I have a massive database that needs quantum processing for the best optimization outcome. Can I just plug it in over the cloud?
NO.
Today's quantum information processing has not matured to connect to a database directly. There is a preparation phase to stage the data. It entails plenty of expertise and knowledge of circuit, gate, algorithm, and data processing knowledge. Furthermore, you must be an expert in your "problem" that seeking to solve. For an expert software developer, it may take about a year using low-level SDKs, (working with the expert in the "business problem", but there are other software solutions ($$$$) that can substantially speed up the process without knowing a lot about the quantum back-end. It means, if you know your "problem", you can ask and get solutions using some third party software. The key is to learn these tools.
What is the nature of your "problem?
Although quantum processors are highly efficient in solving some problems, not suitable for all situations; in fact, running specific issues on a quantum platform offers no advantage. For example, quantum computers are far more efficient in processing giant databases, especially optimization problems. Our best classical systems often choke when processing specific data optimization tasks. The difference in calculation time can quickly go from years with a classical computer to minutes with a quantum computer. For example, many classical optimizers use algorithms that result in a solution after running for weeks using a supercomputer. So in the near term, the most apparent and well-scoped problems around portfolio optimization, arbitrage optimization strategy, and training cost optimization. These solutions require quick results; weeks-long result-oriented solutions have no value in many cases.
For example, Arbitrage is the simultaneous purchase and sale of the same asset in different markets in order to profit from tiny differences in the asset's listed price. For that, you need high-speed computing far beyond classical supercomputers.
The way quantum computers work seems complicated, but data is prepared and delivered to the quantum computer thanks to classical computers. Once trained and submitted, QC place all the data into qubits (bits in classical computers), staying resident in the system while the QC circuit and gates are running and the data is processed. There are no database interactions as you would use a classical problem-solving project.
In a nutshell, the information prepared creates qubits submitted to the Quantum processor, leading to a solution. Currently, technology doesn't allow combining enough quantum processors or qubits to run large-scale solutions, although the time is approaching rapidly. Today, we can run smaller problems, test and evaluate quantum processors, such as Cold Atom, Superconducting, Ion Trap, photonics, or annealing hardware. Today's quantum information processing has not matured to connect to a database directly. There is a preparation stage to stage the data.
So what happens inside the QPU?
So suppose you prepare the data and submit it. Here is what happens next.
Here is a bit of spooky facts; unlike classical computers relying on binary 0 and 1 9for computation,) quantum computers are multi-dimensional. It means that "data" stays in different states in a multi-dimensional space. I know it is a bit hard to wrap your mind around this, but that is precisely what's happening.
A piece of data can be a 1 and a 0 simultaneously – depending on what's happening in the other dimensions that impact that piece of data, an oversimplified explanation of superposition.
So what happens, this "multi-dimensional space" I just explained can be "energized" to create probabilistic models of potential outcomes to business optimization problems. For example, if using an ion trap QPU to run your circuits, the system uses a laser pulse designed for a specific amplitude to flip the qubits and run gates.
All these processes require different sets of coding to manage the process. A third-party software product can speed up time to submit from a year to a few weeks. For example, I am learning the Cirq for superconducting submission and looking into Ocean SDK for the Annealing platform. I have a long way to go, but I am moving right along.
תגובות