical circuits and couple them together. And so it was like,
if I can build something using electrical circuits and they're big,
(07:15):
that that's the best way that you cancide to test
and understand whether quantum mechanics breaks down at a macroscopic
scale or not. Can we actually make them behave as cubits?
And I agree. When I came to Yale, the cubits
were not very good. They were actually a couple of nanoseconds.
They were unstable. Electron would jump onto the chip and
(07:37):
then they would change all their configurations, so you have
to restart your experiment. And so for the first time
at Yale, it's kind of what the challenge there was,
how do we make a cubit? How do we make
a stable cubit? And that took about five years and
that took us up to two thousand and seven. And
I think the rest of the world looks and says
quantums like just blowing up. But it's actually been like
(08:00):
all most phases theory, showing that we got the algorithms,
how do we make a cubit? How do we couple
of the cubits together? And now we're in the scaling phase.
Speaker 3 (08:09):
Describe for us, because many people in this room, me included,
have only a kind of surface level understanding of what
we mean when we use that phrase. What is the
difference between classical computing and quantum computing? What does that
word mean?
Speaker 4 (08:25):
Yeah, so you can go down the physics way and
talk about supersition and entanglement, which we can go in later,
but actually feel it's a bit of a distraction. So
when you think of classical computers, what they were is
there were machines that were very good at adding numbers together,
like simple addition, and they really showed that they could
(08:47):
add these numbers together really really fast. And now with
GPUs and other AI accelerators, we can add those numbers
together in parallel, and so the whole classical computing can
come down to just arithmetic, just adding numbers together. It
turns out that there's a math that is the quantum
mechanics shown.
Speaker 3 (09:07):
To be true.
Speaker 4 (09:08):
It's more like a group theory type structure. And the
way quantum works is it has a different mathos are primitive,
and if we can exploit that new math and build
a machine that does it, it allows us to answer
different questions. And so think of it as a branching
from classical compute that is very good at adding just
numbers together to something that allows us to work with
(09:30):
an algebra that is much much harder to represent with addition.
And that algebra happens to be the same algebra that
defines the fundamental equations of nature shirting as equation. So
this is why you say it computes the same way
nature does. But there are many other interesting problems. So
the way I explain it to people is think of
(09:51):
it as bringing a new primitive to computer science and
allowing us to work how to go with it. And
I like the analogy. Well, actually, maybe go back. So
if you went back in time, so we're one hundred
years of quantum, and you went back in time and
you asked, what is the foundation? Is a chemistry or physics.
(10:12):
What would have probably the scientists of one hundred years
ago would have said is they would have said, you know,
chemistry is about the small, physics is about planets and
things like this. And one hundred years ago when Heisenberg
or Einstein, all the greats, Schrodinger himself invented quantum mechanics,
it was this concept that nature is discreete not continuous.
(10:34):
It actually brought all the physical sciences together. And now
quantum mechanics is like it is the foundation of the science.
And so now what quantum computing is by that analogy
is computer science. The foundation of the math is coming
together with the physical science to allow us to compute
using math that if you were to try to represent
(10:55):
it with classical computers, it takes exponential time.
Speaker 3 (10:59):
Yeah, it was a classical computer and an expense in
a way that someone is well informed as I am
can understand it. A customer computer works primarily on problems
that can be easily represented in numerical form in numbers. Yes,
quantum allows you to step outside to a class of
problems that don't necessarily have a simple numerical representation.
Speaker 4 (11:20):
Yeah, and so imagine I got some medicine or or
some set of operation, but call it A and I
then follow it by a different operation B if A
followed by B gave a different answer than B first
followed by A. So in mathematics we call that commuting.
But like you can think of a correlation there one
(11:41):
one gives you a different outcome to the other. That
means there's an algebra behind it. That Representing that algebra
traditionally on classical computers is really really hard, whereas that algebra,
if we can get creative, we can come up with
ways of representing that math. So we step as you say,
we step out aside of the simple math to a
(12:02):
new math to allow us to calculate interesting problems.
Speaker 3 (12:06):
So quite in a sense, compliments it doesn't replace judicial.
That's good.
Speaker 4 (12:11):
I think this is one of the this is you're
exactly on is people think quantum is going to be
replacing classical If your problem is good at adding numbers together,
you should just keep using classical computers. I think the
future is going to be heterogeneous accelerators, and it will
definitely have quantum as one. But in some sense, the
next generation of superstars are going to be those applied
(12:32):
mathematicians that know, how do I write a problem using
the simple math of classical computers or the more complicated
math for quantum computers, and how do I actually iterate
between them? And things like this? This is where I
think the next generation of students are going to come
up with much more novel ideas. I can give you
examples of what we want to do on quantum, but like,
(12:54):
you're giving them a fundamental, foundational new thing, and so
I'm optimistic that will do much better jobs than my generation.
Speaker 3 (13:03):
Well, yeah, we're to get to some of the albums
in a moment, but I wanted you to the most
kind of down to you said, as a kid, you
thought you might want to be a mechanic because you'd
like to build things. Describe to me what it takes
to build a quantum computer, Like, what are you doing
that's different from building a classical computer.
Speaker 4 (13:22):
Yeah, so maybe I'll give you analogy and then i'll
go in so the way classical computers, we've got them
to get to smaller and smaller sizes like five seven animeters,
five animeters and things is actually inventing material to kill
quantum effects, so you actually put dielectrics and other things
(13:43):
in there. To kill the quantum tunneling effects, and you
want them to behave more classically. In the quantum world,
you want to get rid of all the classical effects,
So you want to get rid of the ability of
the cubits to interact with the environment. And in the
in the sort of technical world, we call it this
quantum conflict. The more ways you want to control the
(14:05):
quantum computer, you open it up to interacting with everything else,
like interacting with its environment. So the biggest challenge has
always been how do we give more control but don't
bring in other sources of noise. So I want to
be able to do gates on the cubit, but I
don't want it to decohere. I want to couple the cubits,
(14:28):
but I don't want them to couple to other things.
So the hardest challenge is the energy inside the cubits
is a nine gigahertz, and if your tames that by
HBO tend to the neggave thirty four with nine, you're
at a tender the negative twenty like three or something
in energy. That's a tiny amount of energy. So you're
(14:48):
trying to have a tiny, tiny amount of energy to control,
and you don't want that to interact with anything, So
you have to cool them down, you have to isolate them,
and you have to make the quantum effects dominate over
the classical effects.
Speaker 3 (15:03):
So practically, if I'm trying to do that right now,
how big are these machines.
Speaker 4 (15:08):
So the cubits themselves are not that big. So the
cubits themselves are like a few microns. But yeah, most
of the size so you can see some of our
I got the pleasure of showing you around to one
of the machines in Yorktown. You saw that they're like
twenty foot by twenty foot in size. Most of that
is all that equipment to isolate the cubit chip, which
(15:30):
is only a few millimeters when you put it together
from the rest of the environment. We will, as we
get better at that, miniaturize all the isolation. But that's
cooling it down to a few milli calvin, so about
a thousand times colder than outer space. It's isolating the
noise on any electrical signal so that no noise from
(15:52):
the outside world gets into the system. And so that's
a lot of isolators, filters, and things like that that
we've had to invent and to allow us to make
the quantum properties of this chip go.
Speaker 3 (16:04):
It's like the Princess and the pe mounds and mounds
and mounds of mattresses trying to isolate the impact of
this little thing, and that maybe.
Speaker 4 (16:12):
That's the best way to describe it. Yeah, and you've
got to keep it really really prestige.
Speaker 3 (16:17):
But that when you show me. So in the in
the lobby of the Watson Research Center in New Yorktown,
which by the way, is just the coolest building. It's
like a it's like a modernist it's also master piece. Anyway,
in the lobby there's there are these is it two machines.
Speaker 4 (16:34):
It's it's inside a container that has three machines machines.
Speaker 3 (16:39):
So what can you can you tell me what would
one of those machines cost to build right now?
Speaker 4 (16:44):
So typically we put them together in a way where
we upgrade them because we want to as I as
I was talking about before, one of the things we
want to do is always get algorithms done on our machines.
And I've got a roadmap of build bigger and bigger machines.
So usually one of those quantum processes today is out
(17:05):
of date in six months so we want to build
this future of computing that leverages quantum computing, where every
six months we've outdated a quantum processor. Eventually, hopefully we
get to a point where it's like stable and it
can be many years operating. But we want to get
(17:25):
as large a quantum computer in the hands of people
to explore the math as possible to come up with
those new algorithms. So we've had a philosophy of having
them open, working with universities and things like that. So
to answer a question of costs, yes, there's cost in
building the system, but we are operating in them much
more in a service model where people pay to use
(17:46):
the machine because we have to continuously calibrate it and
operate it and so depending on various different things. Professors,
we have a credits program where they get free access
some universities and animal prizes. They can buy premium access
and get more access. So think of not like a
cost of it, because it's almost like a continuum. I
(18:08):
want to make sure that the best quantum processors that
I can build get in the hands of students and
professors and interested enterprises that want to explore these machines
as fast as possible, and typically every six months we
upgrade it.
Speaker 3 (18:22):
Yeah, you don't start over, you upgrade.
Speaker 4 (18:25):
We upgrade various different pieces, the processor, the electronics. Some
upgrades are just simply replaced the processor. But as an example,
I think many people have probably seen photos of quantum
computers and you see this scary thing with all these
wires hanging down, as I've referred to as the chandelier,
and it's got all these wires with loops and things
(18:47):
like that. They're called co x keebles. When we first
put the quantum computer on the cloud in twenty sixteen,
you could probably only fit about fifty cubits inside one cryostat.
We've had to upgrade all those cables so that we
can fit around one thousand. I want to get to
three thousand, and that's about miniaturizing it. So to answer
(19:07):
your question, an upgrade, it depends. It can be either
just the processor or it can be the complete insides.
And we're actually in our third generation of our electronics
to control the systems, to make them faster, less noise. Internally,
we've got exciting results of going to something like cold cryocemos,
so you can bring down the cost in terms of
(19:29):
energy of running these quantum computers almost negligible, and you
could imagine future quantum computers. I'm not going to require
much energy to run, so unlike classical compute that requires
lots of energy. The biggest machines that we envision is
only in the few megawatts. But we have to upgrade
to future controls that use less energy. So it depends.
(19:51):
It's my long answer, short answer to how it upgrades,
and it depends on what it is.
Speaker 3 (19:57):
The only observation that I felt I could with making
when you showed me the quantum machine is it's gorgeous.
I look at art.
Speaker 4 (20:05):
I've always believed that, and I think that there's an
IBM saying good design is good business. But we've always
taken pride in making sure what we build. I don't know.
I feel if you're going to build something that is
new that can change, you should take the time to
make sure it looks and feels good.
Speaker 3 (20:26):
Will you donated to MoMA when you're through with that
particular Actually, I think we just put.
Speaker 4 (20:32):
An old version of one of our insights with the
United Airlines and the AAPS, which is the American Physical Society,
and the University of Chicago. There's a replica right now.
If you fly into one of the terminals in Chicago,
you can walk and see one.
Speaker 3 (20:48):
Oh really yeah, well the most advanced thing at all air.
Speaker 4 (20:51):
I'm sure probably, but yeah, I hopefully, I think, yeah,
we're open to that. But yeah, I appreciate that you
love the design. It was beautiful.
Speaker 3 (21:00):
So I last week I interviewed for another episode of
Smart TALX your CEO, Ivin Kushner, And when we got
to the quantum question, I mean, he's always alliant and brilliant,
but quantum, he's like lit up. I mean right in
thinking that IBM is much more invested in quantum than
(21:22):
anybody else. Is that a fair statement? Oh yeah, most definitely.
Why Why did IBM choose to kind of make this
such a priority.
Speaker 4 (21:29):
So when I took to the history of the physics side,
there's this interesting thing in the history of computing. So
we build computer classical computers today using bits and sea moss,
and they consume energy. Do you know that there is
a way in classical where you can actually compute without
using energy. It's called reversal computing. Turns out to be
(21:51):
a terrible idea. It's not practical to build. But IBM
investigated that with Ralph Laura and Charlie ban It early
on and they proved the concept that reversible computing. The
first use of quantum information theory, one of the first
actually was from IBM. When I did my PhD, I
(22:12):
remember actually picking up this paper on quantum teleportation and
seeing IBM written there, and at the time I remember
thinking that they make PCs. Well, what the hell are
they doing this foundational paper on quantum teleportation? Why are
they doing it? So to answer your question, actually, IBM
was the first in quantum information science because it's the
(22:33):
fundamental of computation. Can we actually come up with compute
that we can go beyond the classical So way before
anyone was talking about it, they were doing fundamental theory.
And then as we've built it, we've always When I
first came there, the experimental team was small. In twenty eleven,
we've had a small team that we're focusing on single
(22:57):
cubitts coupling them. I think in two thy twelve was
the first time we showed really good two Cuba gates
and no one was talking about quantum computing then. And
then I remember in about twenty sixteen, I said to
actually Arvin was the director of research, then can we
actually put our quantum computer on the cloud? Well that's
(23:20):
probably twenty fifteen, and it was always supporting that. So
as we've done more and more we've been able to
do it. It's had this program going now, I agree
is very visible, like because we're in this scaling phase
and so we're invested to keep scaling it and to
get why is At IBM research, what we always do
(23:43):
is answer what is the future of computing? Whether it's
coming up with new algorithms, coming up with better AI,
coming up with quantum, or coming up with just how
do different accelerators go together. It's our DNA to answer
the question of what is the future?
Speaker 3 (23:59):
Need a perfect problem for IBM because you kind of
need to have a legacy of building stuff, building actual
physical machines.
Speaker 4 (24:07):
Yeah, that's why I came to IBM. I wanted the experience,
the culture of building hard things that others have not
done before.
Speaker 3 (24:19):
Where do you imagine we are in the timeline of
this technology? It will come a point when it will mature.
My cell phone is a mature technology at this point.
How far are we from that point? With condom.
Speaker 4 (24:32):
So I think there's various aspects of it. So we
sat in twenty and seventy we set our goal that
in twenty twenty three we would be able to build
a machine that was beyond classical computers to simulate it.
And we achieved that in twenty twenty three. So to
run a biggo we call it a quantum circuit. The
details of it don't matter, but to run a quantum
(24:54):
workload that if you were to simulate that workload how
a quantum computer operates, it's on a classical computer, you
couldn't do it. So we said that as our first
and now I've made it publicly that by twenty twenty
nine we'll build the first fault tolerant quantum computer. That is,
one that can completely handle the noise to the level
(25:16):
to allow you to run a very very large, large problem.
Speaker 3 (25:19):
So an example of a large problem.
Speaker 4 (25:21):
Yeah, a large quantum problem. So for around a couple
of one hundred cubits and one hundred million operations, you're
talking still interesting science problems like simulating a molecule, or
calculating a small optimization problem, or calculating, say some part
(25:42):
of a matrix update. In some type of differential so
it'll still be scientific, but it'll be at the point
where it's beyond, well beyond any classical approximate method. And
then I think that's twenty twenty nine. That's twenty twenty nine,
So we're four.
Speaker 3 (25:58):
Years away from something that can start to handle.
Speaker 4 (26:01):
Interesting problem, serious problems. I do believe the scientists will
find interesting heuristic problems before that, And so over the
next four years, you're going to continue to see more
and more let's call them heuristic not provable quantum problems
that run on quantum computers that come out. We're see
more and more come from many of our partners and ourselves.
(26:22):
Heuristic problems have value, but they have to be tested,
they have to stand up over time. You have to
run them many, many times, you have to try different ones,
and many times heuristic can lead to formal problems. So
you're going to see, because we're beyond now the point
that you can simulate these quantum computers with any classical computer.
They're kind of like a scientific tool. So they're exploring
(26:43):
the heuristic.
Speaker 3 (26:44):
What do you have to get done between now and
twenty twenty nine to get there?
Speaker 4 (26:47):
So we had to reinvent how we wanted to do
error correction. So we have to demonstrate modules and if
we can demonstrate these error corrected module and our goal
is actually it's called Kooko, but I name all that
chips after birds, so it's called Kuoko. Borrow is named
after an Australian vert. I think I still say Cuoko
Borrow the way Australians do. We need to then show
(27:10):
that we can make a single module and then we
want to connect two of those modules together, and I
call that one Cockatoo, which is another Australian vert. And
then if we can do that, so that's twenty six
and twenty seven, and then we want to scale them
scale those modules, and that we call Starling, and we
want to scale that in twenty twenty nine. So get
(27:30):
a module, join two modules together, and scale and so
each module is going to be around one thousand cubits.
Speaker 3 (27:37):
The challenge to getting there is it finding the right
material or how would you describe what that's.
Speaker 4 (27:43):
The beauty to be done. That's the beauty of it
is if we would have been here two years ago,
I couldn't tell you how it would be done. So
we had a huge breakthrough we came up with a
new code, a new quantumeric Russian code, and that code.
The biggest in part of that code that is the
most important is it's modular in nature. So previous codes,
(28:07):
without getting too technical, they were very monolithic and you
had to build a very big device and I wouldn't
have known we would have to invent tools like new
simos tools to do that. So we came up with
this new code. We started on twenty nineteen, we published
in twenty twenty four. We kind of had most of
things worked out in twenty twenty three. That's why we
(28:28):
got confident to release the thing. So the biggest breakthrough
we had is coming up with a code that's modular
in nature. And think of that as a like a blueprint.
And so now we have the blueprint and now we're
doing engineering tasks to implement every part of that blueprint.
Speaker 3 (28:45):
And so the minute you had that breakthrough, then you
began to have confidence at something exactly these goals could
be met.
Speaker 4 (28:51):
And then you can't. And then anyone that's done engineering
will know what I'm talking about when I say this
is cycles are learning. It takes so long from test
idea to build two tests in hardware, the cycles of
learning are much much lower than software, Like you can
be really, really faster in the software. So then we've
(29:11):
planned out our iterations over the next few years, and
so we have to successfully demonstrate them. I may slip
because sometimes you may estimate your time wrong, but we
now have exactly what we want to do for the
next four years.
Speaker 3 (29:28):
I want to go back to that breakthrough for a moment.
What does the word breaks we mean in that context, Like,
it's not that you get a call in the morning
from somebody who says.
Speaker 2 (29:36):
I did it?
Speaker 3 (29:37):
Do you see it coming? Or is it a surprise
when they get there.
Speaker 4 (29:40):
So the way this one worked is Sogo Brave, who's
an algorithm person at IBM, one of the smartest and
quantum information.
Speaker 3 (29:48):
Don't mention his name to everyone. You'll come for him.
Speaker 4 (29:52):
Everyone in quantum already knows his name. I don't think
there's an idea that has not originated from him in quantum.
So we're looking at other codes and we'll go all right,
we've got to get serious about these codes. And others
were starting to propose to bring these and then we
call them LDPC codes from the classical space into the quantum,
(30:16):
and I asked him, we need to get ahead of
this and understand what they're doing it. He's like, the
most modest perfuse late, Jay, let me learn about them
and I'll generate a report for us and we'll read
through it. And then I said great. Then I don't know.
Six months later, he comes back with one hundred page
report on everyone everyone had done in LTPC codes. I'm like, awesome.
(30:38):
So I started then to read from them. And then
we said, all right, how do we under the assumptions
of the hardware we can build? Can we get an
LTPC code knowing what we can build? And that's a
great question, and so we put a small team together
to investigate and honestly took two to three years, and
(31:01):
we iterated and we used the constraints, so we had
the sort of theory, and then we had the constraints
of what we could build. And we iterated for a
few years, and then at the end of that we
came out with a solution that yes, it is possible
to meet all the constraints of the hardware and build
a code that will work.
Speaker 3 (31:22):
I'm just curious about So you had this task, this problem,
you want to solve. And when you set out on
the task of trying to solve the problem, what's your
certainty level that you'll get a solution.
Speaker 4 (31:34):
Well, that's the beauty of science for things where you
kind of have a few ideas. My philosophy is try
a few for the ones that need to be in
that like wow moment. It's honestly, you've got to set
the ambition really, really high, but know when to stop.
It was a great team that went together to get
(31:56):
that breakthrough, and we knew that we needed to come
up with a code that met the requirements of the experiment.
And I think what was different before then is the
theorists that were doing error correction codes didn't necessarily know
the constraints of experiments, so it was like really more
pen and paper. So this became one, all right, given
(32:18):
these sets of constraints, is it possible when LA's.
Speaker 3 (32:22):
Questions about this? Sorry, I love these kind of moments
when things become clear. At the time the problem was solved,
were you aware of the implications of the solution or
did that takes you knew exactly what.
Speaker 4 (32:35):
We set out exactly like either we were going to
have to work out how to cool down a very
large piece of silicon, which would require a lot of
engineering and building tools beyond what anyone has ever built
in the silicon semoss industry to implement the known codes,
or we had to come up with a different one,
(32:56):
and once I knew that we had one, that I
didn't need to re invent any tools to build. The
implications are clear how.
Speaker 3 (33:04):
Much time elapsed between the time you heard the problem
was solved and the time you told Arvin Krishna, the CEO,
the problem was solved.
Speaker 4 (33:13):
I'm sure the next time I spoke to him, I update,
but I don't remember. The beauty of Avin is he
trusts the scientists will do it, and so he doesn't
really check on us. We update him when it is,
and he he empowers us to do really hard problems.
Speaker 3 (33:27):
Yeah, so let's talk about uses. I mean, they're really
like cool, big shiny machine. I think you'll get pay
twenty twenty nine. But there's all kinds of really interesting
problems you're already working on.
Speaker 4 (33:40):
Yes, this is like another interesting area is I can
prove in pen and paper algorithms that we want to
run that. Like, it's not that we don't know what
to do with a quantum computer, there are hundreds of algorithms.
So you can go to I think it's called quantumzoo
dot com and you can see many many algorithms. People
are coming up with more of more of them that
(34:02):
they prove by pen and paper. Imagine, now we have
a machine that you can't simulate. How do you actually
discover algorithms in a scientific way? How do you look
and discover algorithms using a quantum computer. We're in this
exciting period right now, and so even though I can
(34:24):
prove these ones that we can run in the future,
there's a big white space between what the machines we
have and we're going to build and continue to do
and those ones that want the provable ones. And I'm
an optimistic person by nature. I think getting those machines
in the hands of students to explore and look at
(34:45):
heuristic algorithms, so looking at the equivalent of doing numerical
algorithms on computers, which there's many histories of numerical algorithms
being discovered on classical computers before we had formal pre
that we rely on today. People would even argue the
way AI works was driven numerically, even though we have
(35:07):
input into it. There are ones in optimization driven numerically.
We are entering that phase. So the computer scientists now
need to go play with these primitives. Our prediction is
over the next couple of years we're going to see
valuable numerical equivalent algorithms emerge. And where the scientists are
(35:28):
going is in four categories. One is simulating nature, so
looking at either Haanji physics, chemistry, light problems. As an example,
with our partners in Japan, they took one of our
quantum computers and for Gackle, a very large classical supercomputer,
and they ran a problem where quantum was just a
(35:50):
sub routine of the problem that was running on all
of for garcule, and they were able to look at
an interesting molecule, a molecule that if you would go
by pen and paper you would have said to take
me a very long time to run that. They were
able to run that quite accurately, heuristically, and already get
results that are comparable with the best classical methods. So
they are extremely excited because they want to push that
(36:11):
further and they're sort of showing that you can take
a classical supercomputer with quantum as a subroutine and start
to push the level they were.
Speaker 3 (36:19):
This was trying to solve a medical problem.
Speaker 4 (36:22):
Is this one is a like most people don't realize,
Like iron sulfur, just something as simple as iron and sulfur,
we can't solve that exactly, Like iron sulfur, molecules are
too hard, so really small small molecules are really really hard,
too hard for classical computers to solve. People think we
can solve a lot of things. It actually turns out
(36:44):
we can't solve very much.
Speaker 3 (36:45):
You say solve its instance, you know precisely how that
molecule works and it's constructed.
Speaker 4 (36:50):
No precisely what the energy levels of that molecule is
and how they come together, and then be able to
do that on a classical computer and compare it talk.
Speaker 3 (37:00):
It would be really really useful to know that specifically, because.
Speaker 4 (37:04):
If you can have energy levels, then you can estimate
reaction rates. If you can estimate reaction rates, you can
see how different types of chemicals will react. That can
then lead to better informing eventually how to build materials
or even drug design. I just want to be careful
and not say, oh, we're going to solve drug design
or that because there's many scientific steps to make that so.
(37:28):
And so what quantum gives you as a different tool
to give you more accuracy and then lead to making
the different methods work.
Speaker 3 (37:36):
You can subcontract out aspects of a problem quantum right now,
and that just gets you further along than you would
have been.
Speaker 4 (37:44):
So at the moment, even this result still does not
beat the best approximate classical method. It's comparable. So the
art of chemistry for the last hundred years has been
about approximating. So what we've done is we have got
very very good at coming up with ways of approximating nature.
(38:06):
And a lot of the things that we do and
we exploit and we use to estimate approximations. They don't
a stimulate nature of the way nature is. They approximate it.
And there's I could list many different acronyms of different
methods that go into approximating nature. What quantum gives us
is to eventually get beyond that approximation and do it
(38:28):
the way nature works. And so we aren't beating those
approximation methods, and this is why I think, this is
why it's still in the science. But they're getting comparable,
getting comparable with a new tool where the previous tool
is a dead end makes scientists very excited. Yeah, that
nuance is where it is, and so that's in machine learning.
Sorry Hamiltonian. Then there's examples in differential equations, So can
(38:52):
I actually come up with differential equations and solve them?
And if I can solve them, you could look at
things like an obvious Stokes goes into weather. There's financial
differential equations that you can better predict. So differential equations,
there's many different examples there. And then I would say
that two others are optimization and then there's quantum versions
of machine learning that are very exciting as well.
Speaker 3 (39:16):
Cleveland Clinic one of the organizations that you guys have
worked with. Why would the Cleveland Clinic be calling you up?
Speaker 4 (39:22):
Because that problem that they want to look at. So
they've also done similar problem to the recent lab. So
they've taken that method now and they've looked at molecules
that matter for drug design. So they're fundamentally looking at
those molecules that matter for eventually replacing some of the steps.
(39:42):
So they're investing to see how reliable it can be done.
And so there's a scientist there that's done many iterations
now using the techniques that were done first with the
team in Japan, they've now replicated that for new molecules
that are essential primitives for eventually designed drugs and things
that may matter for medical Yeah.
Speaker 3 (40:03):
And also there's some finance firms yep, HBC, Van Good yep,
and their interest is.
Speaker 4 (40:09):
What so that was the differential equation and optimization. So
if you are doing very large calculations like risk portfolio,
or if you want to model the black Shaw's equation
or things like this that are fundamental for them to
make better predictions, come up with better trades and things
like this. That is a very hard computational task. And
(40:31):
so rather than quantum replacing that whole problem, can quantum
be a subroutine in there? And what HSBC showed is
they showed they could take their real data, they could
take their real classical method and they just replaced a
tiny part of it. They replaced a tiny part of
it with a quantum subroutine that allowed them to come
up with better predictions of the weights that then when
(40:54):
they were to compare trial A versus Trial B, it
was thirty four percent better at algorithmic tun and that's
a big deal for them.
Speaker 3 (41:03):
It's huge.
Speaker 4 (41:04):
Yes, Now do they need to do more trials? Do
they need to see is this a heuristic algorithm? Do
we need to be careful? Is there other classical algorithms
that go into these are great questions that are now
being investigated. So think of this period of heuristic algorithms
is really a period of scientific discovery using these machines,
(41:25):
knowing that we want to continue and build the ones
which have determinist their algorithms that can run.
Speaker 3 (41:33):
Do the people who would profit the most space starting
to run quantum experiments realize that they would profit so
much from running quantum experience And does the world know this?
You've given us a couple of specific examples, but generally speaking,
there must be a very large universe of people who
could gain from at least starting to play in the space.
Speaker 4 (41:55):
So the enterprises that use computation as key for their
understand the limits of classical computation and they're very interested
to get started. The universities are very interested. Could we
get more students doing more algorithms one hundred percent. Some
of the limitations on the rate of algorithm discovery is
(42:17):
because people are thinking through the classical way of writing algorithms.
My belief is yes, So this is why we want
to get more and more students and things, because it's
just starting. But I would say in general most people
are aware of it. Could we get more, could we
accelerate it?
Speaker 3 (42:33):
Yes?
Speaker 4 (42:33):
Do we need to make better hardware, do we need
to come up with better libraries, yes? Do we need
better software yes, But it's all happening over the next
few years.
Speaker 3 (42:43):
Is it hard to get someone who's spent their entire
life thinking in terms of solving problems to classical means
to make the transition to this new paradigm.
Speaker 4 (42:51):
There's a lot of examples when you approach something with
the classical intuition, it's not the right way to do
it when you approach it through the quantum. But if
people are being taught to understand the fundamentals of the math,
then a lot of the techniques carry across. I don't
recommend people need to learn about entanglement or supersition because
(43:14):
whilst the physicists will argue like spooky action a distance
and all these type of things, entanglement is the power. Yes,
that's how physicists are labeled. How quantum is different. But
I would say, do we need some physicists really worrying
thinking about that? Yes, but we need more applied mathematicians
that are realizing they can use this as a as
(43:35):
a different way of looking at the problems.
Speaker 3 (43:36):
Yeah. And when I asked you one question, know is
we're describing a a It's more than a new technology.
We're talking about a new paradigm. It's a way of
thinking about problems. Can you compare this to kind of
previous technological paradigms. I'm thinking at the last couple hundred years,
what does this rank in terms of a new field
(43:58):
that we've opened up.
Speaker 4 (44:00):
It's a hard question to answer, but I often say
the history of computing, this will be the first time
that computation is branched between classical and quantum. I like
thinking reading a lot in the past. One of the
things that I think was a way we changed as
a society was the invention of zero. Before zero, math
(44:22):
was limited. Realizing that numbers have a number a zero
allowed us to develop a whole set of new mathematics
that then went on and defined like everything from waves
to calculus to all of that. Yes, we can describe
it with that same math, but when we describe it
with that math, it gets exponentially big and gets impractical
(44:44):
to do. Now we can actually work on it, I
would say, if I had to give you a quick answer,
maybe going all the way back to when we were accepted.
Speaker 3 (44:54):
Zero, I thought you were going to say, like the airplane,
but in fact, yeah, you went several orders of magnitude
on that.
Speaker 4 (45:00):
Yes, but but I think it's sort of fundamental.
Speaker 3 (45:04):
This is absolutely fascinating. Thank you so much for chatting
with me about it.
Speaker 4 (45:08):
Thank you for your time.
Speaker 3 (45:11):
Hey, listeners. So normally we end this episode here, but
the Tech Week attendees asked Jay some really great questions,
questions I wish I'd asked, so we wanted to include
those here. Enjoy.
Speaker 5 (45:24):
Hi, J, thank you so much for the great presentation.
My name is Trixie Apiado. I work for Willis Towers Watson,
an insurance broker. I help seisos identify and quantify their
cyber risk so they can prepare for threats before they happen,
and so quantum threats keep me up at night. You
mentioned so many good problems that quantum can solve. It
(45:46):
can also break encryptions in our classical computer systems. So
what safeguards or policies do you implement in your teams
to build quantum capabilities responsibly? And what can we do
for people in this room as builders and users to
secure our data in systems before quantum computers become more
(46:09):
energy efficient, cheaper, and more available.
Speaker 4 (46:12):
So it's a great question. So yes, one of the
algorithms for quantum computing is to break our traditional encryption.
So at IBM Research, we were aware of this from
day one. We've come up with algorithms that we believe
and have very strong evidence will not be broken by
a quantum or classical computer, and this has selected them.
(46:36):
So first the scientific technical question, security is saved. There
are algorithms that exist that we can implement that neither
a quantum or classical computer can break. So the technical
answer is we're all okay. The more complicated answer is
a social and society answer. Encryption was built in classical
(47:00):
computing in a way that was never thought of being upgraded.
It's mixed everywhere. Some of it is downstream, some of
it is like software that you may use. Some of
it is software that you've developed. And I get that
if you've got a product and you want to have
it secure for the next ten years, you probably want
(47:20):
to think about how you're going to upgrade it, or
if you have data that needs to be secure for
the next ten years, it needs to upgrade to new encryption.
So the real challenge is more of a social business
problem of how do we actually transition from old encryption
to new encryption knowing this is going to happen. So
(47:41):
we at IBM have been very proactive on this. We've
developed tools where we can determine where encryption is used,
We've developed tools which can show you how to replace it,
and we early on have made sure the Mainframe when
we made these algorithms. So I think it was Z
sixteen that was the first version of the Mainframe to
(48:01):
have these quantum safe algorithms implemented. So my answer to
your question is, yes, there's a real problem, but it's
not a technical problem. It's a social and business problem.
And I'm not minimizing that. I understand that that is
a lot of work you need to start now. You
need to come up and do a you need to
(48:22):
make it part of your IT transformation. You need to
get onto it. And I realize I realize it's not
going to take zero time because it's not an easy
problem to do. So the short answer is one we
developed algorithms that we can't and we're developing tools to
help you in that transformation.
Speaker 6 (48:41):
Thank you so much, Thank you. My name is Emma.
I'm a product manager at Expedia working on software side
of things. My question is around the non technical roles
outside of the researchers, the mathematicians, the builders. How can
the rest of us, whether it be policymakers, those in
the legal fields, those thinking about what use cases quantum
(49:02):
can solve for in the future, what should we be
thinking about and how can we prepare for that.
Speaker 4 (49:07):
It's a good question. I think this is part of
the requirement of the scientists to being able to articulate
where they are. We need a forum for those type
of discussions. I think a lot of this can fit
within the forums that we already have for classical and AI,
and I think we need to just be asking how
do we actually bring them into them. Because I don't
(49:28):
think of quantum as a replacement of compute. I think
of it as an accelerator that expands what is possible,
and I think we can ask those questions in those forums,
are we doing enough now? I think I agree with you. No,
I don't know the answer to it.
Speaker 6 (49:43):
I think it's a really interesting perspective because those existing
forums do start to bring in those other fields as well,
so it could warrant the same sort of discussion and.
Speaker 4 (49:54):
Active And I understand those forums right now AI is
probably dominating, and it should be like we are going
through a period of time where AI is impacting society.
The technology is impacting society in big ways. So I
totally understand that most of their focus should be on AI,
(50:15):
but we should start to ask where is quantum in
that as well?
Speaker 7 (50:19):
Hi, I'm Gobi and I'm a graduating PhD student at
Northwestern and also a member of south Park Commons, which
is a fund here. You mentioned earlier that some problems
are best solved by classical versus some problems are best
solved by quantum. When we're thinking about this, if we're
not experts in quantum, but we're thinking about this from
an AI perspective, could you just clarify when we think
(50:41):
about quantum, what is deterministic and what is not deterministic.
Speaker 4 (50:46):
I think the future of computing we've got to get
our heads around is that not everything is deterministic, and
it's much more going to be probilistic. How do you
handle error bars? How do you put confidence? I think
a lot of those questions which you're referring to in
AI are going to completely apply in quantum. I actually
think it's a mistake to compare AI verse quantum. I
(51:10):
actually think of quantum as much its quantum verse classical compute,
and AI is going to come across on top. So
as we go forward and we get a better understanding,
I'm not going to say quantum is going to replace
the classical compute that enables AI, but I think some
of the math you do in AI will be able
to go to both. So what can we formally prove?
(51:34):
I can come up with a problem where I take
a circle and I color half of it red and
half of it of blue, and then I say I'm going
to apply an operation that takes those dots make it. Say,
let's say ten dots over here red, ten dots over
here blue, and I'm going to wind them around many
many times. I can then show you that if you
(51:54):
feed that into a classical computer, it's a classical random
number generator. You can give your as much data as
you want, you will never be able to say did
the red come from the left side or the right side.
You would take infinite data like it is like you
would have to break a classical random number generator. I
can show you a quantum algorithm that can do that deterministically.
(52:18):
So where we're thinking is when the data appears to
be completely unstructured or you looks essentially like a complete
random number to the classical methods, there are quantum methods
that can actually potentially find that structure.
Speaker 3 (52:37):
That's it for this episode of Smart Talks with IBM.
If you haven't already, be sure to check out my
conversation with IBM Chairman and CEO Arvind Krishna, and stay tuned.
Another episode is coming soon. Smart Talks with IBM is
produced by Matt Romano, Amy Gains, McQuaid, Trina Menino, and
(52:58):
Jake Harper. Engineering by Bird Lawrence, Mastering by Sarah Buger,
music by Gramoscope, Strategy by Tatiana Lieberman, Cassidy Meyer and
Sofia Derlon. Smart Talks with IBM is a production of
Pushkin Industries and Ruby Studio at iHeartMedia to find more
Pushkin podcasts. Listen on the iHeartRadio app, Apple Podcasts, or
(53:22):
wherever you listen to podcasts. I'm Malcolm Godwell. This is
a paid advertisement from IBM. The conversations on this podcast
don't necessarily represent IBM's positions, strategies, or opinions.
Hey everyone, it's Robert and Joe here. Today we've got
something a little bit different to share with you. It's
a new season of the Smart Talks with IBM podcast series.
Speaker 2 (00:09):
This season on Smart Talks with IBM, Malcolm Gladwell is back,
and this time he's taking the show on the road.
Malcolm is stepping outside the studio to explore how IBM
clients are using artificial intelligence to solve real world challenges
and transform the way they do business.
Speaker 1 (00:25):
From accelerating scientific breakthroughs to reimagining education. It's a fresh
look at innovation in action, where big ideas meet cutting
edge solutions.
Speaker 2 (00:34):
You'll hear from industry leaders, creative thinkers, and of course
Malcolm Gladwell himself as he guides you through each story.
Speaker 1 (00:41):
New episodes of Smart Talks with IBM drop every month
on the iHeartRadio app, Apple Podcasts, or wherever you get
your podcasts. Learn more at IBM dot com slash smart Talks.
This is a paid advertisement from IBM.
Speaker 3 (00:56):
Hello, this is Malcolm Gladwell and you're listening to Smart
Talks with IBM. Every year, Tech Week brings thousands of
people together to network and learn about what's emerging across
the technology ecosystem, and at this year's conference in San Francisco,
I had an amazing opportunity to sit down in front
of a live audience with Jay Gambetta. Jay has been
(01:16):
with IBM for years and was recently promoted to Director
of Research. In this job, Jay has an important mission
helping the company build the future of computing. In the
last episode of smart Talks, I began to learn about
quantum computing from IBM Chairman and CEO Arvind Krishna. But
this conversation I had with Jay went even deeper and
(01:38):
convinced me that the development of quantum isn't just a fun,
exciting new paradigm of computing, it may be one of
the most important scientific achievements of my lifetime. Jay, Good morning, morning,
Welcome to Smart Talks with IBM. Thank you special live
(01:59):
recording here for tech Week and congratulations. How long have
you been Head of Research at IBM?
Speaker 4 (02:05):
Since October one? It's October tenth today, since nine days,
nine days.
Speaker 3 (02:10):
Can you just talk a little about the position. This
is one of the most important positions in research in
the world.
Speaker 4 (02:18):
IBM research has been around for eighty years and it's
done some tremendous technology, a lot of inventions and fundamentals
for semiconductors, algorithms, AI. Yeah, I think if we look
back to where a lot of the innovation and the
technology of the world comes from, I think you can
find Ibram's footprints on it, and you can find IBM research.
(02:39):
So yeah, I'm very excited for the opportunity, but I'm
also aware that there's big shoes to fill, and I'm
looking forward to how we take IBM research forward. Obviously,
I'm going to be bringing a lot of the quantum side,
which we're going to talk about later. Beyond quantum, there's
important work that needs to happen in AI hybrid cloud,
(03:01):
and I think we're going to also enter into this
new period of mathematics where we get to use quantum
machines and also AI machines. And there's some really good
hard mathematical questions to answer.
Speaker 3 (03:13):
How many people do you have working for you?
Speaker 4 (03:15):
I've been researchers in the three thousand researchers across many
different labs around the world. Our main lab is in Yorktown,
but then we have the lab actually out on the
West coast in Armadan or SBL Now and then we
have one in Zurich, Japan, and a few others around
the world.
Speaker 3 (03:32):
Tell me a little bit before we get into quantum.
I'm just curious about your path. So you're Australian. Yep,
we were talking about earlier backstage. Your accent has become muted.
You should crank it up because it's.
Speaker 4 (03:45):
Yeah, I'm slowly losing my Australian accent. I've been in
the US since two thousand and four, so accent, you know,
to sound very Australian. Yeah, but how do you practice it?
Maybe I got to go back to Australia. Here a
more Australians say gooday, how's it going?
Speaker 2 (04:00):
Like that?
Speaker 3 (04:01):
And you you didn't grow up thinking you're going to
be a scientist one day, now.
Speaker 4 (04:05):
I grew up in a pretty normal life. My dreams
as a kid was building things, so I was either
going to be a carpenter or a mechanic. But I
had some great teachers that inspired me to go to university.
And I didn't even know, honestly what a scientist was.
And then I found myself at university doing science, particular physics,
(04:26):
and I ended up loving it.
Speaker 3 (04:27):
So you go from there to what do you do
your PhD.
Speaker 4 (04:31):
So I did my undergrad in Australia. I did it
actually in laser science, so I think I watched some
TV show in lasers seemed interesting, so I wanted to
learn about lasers. And then I realized in trying to
understand lasers there was this quantum mechanics, and so I
was like, all right, I want to actually understand this
(04:53):
quantum mechanics. So I did my equivalent of what you
and the US school masters. We call it honors in Australia,
but we do a research project. I said, I wanted
to shoot lasers into atoms and measure cross sections and
I got really into quantum physics. So then I decided,
all right, I don't understand this quantum physics. I want
to do my PhD in interpretations of quantum mechanics. So
(05:16):
I jumped in and said, all right, what is this
quantum mechanics? Why is everyone arguing on these different interpretations.
Then I finished my PhD in Australia doing that. Then
I moved over at the end of my PhD interpretations,
it's more people arguing about the equations whilst I think
it's really important. I decided if it's going to be
(05:37):
like a collapse equation versus many worlds, or a hidden
variable model, or that just quantum mechanics decoheres because we
don't see supersitions in the everyday world, because it interacts
with environment. The only way to answer that question was
to build a quantum computer. And so then I decided
at the end of my PhD, I wanted to work
(05:59):
out how to build a quant computer. And then I
left there and I went to Yale, and then at Yale,
that's where I got into superconducting cubits, which just a
few days ago one of the professors there just won
the Nobel Prize this year.
Speaker 3 (06:12):
Oh wow, Yeah, I'm very interested in tracing because your
career follows the arc of quantum computing in a certain way.
Right at the time when you asked the question, what
I really want to do is to figure out how
to build a quantum computer. Where are we in quantum
computing at that point?
Speaker 4 (06:30):
Yeah, So that would have been nineteen ninety So there
was Shaw's algorithm came out, let's say ninety five. There
was a lot of theory. And then the reason I
went to Yale is because people had started to show
that they could see quantum effects in electrical circuits. So
these macroscopic objects they were starting to behave quantum mechanical
(06:52):
There was a really significant breakthrough in nineteen ninety nine
where Yazoo Nakamura in Japan showed that a qubit could
exist in these electrical circuits. And then I found out
the group at Yale were really trying to take these
electrical circuits and couple them together. And so it was like,
if I can build something using electrical circuits and they're big,
(07:15):
that that's the best way that you cancide to test
and understand whether quantum mechanics breaks down at a macroscopic
scale or not. Can we actually make them behave as cubits?
And I agree. When I came to Yale, the cubits
were not very good. They were actually a couple of nanoseconds.
They were unstable. Electron would jump onto the chip and
(07:37):
then they would change all their configurations, so you have
to restart your experiment. And so for the first time
at Yale, it's kind of what the challenge there was,
how do we make a cubit? How do we make
a stable cubit? And that took about five years and
that took us up to two thousand and seven. And
I think the rest of the world looks and says
quantums like just blowing up. But it's actually been like
(08:00):
all most phases theory, showing that we got the algorithms,
how do we make a cubit? How do we couple
of the cubits together? And now we're in the scaling phase.
Speaker 3 (08:09):
Describe for us, because many people in this room, me included,
have only a kind of surface level understanding of what
we mean when we use that phrase. What is the
difference between classical computing and quantum computing? What does that
word mean?
Speaker 4 (08:25):
Yeah, so you can go down the physics way and
talk about supersition and entanglement, which we can go in later,
but actually feel it's a bit of a distraction. So
when you think of classical computers, what they were is
there were machines that were very good at adding numbers together,
like simple addition, and they really showed that they could
(08:47):
add these numbers together really really fast. And now with
GPUs and other AI accelerators, we can add those numbers
together in parallel, and so the whole classical computing can
come down to just arithmetic, just adding numbers together. It
turns out that there's a math that is the quantum
mechanics shown.
Speaker 3 (09:07):
To be true.
Speaker 4 (09:08):
It's more like a group theory type structure. And the
way quantum works is it has a different mathos are primitive,
and if we can exploit that new math and build
a machine that does it, it allows us to answer
different questions. And so think of it as a branching
from classical compute that is very good at adding just
numbers together to something that allows us to work with
(09:30):
an algebra that is much much harder to represent with addition.
And that algebra happens to be the same algebra that
defines the fundamental equations of nature shirting as equation. So
this is why you say it computes the same way
nature does. But there are many other interesting problems. So
the way I explain it to people is think of
(09:51):
it as bringing a new primitive to computer science and
allowing us to work how to go with it. And
I like the analogy. Well, actually, maybe go back. So
if you went back in time, so we're one hundred
years of quantum, and you went back in time and
you asked, what is the foundation? Is a chemistry or physics.
(10:12):
What would have probably the scientists of one hundred years
ago would have said is they would have said, you know,
chemistry is about the small, physics is about planets and
things like this. And one hundred years ago when Heisenberg
or Einstein, all the greats, Schrodinger himself invented quantum mechanics,
it was this concept that nature is discreete not continuous.
(10:34):
It actually brought all the physical sciences together. And now
quantum mechanics is like it is the foundation of the science.
And so now what quantum computing is by that analogy
is computer science. The foundation of the math is coming
together with the physical science to allow us to compute
using math that if you were to try to represent
(10:55):
it with classical computers, it takes exponential time.
Speaker 3 (10:59):
Yeah, it was a classical computer and an expense in
a way that someone is well informed as I am
can understand it. A customer computer works primarily on problems
that can be easily represented in numerical form in numbers. Yes,
quantum allows you to step outside to a class of
problems that don't necessarily have a simple numerical representation.
Speaker 4 (11:20):
Yeah, and so imagine I got some medicine or or
some set of operation, but call it A and I
then follow it by a different operation B if A
followed by B gave a different answer than B first
followed by A. So in mathematics we call that commuting.
But like you can think of a correlation there one
(11:41):
one gives you a different outcome to the other. That
means there's an algebra behind it. That Representing that algebra
traditionally on classical computers is really really hard, whereas that algebra,
if we can get creative, we can come up with
ways of representing that math. So we step as you say,
we step out aside of the simple math to a
(12:02):
new math to allow us to calculate interesting problems.
Speaker 3 (12:06):
So quite in a sense, compliments it doesn't replace judicial.
That's good.
Speaker 4 (12:11):
I think this is one of the this is you're
exactly on is people think quantum is going to be
replacing classical If your problem is good at adding numbers together,
you should just keep using classical computers. I think the
future is going to be heterogeneous accelerators, and it will
definitely have quantum as one. But in some sense, the
next generation of superstars are going to be those applied
(12:32):
mathematicians that know, how do I write a problem using
the simple math of classical computers or the more complicated
math for quantum computers, and how do I actually iterate
between them? And things like this? This is where I
think the next generation of students are going to come
up with much more novel ideas. I can give you
examples of what we want to do on quantum, but like,
(12:54):
you're giving them a fundamental, foundational new thing, and so
I'm optimistic that will do much better jobs than my generation.
Speaker 3 (13:03):
Well, yeah, we're to get to some of the albums
in a moment, but I wanted you to the most
kind of down to you said, as a kid, you
thought you might want to be a mechanic because you'd
like to build things. Describe to me what it takes
to build a quantum computer, Like, what are you doing
that's different from building a classical computer.
Speaker 4 (13:22):
Yeah, so maybe I'll give you analogy and then i'll
go in so the way classical computers, we've got them
to get to smaller and smaller sizes like five seven animeters,
five animeters and things is actually inventing material to kill
quantum effects, so you actually put dielectrics and other things
(13:43):
in there. To kill the quantum tunneling effects, and you
want them to behave more classically. In the quantum world,
you want to get rid of all the classical effects,
So you want to get rid of the ability of
the cubits to interact with the environment. And in the
in the sort of technical world, we call it this
quantum conflict. The more ways you want to control the
(14:05):
quantum computer, you open it up to interacting with everything else,
like interacting with its environment. So the biggest challenge has
always been how do we give more control but don't
bring in other sources of noise. So I want to
be able to do gates on the cubit, but I
don't want it to decohere. I want to couple the cubits,
(14:28):
but I don't want them to couple to other things.
So the hardest challenge is the energy inside the cubits
is a nine gigahertz, and if your tames that by
HBO tend to the neggave thirty four with nine, you're
at a tender the negative twenty like three or something
in energy. That's a tiny amount of energy. So you're
(14:48):
trying to have a tiny, tiny amount of energy to control,
and you don't want that to interact with anything, So
you have to cool them down, you have to isolate them,
and you have to make the quantum effects dominate over
the classical effects.
Speaker 3 (15:03):
So practically, if I'm trying to do that right now,
how big are these machines.
Speaker 4 (15:08):
So the cubits themselves are not that big. So the
cubits themselves are like a few microns. But yeah, most
of the size so you can see some of our
I got the pleasure of showing you around to one
of the machines in Yorktown. You saw that they're like
twenty foot by twenty foot in size. Most of that
is all that equipment to isolate the cubit chip, which
(15:30):
is only a few millimeters when you put it together
from the rest of the environment. We will, as we
get better at that, miniaturize all the isolation. But that's
cooling it down to a few milli calvin, so about
a thousand times colder than outer space. It's isolating the
noise on any electrical signal so that no noise from
(15:52):
the outside world gets into the system. And so that's
a lot of isolators, filters, and things like that that
we've had to invent and to allow us to make
the quantum properties of this chip go.
Speaker 3 (16:04):
It's like the Princess and the pe mounds and mounds
and mounds of mattresses trying to isolate the impact of
this little thing, and that maybe.
Speaker 4 (16:12):
That's the best way to describe it. Yeah, and you've
got to keep it really really prestige.
Speaker 3 (16:17):
But that when you show me. So in the in
the lobby of the Watson Research Center in New Yorktown,
which by the way, is just the coolest building. It's
like a it's like a modernist it's also master piece. Anyway,
in the lobby there's there are these is it two machines.
Speaker 4 (16:34):
It's it's inside a container that has three machines machines.
Speaker 3 (16:39):
So what can you can you tell me what would
one of those machines cost to build right now?
Speaker 4 (16:44):
So typically we put them together in a way where
we upgrade them because we want to as I as
I was talking about before, one of the things we
want to do is always get algorithms done on our machines.
And I've got a roadmap of build bigger and bigger machines.
So usually one of those quantum processes today is out
(17:05):
of date in six months so we want to build
this future of computing that leverages quantum computing, where every
six months we've outdated a quantum processor. Eventually, hopefully we
get to a point where it's like stable and it
can be many years operating. But we want to get
(17:25):
as large a quantum computer in the hands of people
to explore the math as possible to come up with
those new algorithms. So we've had a philosophy of having
them open, working with universities and things like that. So
to answer a question of costs, yes, there's cost in
building the system, but we are operating in them much
more in a service model where people pay to use
(17:46):
the machine because we have to continuously calibrate it and
operate it and so depending on various different things. Professors,
we have a credits program where they get free access
some universities and animal prizes. They can buy premium access
and get more access. So think of not like a
cost of it, because it's almost like a continuum. I
(18:08):
want to make sure that the best quantum processors that
I can build get in the hands of students and
professors and interested enterprises that want to explore these machines
as fast as possible, and typically every six months we
upgrade it.
Speaker 3 (18:22):
Yeah, you don't start over, you upgrade.
Speaker 4 (18:25):
We upgrade various different pieces, the processor, the electronics. Some
upgrades are just simply replaced the processor. But as an example,
I think many people have probably seen photos of quantum
computers and you see this scary thing with all these
wires hanging down, as I've referred to as the chandelier,
and it's got all these wires with loops and things
(18:47):
like that. They're called co x keebles. When we first
put the quantum computer on the cloud in twenty sixteen,
you could probably only fit about fifty cubits inside one cryostat.
We've had to upgrade all those cables so that we
can fit around one thousand. I want to get to
three thousand, and that's about miniaturizing it. So to answer
(19:07):
your question, an upgrade, it depends. It can be either
just the processor or it can be the complete insides.
And we're actually in our third generation of our electronics
to control the systems, to make them faster, less noise. Internally,
we've got exciting results of going to something like cold cryocemos,
so you can bring down the cost in terms of
(19:29):
energy of running these quantum computers almost negligible, and you
could imagine future quantum computers. I'm not going to require
much energy to run, so unlike classical compute that requires
lots of energy. The biggest machines that we envision is
only in the few megawatts. But we have to upgrade
to future controls that use less energy. So it depends.
(19:51):
It's my long answer, short answer to how it upgrades,
and it depends on what it is.
Speaker 3 (19:57):
The only observation that I felt I could with making
when you showed me the quantum machine is it's gorgeous.
I look at art.
Speaker 4 (20:05):
I've always believed that, and I think that there's an
IBM saying good design is good business. But we've always
taken pride in making sure what we build. I don't know.
I feel if you're going to build something that is
new that can change, you should take the time to
make sure it looks and feels good.
Speaker 3 (20:26):
Will you donated to MoMA when you're through with that
particular Actually, I think we just put.
Speaker 4 (20:32):
An old version of one of our insights with the
United Airlines and the AAPS, which is the American Physical Society,
and the University of Chicago. There's a replica right now.
If you fly into one of the terminals in Chicago,
you can walk and see one.
Speaker 3 (20:48):
Oh really yeah, well the most advanced thing at all air.
Speaker 4 (20:51):
I'm sure probably, but yeah, I hopefully, I think, yeah,
we're open to that. But yeah, I appreciate that you
love the design. It was beautiful.
Speaker 3 (21:00):
So I last week I interviewed for another episode of
Smart TALX your CEO, Ivin Kushner, And when we got
to the quantum question, I mean, he's always alliant and brilliant,
but quantum, he's like lit up. I mean right in
thinking that IBM is much more invested in quantum than
(21:22):
anybody else. Is that a fair statement? Oh yeah, most definitely.
Why Why did IBM choose to kind of make this
such a priority.
Speaker 4 (21:29):
So when I took to the history of the physics side,
there's this interesting thing in the history of computing. So
we build computer classical computers today using bits and sea moss,
and they consume energy. Do you know that there is
a way in classical where you can actually compute without
using energy. It's called reversal computing. Turns out to be
(21:51):
a terrible idea. It's not practical to build. But IBM
investigated that with Ralph Laura and Charlie ban It early
on and they proved the concept that reversible computing. The
first use of quantum information theory, one of the first
actually was from IBM. When I did my PhD, I
(22:12):
remember actually picking up this paper on quantum teleportation and
seeing IBM written there, and at the time I remember
thinking that they make PCs. Well, what the hell are
they doing this foundational paper on quantum teleportation? Why are
they doing it? So to answer your question, actually, IBM
was the first in quantum information science because it's the
(22:33):
fundamental of computation. Can we actually come up with compute
that we can go beyond the classical So way before
anyone was talking about it, they were doing fundamental theory.
And then as we've built it, we've always When I
first came there, the experimental team was small. In twenty eleven,
we've had a small team that we're focusing on single
(22:57):
cubitts coupling them. I think in two thy twelve was
the first time we showed really good two Cuba gates
and no one was talking about quantum computing then. And
then I remember in about twenty sixteen, I said to
actually Arvin was the director of research, then can we
actually put our quantum computer on the cloud? Well that's
(23:20):
probably twenty fifteen, and it was always supporting that. So
as we've done more and more we've been able to
do it. It's had this program going now, I agree
is very visible, like because we're in this scaling phase
and so we're invested to keep scaling it and to
get why is At IBM research, what we always do
(23:43):
is answer what is the future of computing? Whether it's
coming up with new algorithms, coming up with better AI,
coming up with quantum, or coming up with just how
do different accelerators go together. It's our DNA to answer
the question of what is the future?
Speaker 3 (23:59):
Need a perfect problem for IBM because you kind of
need to have a legacy of building stuff, building actual
physical machines.
Speaker 4 (24:07):
Yeah, that's why I came to IBM. I wanted the experience,
the culture of building hard things that others have not
done before.
Speaker 3 (24:19):
Where do you imagine we are in the timeline of
this technology? It will come a point when it will mature.
My cell phone is a mature technology at this point.
How far are we from that point? With condom.
Speaker 4 (24:32):
So I think there's various aspects of it. So we
sat in twenty and seventy we set our goal that
in twenty twenty three we would be able to build
a machine that was beyond classical computers to simulate it.
And we achieved that in twenty twenty three. So to
run a biggo we call it a quantum circuit. The
details of it don't matter, but to run a quantum
(24:54):
workload that if you were to simulate that workload how
a quantum computer operates, it's on a classical computer, you
couldn't do it. So we said that as our first
and now I've made it publicly that by twenty twenty
nine we'll build the first fault tolerant quantum computer. That is,
one that can completely handle the noise to the level
(25:16):
to allow you to run a very very large, large problem.
Speaker 3 (25:19):
So an example of a large problem.
Speaker 4 (25:21):
Yeah, a large quantum problem. So for around a couple
of one hundred cubits and one hundred million operations, you're
talking still interesting science problems like simulating a molecule, or
calculating a small optimization problem, or calculating, say some part
(25:42):
of a matrix update. In some type of differential so
it'll still be scientific, but it'll be at the point
where it's beyond, well beyond any classical approximate method. And
then I think that's twenty twenty nine. That's twenty twenty nine,
So we're four.
Speaker 3 (25:58):
Years away from something that can start to handle.
Speaker 4 (26:01):
Interesting problem, serious problems. I do believe the scientists will
find interesting heuristic problems before that, And so over the
next four years, you're going to continue to see more
and more let's call them heuristic not provable quantum problems
that run on quantum computers that come out. We're see
more and more come from many of our partners and ourselves.
(26:22):
Heuristic problems have value, but they have to be tested,
they have to stand up over time. You have to
run them many, many times, you have to try different ones,
and many times heuristic can lead to formal problems. So
you're going to see, because we're beyond now the point
that you can simulate these quantum computers with any classical computer.
They're kind of like a scientific tool. So they're exploring
(26:43):
the heuristic.
Speaker 3 (26:44):
What do you have to get done between now and
twenty twenty nine to get there?
Speaker 4 (26:47):
So we had to reinvent how we wanted to do
error correction. So we have to demonstrate modules and if
we can demonstrate these error corrected module and our goal
is actually it's called Kooko, but I name all that
chips after birds, so it's called Kuoko. Borrow is named
after an Australian vert. I think I still say Cuoko
Borrow the way Australians do. We need to then show
(27:10):
that we can make a single module and then we
want to connect two of those modules together, and I
call that one Cockatoo, which is another Australian vert. And
then if we can do that, so that's twenty six
and twenty seven, and then we want to scale them
scale those modules, and that we call Starling, and we
want to scale that in twenty twenty nine. So get
(27:30):
a module, join two modules together, and scale and so
each module is going to be around one thousand cubits.
Speaker 3 (27:37):
The challenge to getting there is it finding the right
material or how would you describe what that's.
Speaker 4 (27:43):
The beauty to be done. That's the beauty of it
is if we would have been here two years ago,
I couldn't tell you how it would be done. So
we had a huge breakthrough we came up with a
new code, a new quantumeric Russian code, and that code.
The biggest in part of that code that is the
most important is it's modular in nature. So previous codes,
(28:07):
without getting too technical, they were very monolithic and you
had to build a very big device and I wouldn't
have known we would have to invent tools like new
simos tools to do that. So we came up with
this new code. We started on twenty nineteen, we published
in twenty twenty four. We kind of had most of
things worked out in twenty twenty three. That's why we
(28:28):
got confident to release the thing. So the biggest breakthrough
we had is coming up with a code that's modular
in nature. And think of that as a like a blueprint.
And so now we have the blueprint and now we're
doing engineering tasks to implement every part of that blueprint.
Speaker 3 (28:45):
And so the minute you had that breakthrough, then you
began to have confidence at something exactly these goals could
be met.
Speaker 4 (28:51):
And then you can't. And then anyone that's done engineering
will know what I'm talking about when I say this
is cycles are learning. It takes so long from test
idea to build two tests in hardware, the cycles of
learning are much much lower than software, Like you can
be really, really faster in the software. So then we've
(29:11):
planned out our iterations over the next few years, and
so we have to successfully demonstrate them. I may slip
because sometimes you may estimate your time wrong, but we
now have exactly what we want to do for the
next four years.
Speaker 3 (29:28):
I want to go back to that breakthrough for a moment.
What does the word breaks we mean in that context, Like,
it's not that you get a call in the morning
from somebody who says.
Speaker 2 (29:36):
I did it?
Speaker 3 (29:37):
Do you see it coming? Or is it a surprise
when they get there.
Speaker 4 (29:40):
So the way this one worked is Sogo Brave, who's
an algorithm person at IBM, one of the smartest and
quantum information.
Speaker 3 (29:48):
Don't mention his name to everyone. You'll come for him.
Speaker 4 (29:52):
Everyone in quantum already knows his name. I don't think
there's an idea that has not originated from him in quantum.
So we're looking at other codes and we'll go all right,
we've got to get serious about these codes. And others
were starting to propose to bring these and then we
call them LDPC codes from the classical space into the quantum,
(30:16):
and I asked him, we need to get ahead of
this and understand what they're doing it. He's like, the
most modest perfuse late, Jay, let me learn about them
and I'll generate a report for us and we'll read
through it. And then I said great. Then I don't know.
Six months later, he comes back with one hundred page
report on everyone everyone had done in LTPC codes. I'm like, awesome.
(30:38):
So I started then to read from them. And then
we said, all right, how do we under the assumptions
of the hardware we can build? Can we get an
LTPC code knowing what we can build? And that's a
great question, and so we put a small team together
to investigate and honestly took two to three years, and
(31:01):
we iterated and we used the constraints, so we had
the sort of theory, and then we had the constraints
of what we could build. And we iterated for a
few years, and then at the end of that we
came out with a solution that yes, it is possible
to meet all the constraints of the hardware and build
a code that will work.
Speaker 3 (31:22):
I'm just curious about So you had this task, this problem,
you want to solve. And when you set out on
the task of trying to solve the problem, what's your
certainty level that you'll get a solution.
Speaker 4 (31:34):
Well, that's the beauty of science for things where you
kind of have a few ideas. My philosophy is try
a few for the ones that need to be in
that like wow moment. It's honestly, you've got to set
the ambition really, really high, but know when to stop.
It was a great team that went together to get
(31:56):
that breakthrough, and we knew that we needed to come
up with a code that met the requirements of the experiment.
And I think what was different before then is the
theorists that were doing error correction codes didn't necessarily know
the constraints of experiments, so it was like really more
pen and paper. So this became one, all right, given
(32:18):
these sets of constraints, is it possible when LA's.
Speaker 3 (32:22):
Questions about this? Sorry, I love these kind of moments
when things become clear. At the time the problem was solved,
were you aware of the implications of the solution or
did that takes you knew exactly what.
Speaker 4 (32:35):
We set out exactly like either we were going to
have to work out how to cool down a very
large piece of silicon, which would require a lot of
engineering and building tools beyond what anyone has ever built
in the silicon semoss industry to implement the known codes,
or we had to come up with a different one,
(32:56):
and once I knew that we had one, that I
didn't need to re invent any tools to build. The
implications are clear how.
Speaker 3 (33:04):
Much time elapsed between the time you heard the problem
was solved and the time you told Arvin Krishna, the CEO,
the problem was solved.
Speaker 4 (33:13):
I'm sure the next time I spoke to him, I update,
but I don't remember. The beauty of Avin is he
trusts the scientists will do it, and so he doesn't
really check on us. We update him when it is,
and he he empowers us to do really hard problems.
Speaker 3 (33:27):
Yeah, so let's talk about uses. I mean, they're really
like cool, big shiny machine. I think you'll get pay
twenty twenty nine. But there's all kinds of really interesting
problems you're already working on.
Speaker 4 (33:40):
Yes, this is like another interesting area is I can
prove in pen and paper algorithms that we want to
run that. Like, it's not that we don't know what
to do with a quantum computer, there are hundreds of algorithms.
So you can go to I think it's called quantumzoo
dot com and you can see many many algorithms. People
are coming up with more of more of them that
(34:02):
they prove by pen and paper. Imagine, now we have
a machine that you can't simulate. How do you actually
discover algorithms in a scientific way? How do you look
and discover algorithms using a quantum computer. We're in this
exciting period right now, and so even though I can
(34:24):
prove these ones that we can run in the future,
there's a big white space between what the machines we
have and we're going to build and continue to do
and those ones that want the provable ones. And I'm
an optimistic person by nature. I think getting those machines
in the hands of students to explore and look at
(34:45):
heuristic algorithms, so looking at the equivalent of doing numerical
algorithms on computers, which there's many histories of numerical algorithms
being discovered on classical computers before we had formal pre
that we rely on today. People would even argue the
way AI works was driven numerically, even though we have
(35:07):
input into it. There are ones in optimization driven numerically.
We are entering that phase. So the computer scientists now
need to go play with these primitives. Our prediction is
over the next couple of years we're going to see
valuable numerical equivalent algorithms emerge. And where the scientists are
(35:28):
going is in four categories. One is simulating nature, so
looking at either Haanji physics, chemistry, light problems. As an example,
with our partners in Japan, they took one of our
quantum computers and for Gackle, a very large classical supercomputer,
and they ran a problem where quantum was just a
(35:50):
sub routine of the problem that was running on all
of for garcule, and they were able to look at
an interesting molecule, a molecule that if you would go
by pen and paper you would have said to take
me a very long time to run that. They were
able to run that quite accurately, heuristically, and already get
results that are comparable with the best classical methods. So
they are extremely excited because they want to push that
(36:11):
further and they're sort of showing that you can take
a classical supercomputer with quantum as a subroutine and start
to push the level they were.
Speaker 3 (36:19):
This was trying to solve a medical problem.
Speaker 4 (36:22):
Is this one is a like most people don't realize,
Like iron sulfur, just something as simple as iron and sulfur,
we can't solve that exactly, Like iron sulfur, molecules are
too hard, so really small small molecules are really really hard,
too hard for classical computers to solve. People think we
can solve a lot of things. It actually turns out
(36:44):
we can't solve very much.
Speaker 3 (36:45):
You say solve its instance, you know precisely how that
molecule works and it's constructed.
Speaker 4 (36:50):
No precisely what the energy levels of that molecule is
and how they come together, and then be able to
do that on a classical computer and compare it talk.
Speaker 3 (37:00):
It would be really really useful to know that specifically, because.
Speaker 4 (37:04):
If you can have energy levels, then you can estimate
reaction rates. If you can estimate reaction rates, you can
see how different types of chemicals will react. That can
then lead to better informing eventually how to build materials
or even drug design. I just want to be careful
and not say, oh, we're going to solve drug design
or that because there's many scientific steps to make that so.
(37:28):
And so what quantum gives you as a different tool
to give you more accuracy and then lead to making
the different methods work.
Speaker 3 (37:36):
You can subcontract out aspects of a problem quantum right now,
and that just gets you further along than you would
have been.
Speaker 4 (37:44):
So at the moment, even this result still does not
beat the best approximate classical method. It's comparable. So the
art of chemistry for the last hundred years has been
about approximating. So what we've done is we have got
very very good at coming up with ways of approximating nature.
(38:06):
And a lot of the things that we do and
we exploit and we use to estimate approximations. They don't
a stimulate nature of the way nature is. They approximate it.
And there's I could list many different acronyms of different
methods that go into approximating nature. What quantum gives us
is to eventually get beyond that approximation and do it
(38:28):
the way nature works. And so we aren't beating those
approximation methods, and this is why I think, this is
why it's still in the science. But they're getting comparable,
getting comparable with a new tool where the previous tool
is a dead end makes scientists very excited. Yeah, that
nuance is where it is, and so that's in machine learning.
Sorry Hamiltonian. Then there's examples in differential equations, So can
(38:52):
I actually come up with differential equations and solve them?
And if I can solve them, you could look at
things like an obvious Stokes goes into weather. There's financial
differential equations that you can better predict. So differential equations,
there's many different examples there. And then I would say
that two others are optimization and then there's quantum versions
of machine learning that are very exciting as well.
Speaker 3 (39:16):
Cleveland Clinic one of the organizations that you guys have
worked with. Why would the Cleveland Clinic be calling you up?
Speaker 4 (39:22):
Because that problem that they want to look at. So
they've also done similar problem to the recent lab. So
they've taken that method now and they've looked at molecules
that matter for drug design. So they're fundamentally looking at
those molecules that matter for eventually replacing some of the steps.
(39:42):
So they're investing to see how reliable it can be done.
And so there's a scientist there that's done many iterations
now using the techniques that were done first with the
team in Japan, they've now replicated that for new molecules
that are essential primitives for eventually designed drugs and things
that may matter for medical Yeah.
Speaker 3 (40:03):
And also there's some finance firms yep, HBC, Van Good yep,
and their interest is.
Speaker 4 (40:09):
What so that was the differential equation and optimization. So
if you are doing very large calculations like risk portfolio,
or if you want to model the black Shaw's equation
or things like this that are fundamental for them to
make better predictions, come up with better trades and things
like this. That is a very hard computational task. And
(40:31):
so rather than quantum replacing that whole problem, can quantum
be a subroutine in there? And what HSBC showed is
they showed they could take their real data, they could
take their real classical method and they just replaced a
tiny part of it. They replaced a tiny part of
it with a quantum subroutine that allowed them to come
up with better predictions of the weights that then when
(40:54):
they were to compare trial A versus Trial B, it
was thirty four percent better at algorithmic tun and that's
a big deal for them.
Speaker 3 (41:03):
It's huge.
Speaker 4 (41:04):
Yes, Now do they need to do more trials? Do
they need to see is this a heuristic algorithm? Do
we need to be careful? Is there other classical algorithms
that go into these are great questions that are now
being investigated. So think of this period of heuristic algorithms
is really a period of scientific discovery using these machines,
(41:25):
knowing that we want to continue and build the ones
which have determinist their algorithms that can run.
Speaker 3 (41:33):
Do the people who would profit the most space starting
to run quantum experiments realize that they would profit so
much from running quantum experience And does the world know this?
You've given us a couple of specific examples, but generally speaking,
there must be a very large universe of people who
could gain from at least starting to play in the space.
Speaker 4 (41:55):
So the enterprises that use computation as key for their
understand the limits of classical computation and they're very interested
to get started. The universities are very interested. Could we
get more students doing more algorithms one hundred percent. Some
of the limitations on the rate of algorithm discovery is
(42:17):
because people are thinking through the classical way of writing algorithms.
My belief is yes, So this is why we want
to get more and more students and things, because it's
just starting. But I would say in general most people
are aware of it. Could we get more, could we
accelerate it?
Speaker 3 (42:33):
Yes?
Speaker 4 (42:33):
Do we need to make better hardware, do we need
to come up with better libraries, yes? Do we need
better software yes, But it's all happening over the next
few years.
Speaker 3 (42:43):
Is it hard to get someone who's spent their entire
life thinking in terms of solving problems to classical means
to make the transition to this new paradigm.
Speaker 4 (42:51):
There's a lot of examples when you approach something with
the classical intuition, it's not the right way to do
it when you approach it through the quantum. But if
people are being taught to understand the fundamentals of the math,
then a lot of the techniques carry across. I don't
recommend people need to learn about entanglement or supersition because
(43:14):
whilst the physicists will argue like spooky action a distance
and all these type of things, entanglement is the power. Yes,
that's how physicists are labeled. How quantum is different. But
I would say, do we need some physicists really worrying
thinking about that? Yes, but we need more applied mathematicians
that are realizing they can use this as a as
(43:35):
a different way of looking at the problems.
Speaker 3 (43:36):
Yeah. And when I asked you one question, know is
we're describing a a It's more than a new technology.
We're talking about a new paradigm. It's a way of
thinking about problems. Can you compare this to kind of
previous technological paradigms. I'm thinking at the last couple hundred years,
what does this rank in terms of a new field
(43:58):
that we've opened up.
Speaker 4 (44:00):
It's a hard question to answer, but I often say
the history of computing, this will be the first time
that computation is branched between classical and quantum. I like
thinking reading a lot in the past. One of the
things that I think was a way we changed as
a society was the invention of zero. Before zero, math
(44:22):
was limited. Realizing that numbers have a number a zero
allowed us to develop a whole set of new mathematics
that then went on and defined like everything from waves
to calculus to all of that. Yes, we can describe
it with that same math, but when we describe it
with that math, it gets exponentially big and gets impractical
(44:44):
to do. Now we can actually work on it, I
would say, if I had to give you a quick answer,
maybe going all the way back to when we were accepted.
Speaker 3 (44:54):
Zero, I thought you were going to say, like the airplane,
but in fact, yeah, you went several orders of magnitude
on that.
Speaker 4 (45:00):
Yes, but but I think it's sort of fundamental.
Speaker 3 (45:04):
This is absolutely fascinating. Thank you so much for chatting
with me about it.
Speaker 4 (45:08):
Thank you for your time.
Speaker 3 (45:11):
Hey, listeners. So normally we end this episode here, but
the Tech Week attendees asked Jay some really great questions,
questions I wish I'd asked, so we wanted to include
those here. Enjoy.
Speaker 5 (45:24):
Hi, J, thank you so much for the great presentation.
My name is Trixie Apiado. I work for Willis Towers Watson,
an insurance broker. I help seisos identify and quantify their
cyber risk so they can prepare for threats before they happen,
and so quantum threats keep me up at night. You
mentioned so many good problems that quantum can solve. It
(45:46):
can also break encryptions in our classical computer systems. So
what safeguards or policies do you implement in your teams
to build quantum capabilities responsibly? And what can we do
for people in this room as builders and users to
secure our data in systems before quantum computers become more
(46:09):
energy efficient, cheaper, and more available.
Speaker 4 (46:12):
So it's a great question. So yes, one of the
algorithms for quantum computing is to break our traditional encryption.
So at IBM Research, we were aware of this from
day one. We've come up with algorithms that we believe
and have very strong evidence will not be broken by
a quantum or classical computer, and this has selected them.
(46:36):
So first the scientific technical question, security is saved. There
are algorithms that exist that we can implement that neither
a quantum or classical computer can break. So the technical
answer is we're all okay. The more complicated answer is
a social and society answer. Encryption was built in classical
(47:00):
computing in a way that was never thought of being upgraded.
It's mixed everywhere. Some of it is downstream, some of
it is like software that you may use. Some of
it is software that you've developed. And I get that
if you've got a product and you want to have
it secure for the next ten years, you probably want
(47:20):
to think about how you're going to upgrade it, or
if you have data that needs to be secure for
the next ten years, it needs to upgrade to new encryption.
So the real challenge is more of a social business
problem of how do we actually transition from old encryption
to new encryption knowing this is going to happen. So
(47:41):
we at IBM have been very proactive on this. We've
developed tools where we can determine where encryption is used,
We've developed tools which can show you how to replace it,
and we early on have made sure the Mainframe when
we made these algorithms. So I think it was Z
sixteen that was the first version of the Mainframe to
(48:01):
have these quantum safe algorithms implemented. So my answer to
your question is, yes, there's a real problem, but it's
not a technical problem. It's a social and business problem.
And I'm not minimizing that. I understand that that is
a lot of work you need to start now. You
need to come up and do a you need to
(48:22):
make it part of your IT transformation. You need to
get onto it. And I realize I realize it's not
going to take zero time because it's not an easy
problem to do. So the short answer is one we
developed algorithms that we can't and we're developing tools to
help you in that transformation.
Speaker 6 (48:41):
Thank you so much, Thank you. My name is Emma.
I'm a product manager at Expedia working on software side
of things. My question is around the non technical roles
outside of the researchers, the mathematicians, the builders. How can
the rest of us, whether it be policymakers, those in
the legal fields, those thinking about what use cases quantum
(49:02):
can solve for in the future, what should we be
thinking about and how can we prepare for that.
Speaker 4 (49:07):
It's a good question. I think this is part of
the requirement of the scientists to being able to articulate
where they are. We need a forum for those type
of discussions. I think a lot of this can fit
within the forums that we already have for classical and AI,
and I think we need to just be asking how
do we actually bring them into them. Because I don't
(49:28):
think of quantum as a replacement of compute. I think
of it as an accelerator that expands what is possible,
and I think we can ask those questions in those forums,
are we doing enough now? I think I agree with you. No,
I don't know the answer to it.
Speaker 6 (49:43):
I think it's a really interesting perspective because those existing
forums do start to bring in those other fields as well,
so it could warrant the same sort of discussion and.
Speaker 4 (49:54):
Active And I understand those forums right now AI is
probably dominating, and it should be like we are going
through a period of time where AI is impacting society.
The technology is impacting society in big ways. So I
totally understand that most of their focus should be on AI,
(50:15):
but we should start to ask where is quantum in
that as well?
Speaker 7 (50:19):
Hi, I'm Gobi and I'm a graduating PhD student at
Northwestern and also a member of south Park Commons, which
is a fund here. You mentioned earlier that some problems
are best solved by classical versus some problems are best
solved by quantum. When we're thinking about this, if we're
not experts in quantum, but we're thinking about this from
an AI perspective, could you just clarify when we think
(50:41):
about quantum, what is deterministic and what is not deterministic.
Speaker 4 (50:46):
I think the future of computing we've got to get
our heads around is that not everything is deterministic, and
it's much more going to be probilistic. How do you
handle error bars? How do you put confidence? I think
a lot of those questions which you're referring to in
AI are going to completely apply in quantum. I actually
think it's a mistake to compare AI verse quantum. I
(51:10):
actually think of quantum as much its quantum verse classical compute,
and AI is going to come across on top. So
as we go forward and we get a better understanding,
I'm not going to say quantum is going to replace
the classical compute that enables AI, but I think some
of the math you do in AI will be able
to go to both. So what can we formally prove?
(51:34):
I can come up with a problem where I take
a circle and I color half of it red and
half of it of blue, and then I say I'm going
to apply an operation that takes those dots make it. Say,
let's say ten dots over here red, ten dots over
here blue, and I'm going to wind them around many
many times. I can then show you that if you
(51:54):
feed that into a classical computer, it's a classical random
number generator. You can give your as much data as
you want, you will never be able to say did
the red come from the left side or the right side.
You would take infinite data like it is like you
would have to break a classical random number generator. I
can show you a quantum algorithm that can do that deterministically.
(52:18):
So where we're thinking is when the data appears to
be completely unstructured or you looks essentially like a complete
random number to the classical methods, there are quantum methods
that can actually potentially find that structure.
Speaker 3 (52:37):
That's it for this episode of Smart Talks with IBM.
If you haven't already, be sure to check out my
conversation with IBM Chairman and CEO Arvind Krishna, and stay tuned.
Another episode is coming soon. Smart Talks with IBM is
produced by Matt Romano, Amy Gains, McQuaid, Trina Menino, and
(52:58):
Jake Harper. Engineering by Bird Lawrence, Mastering by Sarah Buger,
music by Gramoscope, Strategy by Tatiana Lieberman, Cassidy Meyer and
Sofia Derlon. Smart Talks with IBM is a production of
Pushkin Industries and Ruby Studio at iHeartMedia to find more
Pushkin podcasts. Listen on the iHeartRadio app, Apple Podcasts, or
(53:22):
wherever you listen to podcasts. I'm Malcolm Godwell. This is
a paid advertisement from IBM. The conversations on this podcast
don't necessarily represent IBM's positions, strategies, or opinions.
Stuff To Blow Your Mind News
Hosts And Creators
Robert Lamb
Joe McCormick
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Ruthie's Table 4
For more than 30 years The River Cafe in London, has been the home-from-home of artists, architects, designers, actors, collectors, writers, activists, and politicians. Michael Caine, Glenn Close, JJ Abrams, Steve McQueen, Victoria and David Beckham, and Lily Allen, are just some of the people who love to call The River Cafe home. On River Cafe Table 4, Rogers sits down with her customers—who have become friends—to talk about food memories. Table 4 explores how food impacts every aspect of our lives. “Foods is politics, food is cultural, food is how you express love, food is about your heritage, it defines who you and who you want to be,” says Rogers. Each week, Rogers invites her guest to reminisce about family suppers and first dates, what they cook, how they eat when performing, the restaurants they choose, and what food they seek when they need comfort. And to punctuate each episode of Table 4, guests such as Ralph Fiennes, Emily Blunt, and Alfonso Cuarón, read their favourite recipe from one of the best-selling River Cafe cookbooks. Table 4 itself, is situated near The River Cafe’s open kitchen, close to the bright pink wood-fired oven and next to the glossy yellow pass, where Ruthie oversees the restaurant. You are invited to take a seat at this intimate table and join the conversation. For more information, recipes, and ingredients, go to https://shoptherivercafe.co.uk/ Web: https://rivercafe.co.uk/ Instagram: www.instagram.com/therivercafelondon/ Facebook: https://en-gb.facebook.com/therivercafelondon/ For more podcasts from iHeartRadio, visit the iheartradio app, apple podcasts, or wherever you listen to your favorite shows. Learn more about your ad-choices at https://www.iheartpodcastnetwork.com
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com