The Potential for Quantum Machine Learning in Industry
How Quantum AI can be used to profit businesses in the future
The future is now, and Artificial Intelligence is all the rage. Machine learning (a subsection of AI) has become such a popular topic of research that there are countless papers and examples of its applications on the web — discussion of neural nets, pruning methods, transformer models, and more. Similarly, Quantum Computing has become a new hot topic in the technology field, with companies like Google and IBM conducting extensive research with their own quantum computers, and numerous papers being written which explore its potential. Even smaller consulting companies like Accenture do research in the realm of quantum, with quantum supremacy becoming more evident every year and its uses meaning greater profit for businesses everywhere.
Overview
So what do you get when you combine AI and QC? Quantum machine learning, of course. It is an entirely new field of study with a lot of potential for development and research. There are two broad classes of QML: methods for running on a quantum computer (this involves preparation, storage, and processing quantum states to retrieve the classical solution) and methods designed to run on quantum annealers that solve optimization problems (this involves the physical evolution of quantum systems according to adiabatic quantum computing). Adiabatic QC is a form of QC which relies on the adiabatic theorem to do calculations. It is closely related to quantum annealing, which is a metaheuristic optimization algorithm that leverages quantum effects to solve quadratic unconstrained binary optimization (QUBO) problems.
Those are a lot of big words, and if you’re anything like me, you might think this to be too nebulous for your understanding. However, what it really boils down to is the idea of classical Machine Learning algorithms and practices being run on a quantum computer, using quantum mechanical concepts which allow for improved efficiency and solutions to existing problems unsolvable by classical methods.
Speculation
Now that we have a better understanding of QML, let’s take a look at the following Venn Diagram below:
Obviously, it’s missing components (I did make it in a midnight Zoom meeting as a joke for a friend). In hindsight, I should have replaced Cryptography with Information Security as a whole, since Crypto is really just a subsection of InfoSec. QCrypto is the concept of quantum encryption algorithms. More specifically, there is QHE, which refers to Quantum Homomorphic Encryption, a form of encryption which allows one to perform calculations on encrypted data without decrypting it first. You can read more about it here (it really is a fascinating topic, but I digress).
Theoretical discussion aside, what are some actual proposed applications of QML in industry? How could businesses (big or small) use QML to solve real-world problems in the modern day or near future?
Well, as previously stated, QML is a developing field, which means there isn’t a lot of evidence of its commercial use right now. We know the University of KwaZulu-Natal’s quantum research group is investigating how quantum theory might improve machine learning and vice versa, and we suspect that if the quantum marketplace continues to grow at its current rate of development, consistent enterprise use of QC and QML is an estimated two to five years away. However, businesses can start innovating now by utilizing existing quantum hardware platforms and software applications, like IBM’s quantum cloud service.
In other words, it’s mostly a matter of speculation — hence the “???” in the center of the diagram. However, there is one such field where it could be extremely useful, a field with which I am more readily familiar, and that is Information Security.
QML in InfoSec
Cybersecurity is fundamental to any business or organization because maintaining the security of a system is vital for its CIA (confidentiality, integrity, and availability, also known as the Security Triad). If any of these components are compromised, productivity dips, personal identifiable information is lost, and the business suffers.
QML could benefit businesses involved in InfoSec because quantum communication networks are much faster and more invulnerable to cyber-attacks, thereby protecting the confidentiality and integrity of stored and transmitted data. Quantum computers have already shown the ability to crack certain encryption algorithms like RSA cryptography, thereby showing the need for quantum-resistance encryption. Likewise, classical ML has shown proficiency in InfoSec tasks like identifying anomalies in network traffic or user behavior and detecting fraud online. This proficiency would only increase when coupled with quantum supremacy.
In classical ML, there exists the innovative GPT-3 (Generative Pre-trained Transformer 3) created by OpenAI, a research business co-founded by Elon Musk. It has been described as the most important and useful advance in AI in years. GPT-3 generates text using algorithms that are pre-trained — they’ve already been fed all of the data they need to carry out their task. Specifically, they’ve been fed around 570GB of text information gathered from the internet like the entirety of Wikipedia.
There is already research being done about this Transformer model’s applications in cybersecurity. For example, we can see GPT-3 is capable of generating code for a machine learning model just by describing the dataset and required output (see here). GPT-3 would therefore also be capable of writing scripts which would perform tests, peruse file systems, run exploits, and attempt attacks to find vulnerabilities. In theory, it could also write malware, which is both thrilling and terrifying to imagine.
All these applications are possible with QML as well, and they would be performed at an exponentially faster rate. All in all, this would mean improved efficiency for businesses in making their systems secure and ensuring productivity.
Disclaimer
Perhaps I should have prefaced with this, but I am not an expert in either Quantum Computing, Machine Learning, or QML. I am just one who studies these concepts and has managed to befriend scholars who have spent more time researching these topics.
That being said, I have provided a list of references below which includes research papers and articles that offer further explanation on the aforementioned topics.
References:
“Quantum Computing Methods for Supervised Learning” (Viraj Kulkarni, Milind Kulkarni, and Aniruddha Pant)
“what is quantum machine learning and how can it help us?”
