Hi my name's Talia Gershon and I'm a scientist

at IBM Research.

Today I've been challenged to explain a topic

with five levels of increasing complexity.

It's a completely different kind of computing called

quantum computing.

Quantum computers approach solving problems

in a fundamentally new way.

And we hope that by taking this new approach

to computation, we'll be able to start

exploring some problems that we can never solve

any other way.

Hopefully by the end of today

everyone can leave this discussion understanding

quantum computing at some level.

What's this?

What do you think that is?

Fancy chandelier.

I think so too.

We jokingly call it the chandelier.

That's real gold you know.

This is a quantum computer.

It's a quantum?

Yeah.

It's a really special kind of computer.

What does it do?

It calculates things but in a totally different

way to how your computer calculates things.

What do you think this is?

An A.

Yeah.

Do you know what your computer thinks that is?

Zero, one.

(laughing)

This really specific combination of zeroes and ones.

Everything that your computer does,

showing you pink panther videos on YouTube,

calculating things, searching the internet,

it does all of that, with a really specific combination

of zeroes and ones.

Which is crazy right?

That would be like saying your computer only understands

these quarters.

For each quarter you need to tell it

that you're gonna use heads, tails.

And you assign it heads or tails.

So I can switch between heads and tails

and I can switch the zeroes and ones in my computer

so that it represents what I want it to represent,

like an A.

And with quantum computers,

we have new rules we get to use too.

We can actually spin one of our quarters.

So it doesn't have to choose just one or the other.

Can computers help you with your homework?

Your really hard homework?

Yeah it can.

Especially if doing your homework involves

calculating something or finding information.

But what if your homework was to discover something

totally new?

A lot of those discovery questions are much harder

to solve using the computers we have today.

So the reason we're building these kinds of computers

is because we think that maybe one day

they're gonna do a lot of really important things,

like help us understand nature better.

Maybe help us create new medicines to help people.

What's your favorite kind of computer?

Smart phones, tablet, regular, laptop, PC?

I've got to go with my iphoness.

So what do you do with your iphoness?

Social media, use it for studying.

Have you ever run out of space on your iphoness?

All the time.

Me too!

Yeah, always when I'm trying to take a photo.

So did you know that there are certain kinds of problems

that computers sort of run out of space almost?

Like your trying to solve the problem

and just like how you run out of space on your iphoness

when you're trying to take a picture,

if you're trying to solve the problem

you just run out of space.

And even if you have the world's biggest supercomputer

did you know that can still happen?

Wow.

So my team is working on building new kinds of computers

altogether, one's that operate by a totally

different set of rules.

So do you know what that is?

I have no clue.

[Talia] It's a quantum computer.

A what?

(laughing)

Have you ever heard of a quantum computer?

I haven't.

Have you ever heard of the word quantum?

No.

Okay so quantum mechanics is a branch of science.

Just like any other branch of science,

it's a branch of physics.

It's the study of things that are either

really really small,

really really well isolated,

or really really cold.

And this particular branch of science

is something we're using to totally reimagine

how computing works.

So we're building totally new kinds of computers

based on the laws of quantum mechanics.

That's what a quantum computer is.

I'm gonna start by telling you about

something called superposition.

So I'm gonna explain it using this giant penny.

Wow, is that like worth a hundred pennies?

I don't know what it's worth, but I can put it face up,

right, and that's heads, I can put it face down.

So at any given time, point in time,

if I ask you is my penny heads or tails,

probably you could answer, right?

Yeah.

Okay but what if I spin the penny?

So let's do it.

Okay so while it's spinning, is it heads or tails?

Heads.

While it's spinning?

Oh, I wouldn't know.

It's sort of a combination of heads and tails, right?

Would you say?

So superposition is this idea that my penny

is not just either heads or tails.

It's in this state which is a combination

of heads and tails.

And that this quantum property is something

that we can have in real physical objects in the world.

So that's superposition.

And the second thing that we'll talk about

is called entanglement.

So now I'm gonna give you a penny.

Wow!

(laughing)

when we use the word entangled

in everyday language, what do we mean?

That something's intertwined or--

Exactly, that there's two things

that are connected in some way.

And usually we can separate them again.

Your hair is tangled, or whatever,

you can unentangled it right?

But in the quantum world, when we entangle things,

they're really now connected and it's much much harder

to separate them again.

So using the same analogy,

we spin our pennies and eventually

eventually they both stop right?

And when they stop it's either heads or tails, right?

So in my case I got tails and you got heads.

You see how they're totally

disconnected from each other, right?

Our pennies, in the real world.

Now if our pennies were entangled

and we both spun them together, right?

When we stop them, if you measured your penny to be a head,

I would measure my penny to be a head.

And if you measured your penny to be a tails,

I would measure my penny to be a tails.

If we measured at exactly the same time,

we would still find that they were both exactly correlated.

That's crazy.

It's so cool, right?

Oh my god.

The way that we are able to actually see

these quantum properties is by making our quantum chips

really really cold.

So that's what this is all about actually.

This is called a dilution refrigerator.

And it's a refrigerator.

It doesn't look a normal refrigerator, right?

But it's something that we use,

actually there's usually a case around it,

to cool our quantum chips down cold enough

that we can create superpositions

and we can entangle qubits,

and the information isn't lost to the environment.

Like what could those chips be used to do?

So one of the things that we're trying

to use quantum computers to do

is simulating chemical bonding.

Use a quantum system to model a quantum system.

Yeah, I mean I'm definitely gonna impress all my friends

when I tell them about this, they're gonna be like,

quantum what?

(laughing)

So what do you think that thing is?

Is it some sort of conjecture circuit?

[Thalia] That is a really good guess.

There's parts of that that are definitely about conducting.

This is the inside of a quantum computer.

Oh wow.

(laughing)

Yeah, this whole infrastructure

is all about creating levels

that get progressively colder as you go from top to bottom

down to the quantum chip, which is how we actually

control the state of the qubits.

Oh wow.

So when you say colder, you mean like physically colder?

Yeah like physically colder.

So room temperature is 300 Kelvin.

As you get down all the way to the bottom of the fridge

it's at 10 millikelvin.

[Amanda] Oh wow.

Amanda what do you study?

So I'm studying computer science, currently a sophomore.

And the track that I'm in is the intelligent systems track.

Machine learning, artificial intelligence.

You ever heard of quantum computing?

From my understanding, with a quantum computer,

rather than using transistors, is using spins.

You can have superposition of spins,

so different states, more combinations means more memory.

So that's pretty good.

So you mentioned superposition, but you can also

use other quantum properties like entanglement.

Have you heard of entanglement?

I have not.

Okay so it's this idea that you have two objects

and when you entangle them together they become connected.

Oh okay.

And then they are sort of permanently

connected to each other and they behave in ways

that are sort of a system now.

So superposition is one quantum property that we use,

entanglement is another quantum property,

and a third is interference.

How much do you know about interference?

Not much.

Okay, so how do noise canceling headphoness work?

They read like ambient wavelengths

and then produce like the opposite one to cancel out.

They create interference.

So you can have constructive interference,

and you can have destructive interference.

So you have constructive interference,

you have amplitudes, wave amplitudes that add.

So the signal gets larger.

And if you have destructive interference

the amplitudes cancel.

By using a property like interference

we can control quantum states and amplify

the kinds of signals that are towards the right answer

and then cancel the types of signals that are leading

to the wrong answer.

So given that you know that we're trying to use

superposition, entanglement, and interference

for computation, how do you think we build these computers?

I have no idea.

So step one is you need to be able to have an object

or a physical device, we call it a qubit

or a quantum bit that can actually handle those things,

can actually be put into superpositions of states.

You know, two qubit states that you can

physically entangle with each other.

That's not really trivial, right,

things in our classical world

you can't really entangle things

in our classical world so easily.

We need to use devices where they can support

a quantum state and we can manipulate that quantum state.

Atoms, ions, and in our case superconducting qubits.

We make qubits out of superconducting materials.

But as like a programmer, how would quantum computing

affect a different way of writing a program?

It's a perfect question.

I mean it's very early for quantum computing

but we're building, assembly languages.

We're building layers of abstraction

that are gonna get you to a point as a programmer

where you can interchangeably be programming something

the way that you already do and then make calls

to a quantum computer so that you can bring it in

when it makes sense.

We're not envisioning quantum computers

completely replacing classical computers anytime soon.

We think that quantum computing

is gonna be used to accelerate the kinds of things

that are really hard for classical machines.

So what exactly are some of those problems?

Simulating nature is something that's really hard.

Because we take something like you know,

modeling atomic bonding and electron orbital overlap,

instead of now writing out a giant summation

over many terms, you try and actually mimic

the system you're trying to simulate

directly on a quantum computer.

Which we can do for chemistry,

and we're looking at ways of doing that

for other types of things.

There's a lot of exciting research right now

on machine learning, trying to use quantum systems

to accelerate machine learning problems.

So would it be like in five years,

or 10 years that I would be able to have

like one of these sitting in my laptop

just in my dorm?

I don't think you're gonna have one in your dorm room

anytime soon but you'll have access to one.

There's three free quantum computers

that are all sitting in this lab here

that anyone in the world can access through the cloud.

Okay so quantum computing creates new possibilities

and new ways to approach problems that classical computers

have difficulty doing.

Couldn't have said it better myself.

So I'm a first year masters student

and I'm studying machine learning,

so it's in the computer science department

but it mixes computer science

with math and probability and statistics.

So have you come up upon sort of any limits

to machine learning?

Certainly, depending on the complexity of your model

then computational speed is one thing.

I have colleagues here that tell me it can take

up to weeks to train certain neural networks, right?

Sure, yeah.

And actually machine learning is one research direction

where we're really hoping that we're gonna find

key parts of the machine learning computation

that can be sped up using quantum computing.

Yeah that's exciting.

So in a classical computer, you know,

you have all sorts of logical gates

that perform operations and they

change an input to some sort of output

but I guess it's not immediately obvious

how you do that with quantum computers.

If you think about even just classical information

like bits, right?

At the end of the day when you store a bit

in your hard drive, there's a magnetic domain

and you have a magnetic polarization, right?

Sure.

You can change the magnetization to be

pointing up or be pointing down, right?

Quantum systems, we're still manipulating a device

and changing the quantum state of that device.

You can imagine if it's a spin

that you could have spin up and spin down

but you can also, if you isolate it enough

you can have a superposition of up and down.

Sure.

So what we do when we try to solve problems

with a quantum computer is we encode parts

of the problem we're trying to solve

into a complex quantum state.

And then we manipulate that state to drive it towards

what will eventually represent the solution.

So how do we actually encode it to start with?

Yeah that's a really good question.

This actually is a model of the inside

of one of our quantum computers.

Okay.

So you need a chip with qubits.

Each qubit is a carrier of quantum information.

And the way we control the state of that qubit

is using microwave pulses.

We send them all the way down these cables

and we've calibrated these microwave pulses

so that we know exactly this kind of pulse

with this frequency and this duration

will put the qubit into superposition.

Or will flip the state of the qubit from zero to one

or if we apply a microwave pulse between two qubits

we can entangle them.

How do we measure it?

Yes exactly, also through microwave signals.

Okay.

The key is to come up with algorithms

where the result is deterministic.

Interesting, so what do those algorithms look like?

There's sort of two main classes of quantum algorithms.

There's algorithms which were developed for decades, right?

Things like Shor's algorithm which is for factoring,

Grover's algorithm for unstructured search,

and these algorithms were designed

assuming that you had a perfect

fault tolerant quantum computer.

Which is many decades away.

So we're currently in a phase where we're exploring

what can we do with these near term quantum computers.

And the answer is gonna be, well we need different

kinds of algorithms to really even explore that question.

Yeah certainly having a search algorithm

is very useful.

Factoring, those are definitely useful things

that I would imagine could be done a lot faster

on a quantum computer.

Yeah, they also unfortunately require fault tolerance.

Right now, the algorithms that we know of today

to do those things on a quantum computer

require you to have millions of error corrected qubits.

Today we're at like 50 and it's actually amazing

that we're at 50.

There's things that we know or we have strong reasons

to believe are gonna be faster to do on a quantum computer.

And then there's things that we'll discover

just by virtue of having one.

Sure, how could someone like me

who's a grad student, get involved in this

or what kinds of challenges are you facing

that someone like me could help out with?

I'm glad you're interested.

I think the place where lots of people can get involved

right now is by going and trying it out and thinking about

what they could do with it.

There's a lot of opportunity to find these near term

applications that are only gonna be found

by trying things out.

I'm a theoretical physicist.

I started out in condensed matter theory,

the theory that studies superconductors

and magnets and I had to learn a new field

of quantum optics and apply those ideas.

One of the nice things about being a theorist

Is you get to keep learning new things.

So Steve tell me about your research

and the work you've been doing in quantum computing.

My main focus right now is quantum error correction

and trying to understand this concept of fault tolerance

which everybody thinks they know it when they see it

but nobody in the quantum case can precisely define it.

It's something that we've already figured out

for classical computing.

Like something that amazes me is all the parallels

between what we're going through now for quantum computing

and what we went through for classical computing.

I was asking a computer scientist recently

where to read about fault tolerance in classical computing.

He said oh they don't teach that in computer science classes

anymore because the hardware has become so reliable.

In a quantum system, when you look at it

or make measurements, it can change

in a way that's beyond your control.

We have the following task,

build a nearly perfect computer

out of a whole bunch of imperfect parts.

Common myth, how many qubits do you have?

That's the only thing that matters.

Just add more qubits, what's the big deal?

Pattern them on your chip.

The great power of a quantum computer

is also it's Achilles's heel.

That it's very very sensitive to perturbations

and noise and environmental effects.

You're just multiplying your problems

if all your doing is adding qubits.

Exactly, so I think something

that frustrates a lot of people about quantum computing

is the concept of decoherence, right?

You can only keep your information quantum for so long.

And that limits how many operations you can do in a row

before you lose your information.

That's the challenge I would say.

As much progress as we've made

it's a frustration to still be facing it.

Let's talk about some of the things we think

need to happen between now and fully fault tolerant

quantum computers, to get us to that reality.

I mean there's so many things that need to happen.

In my mind one of the things we need to do is build

all these different layers of abstraction

that make it easier for programmers to come in

and just enter at the ground level, you know?

Exactly, so I think there's gonna be

a kind of co-evolution of the hardware

and the software up here and the sort of middleware,

and the whole stack.

Another common myth, in the next five years

quantum computing will solve climate change, cancer, right?

(laughing)

Right, in the next five years

there'll be tremendous progress

in the field but people really have to understand

that we're either at the vacuum tube or transistor stage.

We're trying to invent the integrated circuit and scale up.

It's still very very very early in the development

of the field.

One last myth I think we should bust Steve.

Quantum computers are on the verge

of breaking into your bank account

and breaking encryption and cryptography.

There does exist an algorithm, Shor's algorithm,

which has been proven mathematically

that if you had a large enough quantum computer

you could find the prime factors of large numbers.

The basis of the RSA encryption

is the most commonly used thing on the internet.

First we're far away from being able to have

a quantum computer big enough to execute Shor's algorithm

on that scale.

Second, there are plenty of other encryption schemes

that don't use factoring and I don't think

anybody has to be concerned at the moment.

And in the end, quantum mechanics goes to the side

of privacy enhancement.

If you have a quantum communication channel

you can encode information and send it through there

and it's provably secure based on the laws of physics.

You know now that everybody around the world

can access a quantum computer through the cloud,

people are doing all kinds of cool things.

They're building games.

We've seen the emergence of quantum games, right?

What do you think people want to do with them?

I have no idea what people are going to end up

using them for I mean if you had gone back

30 years and handed somebody an iphoness

they would have called you a wizard, so.

(laughing)

Things are gonna happen that we just can't foresee.

(soft music)

So I hope you enjoyed that foray into the field

of quantum computing.

I know I've personally enjoyed getting to see

quantum computing through other people's eyes.

Coming at it from all these different levels.

This is such an exciting time in the history

of quantum computing.

Only in the last couple years have real quantum computers

become available to everyone around the world.

This is the beginning of a many decade adventure

where we'll discover so many things about quantum computing

and what it'll do.

We don't even know all of the amazing things it's gonna do.

And to me that's the most exciting part.

(soft music)