Is digital music still music?
DecaQ’s DecaQuasar Oracle Quantum Computing
To be honest, The Quantum Dragon would prefer to listen to a bard who isn’t trying to slay him, but that’s neither here nor there. He has way too much treasure to ever make that dream practical.
What is quantum computing?
The Quantum Dragon has a record player in his cave. It still works. His favorite album, you guessed it, is “Puff, the Magic Dragon.” And the question posed by the fine folks over at DecaQ is whether or not the song ceases to be music the moment we digitize it and upload it to YouTube, Spotify, or what have you.
The vast majority of us should reply that it is, of course, still music.
And yet we debate this question in quantum computing circles. Is it only a quantum computer if there are analog qubits at its core? DiVincenzo Criteria or bust? Or is it also a quantum computer if CPUs or GPUs are performing literally the exact same computation? Or do we go back in time, before digital computers, when humans were computers by job titles, and ask if a quantum computer is a PhD candidate using paper and a pencil to solve the Schrödinger Equation?
I have observed heated arguments among physicists about this.
Digital Oracles
DecaQuasar by DecaQ falls in the middle. I’ve fed it OpenQASM, as I have many products that we would not argue about. I’ve been shown under the hood a little bit, and I saw a Bloch sphere. Have you ever seen a Bloch sphere in quantum computing? Yeah, it was one of those.
Grossly oversimplifying what I saw, it’s a new mathematical framework based on 10D geometry that they call Oracle Quantum Computing. The reasons given to me for the 10 dimensions are twofold: a) real orthogonal dimensions; and b) our math is based on decimal. This is where all of the “deca” naming comes from, by the way: DecaQ, DecaQuasar, etc. That said, I am leaving the mathematics to those who discovered it, and I’ll just be discussing how the thing works.
The Good
Scalable. The early version I used accepted quantum circuits with up to 200 qubits. The roadmap extends this to 2,000 qubits in January, and that’s just what’ll be accessible to you and me, not what they’ve got going on in the lab. All of this is supposedly being run on a laptop with 32 GB RAM, by the way, so the upper limit is way up there.
Fast. A 200-qubit Bernstein-Vazirani circuit executed in under 7 seconds, a 200-qubit Deutsch-Jozsa circuit in less than 4 seconds, and a 200-qubit iterative phase estimation (IPE) circuit in less than 30 seconds.
Correct. I generated the circuits myself, instead of using the provided samples, and I knew what the outputs should have been. As Gordon Ramsay might say, “spot on.”
Deterministic. It runs only once. It outputs one solution, so there’s no histogram. The answer you get is the answer, folks.
Pricing. Easy-to-calculate pricing ranges from $1.00 per qubit down to $0.50 per qubit.
The Bad
Unroll it. OpenQASM 2.0 has shortcuts that allow some massive circuits to require relatively few lines of code. I had a 200-qubit Bernstein-Vazirani algorithm with only 11 lines of code, but I had to unroll it into many more lines of code. Ironically, it works with OpenQASM 3.0, but I’m an OpenQASM 2.0 aficionado.
Limitations. My hands-on experience has been limited to the Bernstein-Vazirani, Deutsch-Jozsa, and Iterative Phase Estimation (IPE) algorithms. That’s not the limit of the software, though; this is a consequence of getting an early preview.
The Ugly
Scalability. This isn’t ugliness about DecaQuasar, but the company claims that 10,000 qubits with 1 billion gates have been simulated in the lab. The challenge, as we would have with a 10,000-logical-qubit quantum computer, is verifying the computation. Something like Simon’s algorithm can be verified, but something like quantum phase estimation (QPE) still needs to be checked somehow. A 312-qubit QPE executed in only 11 seconds, but I literally can’t run that anywhere else to verify it. But if it can be verified, this would seem to push the timeline for useful quantum computers way back.
Weaponization. There is no desire to demonstrate Shor’s algorithm, nor is there any desire to let anyone else do it. They’re very clear about that. But if you can quickly and accurately simulate thousands of qubits, and you can run Simon’s algorithm, the precursor to Shor’s algorithm, somebody’s going to be having impure thoughts and trying it.
Conclusion
The Quantum Dragon is still laughing at my use of the phrase “heated arguments.”
Gee whiz. He’s got a point, but it’s not THAT funny.
Anyway, DecaQuasar has been evolving as I’ve been assembling this article. Though the site offered me only 3 algorithms, it’ll eventually be universal. I’ve already seen QPE and QAOA, the latter of which is downright fascinating. The Quantum Approximate Optimization Algorithm (QAOA) is supposed to be approximate, as its name suggests, and yet DecaQ’s version solves problems deterministically. And quickly.
You may be wondering how the heck we’re supposed to generate quantum circuits large enough to use this thing? QuantumGear is due in January, and I’m looking forward to putting it to the test.
Finally, DecaQ would like to issue a challenge. Not to a firebreathing dragon, obviously, but to human researchers. What do you have that can’t run anywhere, but that you could verify a correct solution for? DecaQuasar made easy work of my entire library, and it’s all accurate at scales that I can test it. The question is: can it solve something of real commercial value?
Epilogue
DecaQ is confident in DecaQuasar’s accuracy because they can demonstrate a large-scale Simon’s algorithm and verify the correct answer. And to their point, it is accurate to the extent I can verify it. However, I have observed with other software how toy-scale accuracy deviates as problems scale. Based on that experience, I have to maintain some healthy skepticism until some chemists and materials scientists come along like medieval bards and sing its praises. Oh, how The Quantum Dragon would love that. Unlike the other software, however, which mimic probabilistic computation, DecaQuasar is deterministic and holding up thus far. Like a well-balanced party of adventurers, it needs a tougher challenge, though. Are you up to it?
