I am Sergio Boixo from the Google AI Quantum team,

and today, we're going to talk about an experiment we're

working on, which is known as quantum supremacy.

The latest experimental quantum processor

produced at Google, Bristlecone, has 72 qubits or quantum bits.

We're testing quantum circuits in Bristlecone

with the goal of reducing errors.

By their nature, quantum gates have a probability of errors,

and errors can cross quantum circuits.

As we calibrate quantum circuits,

we bring down the probability of error.

We simulate quantum circuits with traditional computers

to benchmark and calibrate quantum circuits.

As we work to reduce the probability of an error,

simulations gets exponentially harder.

This means that it gets too computationally intensive even

for a supercomputer to keep up.

From this, we get the name quantum supremacy

for this experiment.

This has to do with something called a strong Church-Turing

thesis in computer science.

Traditional computers from the abacus to your laptop

implement equivalent operations or classical gates,

although a modern computer is, of course, much, much faster.

The strong Church-Turing thesis says

that all universal computers are equivalent in this way,

and can simulate each other efficiently.

But according to quantum computing,

the strong Church-Turing thesis is false,

and quantum computers can solve some problems exponentially

faster than other universal computers.

So what we're trying to do is kind

of breaking the strong Church-Turing thesis.

You can think of a qubit as an arrow pointing

to some direction on a sphere.

Quantum gates are operations on qubits.

Similar to classical gates, we often depict quantum gates

as boxes, with the input on one side

and the output on the opposite side.

In a quantum circuit, we apply layers

of gates, one per clock cycle.

A measurement at the end of the quantum circuit

produces a string of beats.

For the quantum supremacy experiment,

we choose the quantum gates at random.

This is a Hello World program for quantum computers.

Crucially, in this case, we have the strongest critical evidence

against the strong Church-Turing thesis.

It takes exponential time to simulate a random quantum

circuit with a classical computer.

According to quantum mechanics, every particle

can also act as a wave, and this applies to qubits.

The quantum state of a quantum computer

contains an exponential number of waves

or computational paths.

This is the property that we are testing.

The output state of a random quantum circuit

looks like the speckles of a laser.

This is a fingerprint of the quantum circuit.

For some bit strings, the computational paths

interfere constructively, and the intensity of the output

probability grows.

For others, the computational paths interfere destructively,

and the output probability decreases.

Simulating interference of the exponential number

of computational paths in the quantum circuit

takes exponential time.

We can check if we obtain the correct fingerprint

in the experiment, and measure the probability of error.

First, we get around a million bit strings

from the quantum computer.

This takes a few seconds.

Then we use an expensive classical simulation

to check if those bit strings have high probability.

If this is the case, the error rate is low,

and the experiment has succeeded.

The implication will be that quantum computers

seem to be breaking the strong Church-Turing thesis.

As we reduce errors farther, we expect

to see a similar exponential speed up

for a practical problem.

So what's next?

Visit the other videos in these series

to learn more about how a quantum computer works

and how to program it.

You can also visit OpenFermion to learn more about how

quantum computers can be used to solve problems

in chemistry and material science,

or check out the links included below.

Thank you.