Quantum Supremacy: When Will Quantum Computers Be a Thing?

In 2019, Google announced that they had achieved quantum supremacy - but what does that mean? And does it even matter?

Hosted by: Hank Green

----------
----------
Huge thanks go to the following Patreon supporters for helping us keep SciShow free for everyone forever:

Kevin Carpentier, Eric Jensen, Matt Curls, Sam Buck, Christopher R Boucher, Avi Yashchin, Adam Brainard, Greg, Alex Hackman, Sam Lutfi, D.A. Noe, Piya Shedden, KatieMarie Magnone, Scott Satovsky Jr, Charles Southerland, Patrick D. Ashmore, charles george, Kevin Bealer, Chris Peters
----------
Looking for SciShow elsewhere on the internet?
----------
Sources:

Image Sources:

[andreaolivo]

Some corrections:
1) Not-really-important nitpicks:

a) Turing Machines (TM) can solve all _computable_ problems (so much so that the definition of a computable problems is kind of based on a TM)

b) Schrödinger's cat was originally about where are we supposed to draw the line between a quantum system and a classical one. In the original version, the cat (a supposedly classical being) is killed by a device that is randomly activated by the decay of an atom (that can actually experience superposition). Schrödinger was criticizing the scholastic interpretation of quantum mechanics by forcing their followers to admit that in this situation the cat had to behave quantumly and be in an absurd superposition of "alive" and "dead". Your description of the experiment makes it seem more a philosophical question than a scientific one.

c) Your description of the qubit was okay, but sadly the popular description of the inner workings of a quantum computer as "trying all the possibilities at once and choosing the right one" is just plain wrong. For one thing, it would give quantum computers the same power of a NTM (Nondeterministic TM), meaning that they could efficiently solve all problems in NP. That's not believed to be the case and has massive consequences on the known hierarchies of computational complexity.


2) More important points (IMHO):

a) "What counts as a practical amount of time"? This question is exactly why we work with complexity classes instead of amount of time. Quantum computer are NOT expected to solve _all_ "the same problems that classical computers can solve, but way faster", only some of them. In fact, at the moment we know of a surprisingly low amount of useful (or even useless) problems where these speedups are significant.

b) Quantum Computational Supremacy (a very poorly chosen word) has nothing to do with demonstrating that quantum computers can efficiently solve _useful_ problems, as you claim in the intro. I want to stress this point because showing supremacy is actually a fairly well-defined goal, contrary to what you said at the end of the video. It is reached whenever it can be experimentally shown that a given problem (even a completely useless one, but clear and concise) requires way less resources if quantum phenomena are exploited than on traditional, classical hardware. As is, the IBM response to Google's result doesn't lower a bit the complexity of solving their problem on a classical computer: instead it stems from an argument that in this case it is possible to trade cpu time for storage, if you have huge amounts (~100 Petabyte) of it. I however agree that it is quite a blow, and someone at Google's is probably banging his head against a wall for not factoring the possibility of exploiting such a huge amount of storage space into the calculations. Anyway, each qubit they add to the 53-qubit machine would translate in DOUBLE the space requirements for the IBM's proposal.

c) If the aim was just generating random numbers, as the video implies, generating a million of them in 200 seconds is not at all an impressive feat (even my phone could do it); certainly IBM wouldn't need hundreds of petabytes and 3 days on a supercomputer to replicate Google's result. And yes, for that classical computers are "just fine", as you say towards the end of the video.

d) And so we get to the crux of the matter, the problem solved by Google, called Random Circuit Sampling (RCS). What is it? It is quite technical, but a classical analogy with Galton boards (look them up) can be helpful. In a Galton board each time the board is flipped you end up with some random configuration of beads at the bottom, but you'll notice that it is not uniform; it has a generic "bell shape", and is called Normal distribution. In more formal language, Galton boards are computational devices that, while not having the full power of a Turing Machine, can still solve the problem of "extracting samples from the normal distribution". The device built by google is similar; you set it up in a specific way (the Circuit part in the name of the problem), feed a standard initial state at the input, run it and read a configuration at the output. As in the case with the Galton board, this configuration (sample) won't be uniform; it follows a specific distribution that depends on the settings of the device. Crucially, not only the specific shape of this distribution is extremely hard to compute on a classical device; it is even extremely hard to extract samples from it. But is this device as powerful as a full (quantum) Turing machine? Most probably not. Can it be used for something useful? We don't know, even if some proposals are out there (look up "Certified randomness generation"), but that was never the point.


3) Ending remarks
If you made this far, you're crazy (but I love you). These topics are fascinating and I glossed over some of the best parts (e.g. how they check that they have extracted good samples when there is no classical counterpart to compare the sample to, or why computer scientists are so convinced that a classical efficient algorithm for RCS cannot exist).

[JM-usfr]

"It can solve ANY mathematical problem"


Not any. Just computable ones.

[quito]

Google's claim of "quantum supremacy" is like their "negative latency" for Google Stadia.

[SciFactsYT]

Interesting fact: Bottled water expiration dates are for the bottle, not the water. 💧

[PopeGoliath]

Some day I'll watch a video about quantum things without getting the introductory superposition/cat spiel. Today is not that day.

[mrmeglomania]

Fun fact:
Magic: The Gathering can be made into a turing machine.

[macvelli]

Never.

I actually sat in the audience of a symposium once where a Comp Sci professor was honest about what the challenges are with quantum computers and how the most advanced thing they had ever been able to do with one was calculate the square root of a number.

The overwhelming issue with quantum computers is that you are trying to use the position of a boson inside the nucleus of an atom as if it were a transistor. The problem is trying to set a quantum bit to a desired value *and* have it maintain that value. Bosons naturally interact with the quantum environment inside the atom and hence they change their orientation within the atom accordingly.

Now imagine trying to build a computer on something that had no way to maintain state. Because that is the problem you have with quantum computers. They come up with various ways to *try* and control the orientation of the boson inside an atom but they end up having algorithms on the order of Big O notation N squared to verify if the boson is in the correct orientation or not and then there is only some probability of less than 100% that the boson is in fact oriented properly.

Would it be awesome if you could use a quantum particle inside an atom to represent state just like a transistor does? Absolutely. But it is a futile effort due to two factors a) the above-mentioned problem of maintaining state within the nucleus of an atom and b) even if that wasn't problem enough you have to have the atoms interact somehow such that you can build logic gates, registers, floating point units, etc i.e. all the things that actually make a CPU work properly. And as far as I know atoms do not propagate the orientation of bosons within their nuclei from one atom to another, nor do they have the ability to *change* that state like a logic gate does in electrical engineering.

So to be honest, I'm not quite sure how they can get a quantum computer to do *anything* that would be considered productive. Other than true random number generation. In fact this may end up being the greatest contribution quantum computing provides is a means of generating true random numbers which is critical for things like encryption and information security. All you would have to do is simply observe the orientation of bosons within a set of atoms at a given point in time and assign bit-like values to the positions you observe. Computer scientists have always had a problem generating true random numbers but this would be trivial with a quantum RNG.

[ridwansetiadi]

John: *asking who needs Quantum RNG
Me: "Hahaha."

100 years later
Streamer: "I guess Quantum RNG isn't with me. I'm so quantum unlucky in this game. Okay chat, let's do the math; let's check the quantum probability."

[TheRealGuywithoutaMustache]

6:06
This is absolutely mind blowing.

[colenoonan]

I can’t wait until these come out so I can run Minecraft with RTX without my computer blowing up

[asiburger]

Haha, Hank. Your pronunciation of Schrödinger made me crack up. :D

[seanspartan]

You guys should read up on Gödel's Incompleteness Theorem...

[ChrisSeltzer]

IBM pointed out another problem with Google's announcement: their quantum computer relied on a classical computer doing computation beforehand in order to work. The total amount of computation time was not measured in seconds but hours.

[businesschicken]

Episode Idea: " How Many Times Can We Make a "When Will Quantum Computers be the Norm?" Episode?"

[eamonahern]

Sounds like a cool album title from a prog rock or technical death metal band. To back up my point, the death metal band Beyond Creation opened up their song Fundamental Process with the line "quantum hypothesis leading to those mystical superpositions".

[terashewchenko]

Super interesting stuff!!!! Also that was the most pg version of Schrödinger's cat I have ever seen 😂😂😂😂

[MrCanuck]

Quantum computers will only be a thing when we actually observe it lol .

[HECKproductions]

field of mathematics that only works with 2 states: exists


computer inventors: hey thats easy to apply to electronics lets use that as a basis for literally all computing ever


quantum computer bois: imma bout to end this mans whole career

[tungstikum]

Wrong! The supremacy test was not about generating random numbers (another quantum computer generated the numbers) - the supremacy test was about confirming that the numbers generated were actually random.

[Bigfoot_With_Internet_Access]

Idk about you humans, but we bigfoots don't support computer racism