Who will win the race to model complex systems?

Technology companies have been investing billions of dollars in the development of quantum computers for many years.expecting them to bring breakthroughs in areas such as finance, drug development and logistics. Expectations are especially high for quantum computing in physics and chemistry, where the unique effects of quantum mechanics can manifest themselves. In theory, these are areas where quantum computers could be significantly superior to classical machines.

However, despite the enormous efforts in developing complex quantum equipment, a new contender has emerged that is already moving in the most promising directions. Artificial intelligence (AI) is being used extensively in areas such as fundamental physics, chemistry and materials science, raising questions about whether quantum computers can maintain their “dominant” position in these areas.

According to Giuseppe Carleoa professor of computational physics at the Swiss Federal Institute of Technology (EPFL), the speed and complexity of quantum systems that can be simulated by AI are increasing every day. In a recent article published in the journal Sciencehe showed that methods based on neural networks are becoming leading for modeling materials with strong quantum properties.

With rapid advances in AI, more scientists are wondering whether AI will be able to solve much of the most exciting problems in chemistry and materials science before large-scale quantum computers become a reality.

“The emergence of these new machine learning approaches is a major blow to the potential of quantum computers,” Carleo says. “In my opinion, companies will sooner or later realize that their investments in quantum technologies are not justified.”

Exponential problems, limitations and applications

One of the main promises of quantum computers is their ability to perform calculations with exponentially faster speed compared to traditional machines. This is achieved thanks to their information processing capabilities based on the principles of quantum mechanics. However, to fully realize this benefit will require much more powerful quantum hardware than what currently exists.

Today, the largest quantum devices contain thousands of qubits. For a quantum computer to effectively compete with classical machines, it will require hardware with tens of thousands, and perhaps millions, of qubits. But even with current limitations, once sufficient power becomes available, quantum algorithms such as Shor's algorithm for breaking cryptographic ciphers will be able to solve problems much faster than classical computers.

This opens up many opportunities, especially in areas such as data security, where the use of quantum algorithms can fundamentally change the way information is protected.

Jay Gambetta from IBM says the main problem with quantum computing today is noise — random errors that arise during the operation of quantum bits (qubits). Because of this, many quantum algorithms face serious difficulties. However, the situation is gradually improving thanks to new developments in error correction and increasing the stability of quantum systems.

For quantum algorithms that may have commercial applications such as database searching, task optimization, or AI support, so far the speed advantage remains negligible. Research conducted by Matthias Troyer of Microsoft shows that quantum computers today are much slower than modern classical chips. This limits their use in tasks requiring fast calculations.

However, quantum computers may be effective in modeling chemical and materials science processeswhere quantum effects are important, since these systems operate according to the laws of quantum mechanics.

Many critical processes in chemistry and materials science, such as the synthesis of new materials or drug development, depend on the quantum interactions of particles, especially their electrons. Quantum computers can help model these interactions, accelerating the discovery of new molecules and the improvement of existing materials.

However, modeling quantum systems is not an easy task. In systems such as molecules, the behavior of particles is difficult to predict due to the phenomenon of quantum entanglement, in which particles can be linked over large distances. This requires complex mathematical techniques that become exponentially more difficult as the number of particles increases.

However, not all quantum systems are difficult to model. In weakly correlated systems, particles do not interact with each other as much, which simplifies the modeling task. Most chemical and materials science processes are weakly correlated, and for such systems there are already classical methods such as DFT (density functional theory) that can give accurate results.

For highly correlated quantum systems (such as superconductors or ultra-precise sensors), classical methods cannot cope with modeling their behavior. These are those systems where the interaction between particles is so strong that standard approaches cannot be applied.

The role of artificial intelligence in quantum chemistry

More recently, artificial intelligence has begun to be actively used to model chemical and biological systems. Neural networks trained on data from DFT can predict the properties of molecules with much less computational effort than traditional methods. This greatly expands the scalability of the models, allowing systems with more than 100,000 atoms to be simulated.

This approach not only reduces computational costs, but also opens up new opportunities for chemists and materials scientists. For example, tasks such as optimizing chemical reactions or developing new materials for batteries become completely solvable.

However, as Alexander Isaev from Carnegie Mellon University notes, The main problem for AI remains the lack of data. To train neural networks, huge amounts of data are needed, which are not always available. However, the use of AI to model chemical systems continues to evolve.

In 2017, Giuseppe Carleo and Matthias Troyer demonstratedWhat neural networks can be useful for modeling highly correlated quantum systems. This approach is similar to how AI learned to play chess or Go using only the rules of the game. By applying this approach to quantum systems, neural networks can search for the minimum energy state of a system, which helps understand its properties.

This method is actively used to simulate complex quantum systems, including systems with exotic properties such as superconductivity. The advantage of neural networks is their ability to compress complex information about quantum states, making it possible to solve problems that previously seemed impossible.

However, as Stephanie Ciszek of the University of Ottawa points out, there are uncertainties about what tasks neural networks will be able to solve effectively. For example, for some highly correlated systems, neural networks perform well, but in other cases the computational cost unexpectedly increases.

The Future of Quantum Computing

One of the most promising areas for quantum computers will be the simulation of quantum systems that evolve over time, giving scientists a new tool to study areas such as statistical mechanics and high-energy physics.

As quantum computers become more powerful and stable, they will likely fill an important niche for solving problems that cannot be solved using classical approaches. However, for most current problems in chemistry and materials science, classical methods and AI will likely continue to dominate.

Conclusion

Currently, artificial intelligence is significantly ahead of quantum computers in speed and availability for practical applications, especially in chemistry and materials science. AI is already actively used to model complex systems and predict the properties of molecules and materials, thanks to its ability to process large volumes of data. While quantum computers have enormous potential for solving problems that require quantum effects, they still need significant development to become stable and scalable.

In the future, a symbiosis of these technologies is likely: quantum computers and AI will complement each otherallowing you to solve problems that cannot be solved using a single approach. In the long term, quantum computing could fill an important niche in solving problems that currently seem intractable to classical methods.

*This article is a translation of the main ideas. Source