I did the math seven years ago, when I had just finished my master’s degree. At the time, there was both excitement and skepticism on the scene. Quantum computing currently stands out as an emerging technology alongside HPC and AI.
The focus has shifted from solely hardware-related research and discussion to applications, software, and algorithms. Quantum is actually not an isolated field, but a tool that can be used across many different fields. One promising but still poorly understood application of quantum computers is quantum machine learning.
Quantum Machine Learning (QML) has become an umbrella term over the past few years. One of the earliest and most important emergences of QML was in 2013 when Google and NASA established the Quantum Artificial Intelligence Laboratory. The institute was tasked with exploring how quantum computers could be used in machine learning applications. Since then, the term QML has appeared in research papers, startup pitches, and conference talks, often with vastly different meanings.
In some cases, it refers to using quantum computers to speed up machine learning. The other parts describe classical algorithms inspired by quantum physics. In other cases, it may simply mean running familiar ML workflows on unfamiliar hardware.
So, even as someone who works and researches quantum computers, I was very confused at first…The first question many people ask when they hear the term “quantum machine learning” is, what exactly is quantum machine learning? quantum?
Answering this question is why I decided to write this article. Simply put, it’s not about speed, it’s not about neural networks, it’s not about vague references to “quantum advantages”. The core of quantum machine learning is defined by how information is represented, transformed, and read. In QML, this is done using the rules of quantum mechanics rather than classical calculations.
This article aims to clarify that distinction, separate substance from hype, and provide a clear conceptual foundation for the rest of this series. I plan to explore the lore of QML and write about its short-term research results and applications.
Machine learning before “quantum”
Before we get into what quantum is all about, let’s take a step back. Machine learning, stripped of its modern trappings, is about using data to learn mappings from inputs to outputs. Whether the model is a linear regression, kernel method, or deep neural network, the structure is almost the same.
- Data is represented numerically (vectors, matrices, tensors).
- A parameterized model transforms that data.
- Parameters are tuned by optimizing the cost function.
- The model is statistically evaluated on new samples.
Neural networks, GPUs, and large datasets are implementation choices and do not define functionality. This abstraction is important because it allows us to ask precise questions.
When will what change? Data and models exist in quantum space?
The advent of quantum mechanics
If quantum information is the basis of computation, then quantum machine learning becomes quantum. This manifests itself in three ways.
1. Data is represented as quantum states.
In classic machine learning models, data is represented as bits or floating point numbers. In contrast, quantum machine learning uses quantum states, which are complex vectors that follow the rules of quantum mechanics. These states are often described by density matrices, and their transformations are represented by unitary matrices.
As a result, information is encoded in complex-valued amplitudes rather than probabilities, and states can exist in a superposition.
this will be do not have This means that all your classical data will suddenly become exponentially compressed or easily accessible. Loading data into a quantum state is often costly, and the extraction of information from a quantum state is fundamentally limited by measurements.
Therefore, the important point is that the model operates in quantum states rather than in classical numbers.
2. The model is quantum evolution
Traditional ML models apply functions to data. Quantum ML models apply quantum operations (typically unitary transformations) on quantum channels. In practice, many QML models are constructed from parameterized quantum circuits. These circuits are a series of quantum gates, the fundamental operations that change the quantum state. The parameters of these quantum gates are adjusted during training, similar to adjusting the weights of a neural network in classical machine learning.
Essentially, what happens in these models is that we start with the state of the system represented by a matrix (more accurately called the Hamiltonian), and then the gates we apply to the system tell us how that system will evolve (change) over a given period of time. Its evolution determines the behavior of the model.
As a result, quantum models explore a hypothesis space that is structurally different from classical models, even though their training loops may appear similar on the surface.
3. Measurement is part of the learning process
In classic ML, reading the output of a model is easy and does not affect the state or behavior of the model (unless you do it intentionally). But in quantum ML, measurements are probabilistic and destroy states. This has a significant impact on the system. The output is determined by repeated executions of the circuit called “shots.” Here, “shot” means running the same quantum circuit multiple times to estimate the result, since quantum measurements are probabilistic.
The gradients (which guide parameter updates during training) are statistically estimated from these measurements, rather than being calculated exactly as in traditional machine learning. As a result, training costs are often dominated not only by computation but also by sampling noise from these repeated measurements.
In other words, there is uncertainty built into the model itself. Any serious discussion of QML must take into account the fact that learning occurs. through Measure, not after.
what to do do not have Create a QML quantum
Quantum computing and QML in particular generate hype and misconceptions. Much of what is called “quantum machine learning” today is quantum in name only. For example:
- Classical ML algorithms run on quantum hardware without making meaningful use of quantum states.
- A completely classic “quantum inspired” method.
- A hybrid pipeline that allows you to remove quantum components without changing model behavior or performance.
If you meet someone talking about QML and you’re not sure how quantum the model they’re discussing is, a good rule of thumb to follow is to ask:
“Can we replace the quantum part with a classical one without changing the mathematical structure of the model?”
If “yes” or “probably” then the approach is probably not fundamentally quantum. While this research may still be valuable, it is outside the core of quantum machine learning.
Where is QML now?
When discussing quantum computing, remember that current hardware is noisy, small, and resource-constrained. For this:
- Currently, there is no general and proven quantum advantage for machine learning tasks.
- Many QML models are more similar to kernel methods than deep networks.
- Data loading and noise often dominate performance.
This is not a failure in the field. That’s where quantum computing is now. Currently, most QML research is exploratory, such as mapping model classes, understanding quantum learning theory, and identifying where quantum structure may be important.
Why quantum machine learning is still worth exploring
Why pursue QML if you can’t expect speedups in the short term?
QML forces us to rethink fundamental questions about machine learning and quantum computing. We need to answer what it means to learn from quantum data, how noise affects optimization, and which model classes exist in quantum systems but not in classical systems.
Quantum machine learning is less about outperforming today’s classical ML and more about expanding what “learning” means in the quantum world.
This is important because advances in science and technology begin with new approaches. Even if your hardware isn’t ready yet, considering QML will prepare you for better hardware in the future.
Final thoughts and future developments
Advances in quantum computing are accelerating. Hardware companies are racing to build fault-tolerant quantum computers. A quantum computer that takes full advantage of the capabilities of quantum mechanics. Software and application companies are exploring problems that quantum computing can meaningfully address.
However, today’s quantum computers cannot run nearly full-scale applications, let alone complex machine learning models. Still, the efficiency potential of quantum computing in machine learning is very interesting and worth considering now in parallel with advances in hardware.
This article focuses on the definition and boundaries of quantum machine learning, paving the way for future articles that explore:
- How classical data is embedded in quantum states.
- Variational quantum models and their limitations.
- Quantum kernels and feature spaces.
- Optimization challenges in noisy quantum systems.
- Where quantum benefits could conceivably emerge.
Before asking whether quantum machine learning is useful, we need to be clear about what it actually is. The further you move away from the hype, the closer you will be to progress.
