Quantum Supremacy Explained - YouTube

Quantum supremacy’s that moment when a quantum computer beats the best supercomputers at

solving some kind of problem, and it’s a very exciting time right now in quantum computing

because, as of recording this video we’re on the brink of having quantum supremacy,

maybe it’s already happened.

So I’m going to explain what quantum supremacy is and why it is so exciting.

So to explain quantum supremacy it’s worth me explaining how quantum computers work.

Us physicists, we call normal computers ‘classical’ computers, and classical computers work with

binary.

There’s loads of bits that can either be in a state of zero or one.

In a quantum computer you have quantum bits, qubits, which can be in the state zero, or

one, or they can be in a special intermediate state and that’s called a superposition.

It’s a special phenomenon in quantum physics that quantum computers take advantage of.

Now when you measure a qubit the result you get is based on a probability, so if you set

this superposition state to be in the middle, you’ve got a 50% chance of getting a zero

and a 50% chance of getting a one.

But you can tune that state so you can make it so that you are more likely to get a zero

than a one, or the other way around.

So that’s the first phenomenon of quantum physics that quantum computers take advantage

of.

The second one is called entanglement, and this is where you bring together several qubits

and you join them together, and now that whole thing has to be treated as one object.

So if you take two qubits and you join them together, now this object can be in the mixed

state of four states, so zero zero, zero one, one zero and one one.

Each time you add a new qubit, you double the number of states that this thing can be

in, and that goes up exponentially.

So if you want to search through a whole load of different states, a classical computer

has to search through them one by one.

But in a quantum computer you’ve got these special quantum algorithms where you can enhance

the probability of the state that you want, and diminish the probability of it giving

you back the state that you don’t want.

So those two phenomena: entanglement and superposition are what give quantum computers their power.

So, what kind of quantum computers exist in the world today?

There are a whole bunch of different companies trying to make a quantum computer: Google,

Intel, IBM, Microsoft, D-Wave, amongst others.

And to complicate matters, there’s not just one kind of quantum computer there’s actually

a whole load of different approaches.

I’ve broken them all down here.

An important thing to point out here are Universal quantum computers.

Universal quantum computing can theoretically simulate any quantum system and so it’s

the fundamental computing machinery of the Universe.

There’s other approaches that aren’t Universal like quantum annealing and ion trap systems.

What they’re focused on are solving certain nice problems better than classical computers

by taking advantage of quantum physics and they’re a valid approach if those kinds

of problems are valuable.

They can be used as a good stepping stone towards a Universal quantum computer, because

Universal quantum computers are very difficult to build, and it’s good to learn stuff along

the way.

I’ve also included the number of qubits in these quantum computers as of today.

Now qubit number is just one measure of how good a quantum computer is.

Just as important is how low the noise is in the qubits, so how high quality the qubits

are, and also how well the qubits are connected together.

So let me describe a bit more about how you’d do the experiment to prove quantum supremacy.

Actually measuring when it has happened is quite a difficult thing to do.

So you take your best quantum computer and you take your best super computer and you

need to give them a problem to solve and then compare the results.

And to begin with the quantum computers can’t do anything better than a supercomputer except

for one thing, and that’s be a quantum computer.

Which might sound absurd, but the logic is this.

You can simulate a quantum computer on a classical computer, but it gets more and more difficult

to do so the more qubits you have.

Now IBM has got the current record for simulating a quantum computer with 56 qubits, so people

are seeing this as the target, that if you build a quantum computer with more than 56

qubits, you can’t simulate that with a classical computer, and so the quantum computer can

do some thing better than what a classical computer can do.

So the problem that they are looking at showing that the quantum computer can beat a classical

at is a certain problem called a sampling problem.

Now if you remember back when I was talking about a qubit being in a superposition state

between zero and one that gives you the probability of which one you’ll get back.

Imagine if you’ve got fifty qubits and they are all entangled together, they are simultaneously

in about quadrillion states, that’s a million billion states at the same time.

But when you measure that computer it just gives you back one state which will just be

a bit string of zero, zero, one, zero, one, one, one, corresponding to each qubit.

Now if you keep measuring that computer over and over again, you’ll get different bit

strings back, so different states back, but over time you’ll build up a distribution

of those states, the more likely ones and the less likely ones, and that’s a probability

distribution.

And each time you measure the quantum computer that’s like taking one sample from that

probability distribution.

Now quantum computers do that completely naturally, whereas simulating that whole thing on a classical

computer is incredibly difficult.

So that’s the problem that they are looking at proving that a quantum computer can do

it better than a classical computer.

So what’s that problem good for?

Absolutely nothing.

To begin with.

I mean there’s some crossover between this kind of sampling problem and a sampling problem

you get in machine learning.

But the shapes of those problems in quantum computing and machine learning are very ver

different, but they come from the same sort of underlying physics which is called statistical

mechanics so there is a tentative overlap there which people like Google are very very

interested in.

But I don’t undersell this achievement because, if you think about it, classical computers

have been around for seventy years and they’ve had literally trillions of dollars worth of

money pumped into research and development to reach them to the incredibly sophisticated

machine they are now.

Quantum computers have been around for twenty years-ish, and have had say a hundred million,

maybe a billion dollars worth of investment, for that technology, this new novel technology

to come along and beat the classical computers at even one narrow thing is a, is a huge achievement

and that’s only going to progress in the future and I’m really excited to see how

that whole thing evolves.

And finally I just want to end with a fact which I think is incredible.

You know I said that you can simulate, using our best supercomputers, you can simulate

a quantum computer with fifty six qubits.

Now if you raise the number of qubits up to two hundred and sixty, the size of the computer

you’d need to simulate that would need more bits then there are atoms in the entire known

Universe, which I think is just absolutely bonkers, but I love it.

Thanks so much for watching, that’s all I’ve got for you today.

And thanks also to the sponsor of this video brilliant.org they’re a website where you

learn by doing.

And I don’t know about you, when I was at University I would sit down, I’d write down

all the lecture notes, but the time where I’d really learn the subject was when I

sat down and did problems and Brilliant do a great job of making this fun and slightly

addictive, they have a way of framing questions that I get slightly obsessed with trying to

work out the answer to.

But that makes it really fun.

And they cover a bunch of different subjects: physics, mathematics, computer science and

many more.

So if you are interested in checking that out go to brilliant.org/dos the link’s also

in the description below, and otherwise, thank you for watching again, and I’ll see you

on the next video.