The incalculable volume of data that is handled today requires an ever-increasing computational capacity to extract its value, especially for the application of artificial intelligence techniques such as machine learning. As a result, research is underway on how to accelerate these procedures by applying algorithms derived from quantum computing to artificial intelligence techniques, originating a discipline known as’Quantum Machine Learning'(QML).

“Quantum learning can be more efficient than classical learning, at least for certain models inherently difficult to learn through conventional computers,” says Samuel Fernandez Lorenzo, a researcher in quantum algorithms who collaborates with BBVA’s New Digital Businesses (NDB) area. “It remains to be investigated to what extent these kinds of models are presented in practical applications.”

Machine learning and artificial intelligence technologies are the two main areas of research in the application of quantum computing algorithms. One of the characteristics of this calculation system is that it allows to represent multiple states at the same time, which is particularly suitable in the use of AI techniques. For example, one case highlighted from Intel is that of voice-activated assistants,which could improve their accuracy by increasing the amount of data they operate with and their computational power. Quantum computing could make it easier to calculate with more variables so that they can respond more similarly to how a person would.

More accurate algorithms

The ability to represent and manipulate so many states makes quantum computing very appropriate for solving problems in different fields. Some of the first applications already being researched are focused on fields such as material science,where small molecule modeling requires enormous computing capacity. Later, larger machines will allow the design of medicines or the optimization of logistics to, for example, find the most efficient route among any number of possible travel routes.

Currently, the largest industrial applications of artificial intelligence come from so-called supervised learning, used in tasks such as image recognition or consumer prediction. “In this area there are already several proposals of QML that anticipate an acceleration—which could become exponential—of some of the most famous algorithms in this field, such as ‘support vector machines’ (SVM), and some types of neural networks,” explains Fernández Lorenzo.

One of the features of this calculation system is that it allows to represent multiple states at the same time, which is particularly suitable in the use of AI techniques

A less explored territory but with great possibilities is found in unsupervised learning. “One particular case is the dimensional reduction algorithms, by which we represent our original data in a smaller space, but still retains most of the properties of the original ‘data set’.” At this point the researcher emphasizes that quantum computing is especially suitable for finding certain global properties of a certain datasheet,and not so much its specific details.

Finally, there is still a long way to go in the area of reinforcement learning to introduce it to the practical problems of the industry. Its potential to deal with complex situations has been evident in applications in the area of video games. The most demanding part here occurs during algorithm training, which requires a lot of computational power and time. “In this context,” adds Fernandez Lorenzo, “some theoretical proposals have already been put forward to accelerate that training using quantum computers, which may help us deliver a very powerful artificial intelligence in the future.”

Applications in the banking sector

In the financial sector,combining AI with quantum computing would help improve and combat fraud detection. On the one hand, models trained with a quantum computer would be able to distinguish patterns that are difficult to detect with conventional equipment. And, at the same time, algorithm acceleration would allow more information to be used and processed than is currently used for this purpose.

Quantum computing: what can it bring to the financial sector?

With uncovered potential, quantum computing gradually makes its way into banking thanks to the different solutions it can provide in services such as product customization and financial studies.

Work is also under way to develop models that allow numerical calculations to be combined with expert opinion to make final financial decisions. One of its main advantages is that they are models “more easily interpret-able than neural network algorithms, so they would have fewer regulatory impediments,”  researcher says.

In addition, one of the trends of banking is to offer products and services increasingly personalized to its customers using recommendation systems. In this sense, there are already proposals for quantum models that could accelerate their performance. “It doesn’t seem foolish to think that this sector can soon offer us investment strategies based on quantum ally inspired algorithms,”Fernandez says.

To get to this point, researchers are turning to discover how to leverage the capabilities of today’s quantum processors, exploring the connections between the newly proclaimed quantum supremacy and machine learning. “In particular, the quantum advantage here could lie in the possibility of building models that would be very difficult to implement on a conventional computer. The applicability of such models in real industry contexts remains to be investigated,” concludes the researcher.

Leave a Reply

Your email address will not be published. Required fields are marked *

Close