Quantum One Step at a Time EP-02: Making Sense of Logical Qubits

Hey everyone. Welcome to another episode of "Quantum, One Step at a Time," featuring Alan Shemel and my friend and long-time collaborator, John Willis, who actually is working on his latest quantum book. Hey John, welcome back.

It's good to see you, my friend. Yeah, no. Great to get another one in the queue here.

Absolutely. Look, before we talk quantum, I can't help but notice the Knicks hat. Yeah.

I got my NBA subscription so I can watch all the games. We're 0 and 3. Yeah, geez.

Yeah. I don't know if it was a last year thing and the magic-- Has Cinderella, has the clock struck midnight? It's way too early.

Way too early. All right, the optimist. So, John, we had an interesting month in quantum, actually.

There was a lot of different announcements. A lot of it was market noise and money noise, but there were some interesting technology announcements. Google announced, I don't know if you want to call it a breakthrough, because if you got to throw away 92 out of every 100 attempts, is that really a breakthrough in quantum computing?

But can you make sense of this for us? What are we talking about? Yeah, I think, it was a tech field day last week, right?

Mm-hmm. And they had one of the vendors postponed, so they asked us if we wanted to give presentations, and I said, "Yeah, I'll give a quantum one," right? And I gave a quantum one, and the title of my presentation was, "The good news is quantum is five years out.

" Has it changed? Yeah, that's the question. The trillion-dollar, billion-dollar, whatever the total market is.

But I think that it's progress. Like all the stuff, the Google, the Bitcoin scare that we had earlier, that they could basically break elliptical curve cryptology, the 256-byte, which is all that stuff that happened earlier this year scared everybody about Bitcoin is going to get destroyed. They're simulations.

We talked about this in the last podcast. This 92% thing is interesting, and what I do is I think if I compare it to the history of neural networks. And I went back, and I looked.

Basically, in 2011, there was a 25%. That's 2011 That's a little over a decade ago. And we're at 25% error in terms of the AlexNet, well, before AlexNet, the ImageNet.

Sorry, ImageNet recognition. And then by 2017, when we get to attention is all you need and reinforced learning, we're under 4%. And what happened there was Nvidia.

There was this inflection of technologies it hit. So I think progress is progress. And so to look at it as are we getting better in quantum?

And it is every day. Have we proved that we can solve the really hard problems yet? Absolutely not.

We talked about this on the last podcast. I asked everybody who seemed to have good answers at the IEEE Quantum Conference back in last month, and there is no example of anybody can show or tell me a story of a quantum breakthrough. In fact, they've even sort of changed the terminology from my 30,000, 40,000-foot view looking in.

What they're calling instead of quantum supremacy or quantum advantage, it seems to be now utility scale. So, they're even downgrading the terminology for what's going to happen in somewhere five years. But again, I think there's too much emphasis.

So I'll give you one more story. One of the gentlemen, I'm sorry, I forget the name, was a really smart guy who you probably know. He was asking me really hard questions during my quantum presentation at Tech Field Day.

" Quantum mechanics. So I think there's a firm belief in everybody that's part of this history that literally the side of the fence, and you could get a little bit of a historian now, but guys like Heisenberg or Niels Bohr, they believed the math worked, and the math does work. The question is, can we actually get computation to be fault-tolerant at a scale where we can solve these polynomial time problems?

That's the question. But there you go with a couple of terms there that I- There you go again. I sound like Ronald Reagan.

I got it. So John, as funny and as crazy as this is going to sound, for many people, it's easier to get their heads around Einstein's theory of special relativity- Mm-hmm ... than it is to get their heads around Niels Bohr or Max Planck's or some quantum kind of math and physics.

Because they seem so diametrically opposed, where relativity, it's almost logical, quantum almost to me defies logic. And then we start getting into the naming of things. We have qubits, then we have fault-tolerant qubits, then we have logical qubits, then we have illogical qubits.

That sounds like something Mr. Spock talks about. Give us a simple primer.

What's the deal with qubits? Yeah. A debate broke out during my presentation at Tech Field Day, and I keep going back to it.

" Richard Feynman, who basically did the sort of canonical presentation and paper on describing what a quantum computer could look like, said that if you think you understand quantum mechanics, you don't understand quantum mechanics. So it is incredibly illogical to humans who think deterministically. So here's the thing.

Just a short history. Quantum computing is made up of three things. I think I might be repeating myself from last time, but primarily three things.

An apology to quantum physicists because me and Alan are not quantum physicists, is it's the notion of a superposition that instead of a bit being on or off, it can be in many positions, or it can be both zero and one, and it can be more complicated than that, but let's leave it at zero one. And then there's the entanglement, that you have two qubits that are entangled means that if you know what one is, and it's just that you know what the other one is, the probability is incredibly consistent. The math is incredibly consistent that if one's zero, the other one will be one.

Okay. And then you have this measurement problem, which is when you observe it, it destroys the state. So quantum computing is all about this sort of superstition, entanglement and the idea that when you observe it, you've lost it.

So Peter Shor- Well, by observing it, you've destroyed the pureness of it or whatever. Yeah, you can't, it's what they call the known copy. It's why the encryption and all that, the teleportation, all that stuff is so fascinating, but we'll save that for another podcast.

" And it was Peter Shor who sort of changed the world because he basically created Shor's algorithm, which showed that you could possibly use this quantum computer that really didn't exist yet to basically figure out really complex factoring problems, which basically could break RSA encryption theoretically. And the long story short is that when you use a quantum algorithm, what you're doing is you're literally creating, again, for apologies to quantum physics, a wave function. And basically, you're not using parallel or byte, sequential or even parallelism.

You're literally creating a probability spectrum of circuits in a quantum, either supercomputer or photonic-based. And when you're measuring it, you're trying to figure out what the probability was of the right answer. So not to go too far into the Shor's algorithm, but basically, he figures out that you can use this sort of math that had been around 400 years old, you could literally use a quantum Fourier transform.

And then there's a possibility that you could literally break encryption. What's interesting here is, not only did Peter Shor change the world by saying, "I think this is possible. Here's the math.

We know the math works. " Some of the criticism he got about all the noise that had been recognized since the 1920s of how these waves work, because there's all sorts of what they call decoherence, sort of temperature, vibrations, all these things, because you're not emulating, you're creating quantum natural things in like photonics or superconductors. And so there's all this stuff that gets in the way.

And so what's ironic is, not only did Peter Shor define how we might break encryption, he actually created the algorithms for error correction. So this is part of my history book. And that's where we start the conversation that turns into logical and physical qubits.

And here was the nut. The problem was that if you think about Shannon's work of information science and stuff like that, what do we do? We could create multiple bits and we could say the parity, in other words, we could literally say if one of them's off, maybe two win the majority.

But you can't do that with quantum because the minute you observe it or you can't copy them, you destroy the state. So what he came up with is creating this idea of a logical qubit that could understand the state of the physical qubits. And so he built basically a nine physical qubit structure with a one logical qubit, and then what he was able to do is break it up into, and then this is where he uses entanglement.

I know this is probably getting crazy for everybody, but he uses entanglement, so he had pairs of three qubits, three pairs of three qubits, and where now he can basically check the entanglement states. And if one was off, then the quorum, and the quorum again is not probably a good quantum physicist term, quantum mechanics term, but that's how he built this. And so what's the weird thing is you have these logical qubits that are things that represent or encapsulate what the physical qubit is.

So you have a lot more physical qubits than you have logical qubits. And I think we talked about this last time. So you might have one logical qubit to 10 or even 100 ...

physical qubits. And it depends on the modality, the modality being photonics or superconductor or ion-based. And then even the algorithms themselves.

And then you fast-forward a little bit. I thought this is interesting, so just to get my history stuff in. There's a guy at IBM, he's 83 years old.

He still works for IBM. He's one of their earliest computer scientists. He won the ACM Turing Award this year.

I want to interview this guy. He's 83, and he is one of the guys that started what they call the NISQ, the Noisy Qubit System. And by the way, he's the guy that literally invented the first conversation about QKD, quantum key distribution.

He's still a fellow at IBM, working at 83. God bless him. Yeah.

No kidding. God bless him. But then you get into the, I know you mentioned magic state, which is a way to build the algorithms in a certain way, and I know I'm high level.

And then you get into what Google did with Willow, and then Quantanium. So here's the nut. That's the quick tutorial on logical physical qubits.

Where we stand today is we still suck right? The bad news is, we're st- Oh, I was going to say, we suck less is the good news. I think, well, no, that we're getting better at showing the math.

And then here's the thing where it gets really interesting is there's a mid-market now of a bunch of companies that are literally focusing on that part of the stack. So I think that's going to help accelerate. And then you got those companies, for the most part, are using neural network acceleration GPUs to try to get better at error correction.

And then, DeepMind just created an alpha qubit. And think about all the things they've done in protein simulation and just across the board. All the alpha.

So, the good news is we know the math works. Neural networks and GPUs, what neural networks do really well or GPUs do really well is matrix multiplication. That works really well for creating better algorithms that can basically get better at this error correction.

But still the problem is we need very low error rates at high qubit, and that's still a scale problem that hasn't been demonstrated. Everybody talks about 30/30, or now 30/33 with DARPA. It's not less than five years.

It's probably maybe within a five to 10-year period that something significant probably will happen. And here's the nut. When it happens, in my opinion, it's going to be equivalent to decoding the DNA sequence.

It'll be IP, and crypto will be important, but it's the optimization algorithms. We talked about that. Right.

People are going to get patents on optimization, and they'll own the toll booth to a new type of plastic or a new type of battery. Those type of things. Materials.

Yeah. Those kind of things are going to be incredibly important to own the patent on those. Absolutely.

" He said, "You don't realize it, but we're living through a new Big Bang, the second Big Bang. First one was about 14 billion years ago. The new Big Bang is just beginning with AI.

" John, we'll go into more of this in future episodes. But the days are long, but the years are short. Yeah, I think that's the premise, the old saying.

And I think that's what we're in now. When you look at it day to day, week to week, it does seem like we're taking baby steps. It does seem like we're making very incremental progress.

But when you look over the arc of time- Yeah ... you see that there's major things going on. And the other thing to close out this week's show is I want to follow the money, because- Yeah ...

clearly people with a lot of money are putting their money where their mouth is here, right? We saw DARPA pick six more finalists, six modalities this week to add to the two they already had. 3 billion.

All stock deal- Right ... I should mention. It's all stock.

And the day after they announced that, I think they announced a $2 billion equity placement at $40 a share or something like that. What I find interesting there, John, is the connection between cyber and quantum. Yeah.

And this isn't just about post-quantum- No ... cryptography. It's part of the story, and it is the bigger part of the story now.

But the bigger part, again, I think in the end, when it's all said and done, it is going to be the IP on optimization. And just go back to that idea about, are we in a Big Bang? The original byline of IT Revolution was something that I created where we called it the Cambrian Explosion.

And it's since been taken off the website. But originally when Gene and I and a couple of us were creating the first ... sort of rollout of the DevOps handbook and IT revolution, I had been reading about Hal Varian, who was the chief economist at Google, and he had this article he wrote where he talked about being in the Cambrian explosion, this digital age.

So to your point about the day being long, you don't get to open AI, and you don't get to what Anthropic's doing without DevOps. You don't get to DevOps without sort of what Google and Amazon did in the '90s. So there's- And you don't get to that without all the fiber that was laid in the ground.

Right. Exactly. So I think this idea of that it isn't just AI, there's a window that started in the '90s, I think, and we can sort of debate where it started, that literally has...

" I think in the end, it's going to be somewhere between 1990 and who knows, 2030, 2040. And that's going to be the Cambrian explosion. Yeah.

That was the beginning of that eon or aeon or whatever they call those archaeological days, geologic eras. Yeah. That's right.

Okay. " What if I told you the genesis of all this was breaking AT&T up in the '80s, giving birth to the whole ILEC, CLEC, open telecommunications that empowered the commercial internet, a trillion dollars worth of fiber that all these companies are built on, and a lot of the fundamental research which came out of Bell Labs and came out of- Yeah ... a lot of these institutions that were out there that we are now the inheritors of.

Right? And that fiber and those phone lines, they themselves are mostly laid in railroad tracks that were given railroad easements back in the 1800s. Yeah.

Right? So everything is built like this, but clearly we reached some sort of spontaneous combustion point, let's call it, in the '90s. Yeah, I think there's some- That we are- ...

inflection points, right? Yep. And there's inflection points that are bigger than others, right?

I think certainly the telecommunications, fiber, all that stuff that was laid out, the fact that... Again, I think if you sort of mapped it, you could go back. Again, the first neural network is, what, 1947, right?

Yeah. Right. But where is the major inflection point for AI?

It probably has to be attention is all you need, in what, 2017, right? Which is... But I think when we look back, I think the ingredients of all this stuff is going to look like somewhere between 1990 and 2030.

But I think it's going to look like a primordial soup. Oh, big time. Right?

Oh, no question. Oh, yeah. No question.

Yeah. And then what creatures come arising out of that soup? And that's why that's important to say it isn't just AI, it isn't just quantum.

It's going to... Because I think the first breakthrough, we talked about this last podcast, is going to be things like quantum neural network or hybrid- Right ... CPUs, GPUs, and QPUs that work in concert together, right?

Those are going to create the first sort of significant breakthroughs. And so then what is it? Is it AI?

Is it quantum? Is it just networking? It's this- It's the future ...

big thing that happened. Yeah, that happened in a 50-year period. It's the state of technology.

Yeah. Hey, man, we got to pop off. We're a couple of minutes over.

But John, we got to get on schedule here. Next week, I'm out. But the next week, I'm back here in, so we'll be on in two weeks.

We'll start preparing our quantum to-do list by then. Yeah. Totally.

Yeah. If we do it, then it really, we might be messing it up, right? Because if we know it exists, but- Yeah.

There you go ... a little quantum joke for the end of the podcast. There you go.

You got any talking or anything you're doing over the next couple of weeks? Yeah, I'm going back to, I got invited back to the Tech Field Day, which I think it's next week, yeah, which is more of the infrastructure. So I'm better at the sort of AI software side, but I guess they think I'm worthy of like popping my head in on the infrastructure stuff.

Yeah, you've got a few qubits in there. Yeah, there you go. Yeah.

Anyway, thanks so much. Thank you for watching. As I mentioned in the beginning, we try to do this every other week.

It's "Quantum, One Step at a Time" with Alan Schimmel and John Willis. You can catch this on Apple Podcast, Spotify, Techstrong TV, Techstrong TV YouTube, or the OTT app for Android, iOS, Apple TV, Roku, Amazon, wherever you got a screen. Right.

John, my friend, I'll talk to you soon. Be well. Sounds great, Alan.

Thank you. Thank you. Bye-bye.