Will AI Make Quantum Computing Less Important?
Ir. J. Philippe Blankert PhD
Researcher and author at Blankertbooks.com
J. Philippe Blankert 24 April 2025
Introduction
Quantum computing and artificial intelligence (AI) are often hailed as twin frontiers poised to redefine computing. Quantum computers leverage qubits and quantum mechanics to tackle problems believed intractable for classical machines, while modern AI (especially deep learning and large language models) harnesses vast data and computing power to achieve human-like problem-solving. Each field has seen rapid advances – AI transitioning from theoretical speculation to practical applications in just the last few years gemslabs.io, and quantum computing moving from lab prototypes toward early commercial devices. This raises a pivotal question: as AI’s capabilities explode, will it eclipse the anticipated role of quantum computing? Recent breakthroughs in AI – from solving complex scientific problems to writing code – hint that tasks once hoped for quantum machines might be achievable with classical AI. In this essay, we examine key domains (cryptography, physical simulation, and optimization) to assess whether AI’s rise diminishes the importance of quantum computing, or if the two technologies will diverge into distinct niches while also converging in synergistic ways. The analysis considers current capabilities and future projections, drawing on academic and industry sources to ground the discussion.
Quantum Computing vs. AI – Different Strengths
At their core, quantum computing and AI approach computation very differently. Quantum computers use qubits that exploit superposition and entanglement, enabling certain computations (like factoring large numbers or simulating quantum physics) to be done exponentially faster than classical bit-based machines cloudsecurityalliance.org. In theory, a fault-tolerant quantum computer can directly solve mathematical problems that underpin encryption or precisely model quantum systems, tasks that classical computers struggle with due to combinatorial explosion. AI, on the other hand, uses classical processors (CPUs, GPUs, etc.) running algorithms – neural networks, search methods, and reinforcement learning – to approximate solutions to complex problems by learning patterns from data. Rather than brute-forcing every possibility, AI excels at generalization and heuristic problem-solving, often achieving “good enough” solutions where exact methods are infeasible.
These distinctions mean that quantum computing and AI shine in different arenas. Quantum computing’s native advantage is solving problems with quantum structure or mathematical structure (e.g. factoring) exactly, albeit requiring exotic hardware. AI’s advantage is flexibility – it can tackle diverse tasks (vision, language, design, etc.) on today’s computers, bypassing the need for new physics. Yet there is overlap: both aim to solve certain optimization or simulation problems beyond reach of straightforward classical brute force. As we explore specific application areas, we will see AI making surprising inroads into domains once envisioned for quantum computing, while quantum technologies still promise unique capabilities beyond AI’s current scope.
Cryptography: Quantum Threats and AI Efforts
Quantum’s role: Modern cryptography relies on problems like factoring large integers and discrete logarithms, which are infeasible for classical algorithms. Quantum computing poses an existential threat here: Shor’s algorithm can factor large numbers and break RSA/ECC encryption given a powerful enough quantum computer cloudsecurityalliance.org. In theory, a quantum computer running Shor’s algorithm could crack 2048-bit RSA in hours, defeating the public-key encryption securing ~90% of today’s secure communications cloudsecurityalliance.org. This looming threat has catalyzed the field of post-quantum cryptography (PQC) to develop quantum-resistant encryption. In short, the importance of quantum computing in cryptography is very clear – it’s the one known technology that can undermine widely used cryptosystems at a fundamental level.
AI’s role: Does AI change this picture? So far, not fundamentally. Despite sensational headlines, AI has not discovered any algorithm that breaks strong encryption in polynomial time. For example, reports in 2023 claimed an AI “cracked” a new PQC algorithm, but in reality the AI was just assisting a side-channel attack, and the underlying cryptography remained sound blog.cloudflare.com. In fact, AI techniques have been used for years in cryptanalysis supportively: machine learning can help find patterns in encryption side-channel emissions or automate parts of cryptanalytic attacks blog.cloudflare.com. However, these AI-driven methods exploit implementation weaknesses (timing, power usage, etc.), not the mathematical structure of RSA or AES themselves. There is no evidence that today’s AI can, say, factor a 2048-bit number or invert a one-way function purely through “learning”; these tasks still appear computationally intractable without a fundamental algorithmic breakthrough.
That said, AI does contribute to cryptography in other ways. It can aid human cryptanalysts by sifting through data or suggesting novel approaches, and it’s being used to strengthen security (e.g. AI-based anomaly detection for cyber threats). But when it comes to replacing the anticipated role of quantum computing in cryptography – i.e. breaking codes – AI is not a substitute. Only a quantum computer running algorithms like Shor’s is known to threaten current public-key cryptography at a fundamental level cloudsecurityalliance.org. Consequently, the importance of quantum computing in this domain remains undiminished by AI’s rise. If anything, AI’s contributions (like automating side-channel attacks) are complementary, highlighting that classical and quantum tools can both be used – in different ways – to attack or improve cryptographic systems blog.cloudflare.com. In summary, for cryptography the divergence is clear: quantum computing represents a future qualitative game-changer for security, whereas AI provides incremental improvements within the classical paradigm. Organizations still must prepare for quantum threats through PQC, as no AI solution is swooping in to render quantum attacks moot.
Simulating Physics and Chemistry: AI Invades Quantum Territory
One of the most touted applications of quantum computing is quantum simulation – using quantum computers to model quantum-mechanical systems like molecules, materials, and chemical reactions. Feynman’s original vision for quantum computers was exactly this: since nature is quantum, quantum hardware should simulate it efficiently. Indeed, problems such as calculating complex molecular ground states or reaction dynamics, which explode exponentially on classical computers, could be handled by a quantum computer “speaking the language of quantum mechanics” natively medium.com. This capability is expected to revolutionize fields like drug discovery, materials science, and superconductivity by allowing exact or highly accurate simulations of systems that are otherwise impossible to model exactly.
AI’s advances: Surprisingly, recent advancements in AI suggest that many of these simulation tasks might be achievable classically – at least approximately – with clever machine learning and data-driven models. Researchers have found that AI can often capture quantum system behaviors without needing quantum hardware. For example, neural networks can learn the wavefunction of certain many-particle systems, reducing the exponential complexity to manageable levels arxiv.org. In a landmark 2017 study, Carleo and Troyer showed that a neural network representation of a quantum state could achieve extremely high accuracy for prototypical interacting spin models in 1D and 2D. In other words, AI can variationally solve the quantum many-body problem for certain cases, a task originally deemed a prime candidate for quantum computers.
More broadly, machine learning models are making inroads into chemistry and materials science problems that were “supposed to be quantum computing’s bread and butter”. Deep learning is now used to predict molecular properties and approximate solutions of quantum chemistry far faster than traditional methods. In drug discovery and materials design, AI-trained models can screen and optimize compounds using knowledge distilled from huge datasets of chemical information, achieving results comparable to expensive quantum chemistry calculations medium.com. In fact, for many weakly correlated systems (where electrons or particles don’t exhibit extreme quantum entanglement), AI-driven methods are not just competing with, but often outperforming, standard computational chemistry approaches like density functional theory (DFT) medium.com. This has real implications: it accelerates innovation in pharmaceuticals and materials at a pace that “makes Moore’s Law look like a leisurely stroll”medium.com.
Crucially, AI is no longer limited to easy cases; it’s starting to tackle strongly correlated quantum systems as well. These are the really hard problems – systems with entangled particles (like high-temperature superconductors or complex chemical reactions) that are thought to require quantum computers to simulate exactly. Yet researchers have begun using advanced neural networks (sometimes with reinforcement learning) to approximate behaviors even in these challenging regimes medium.com. One dramatic example is in quantum chemistry: a leading researcher, Prof. Alexandre Tkatchenko, recently noted that with modern AI techniques “you can really do most of chemistry” on classical computers gemslabs.io. That is a bold statement underscoring how far AI-based simulation has come – tasks once imagined to demand a quantum computer are being handled by clever classical algorithms and learned models gemslabs.io.
Quantum’s remaining edge: Does this mean quantum computing’s importance for simulation is fading? Not entirely. While AI can approximate many quantum systems, there remain scenarios where a quantum computer’s exact and fully quantum approach should win. In particular, for strongly correlated systems at large scale and high precision, or for dynamics with genuinely quantum effects (like entanglement over long times), AI may struggle to capture every nuance. AI models excel at prediction when trained on data, but if a system behaves in novel ways outside the training distribution or requires exponential precision, a quantum computer could still be indispensable medium.com. As one analysis put it, AI can often get “close enough” with approximations, but sometimes in science “close enough isn’t good enough”medium.com – we might need the exactness of quantum simulation. For example, AI might predict properties of a new material with 90% accuracy, whereas a quantum computer could in principle calculate them to 99.999% by brute-force simulation of the quantum mechanics. Certain cryptographic or high-precision physics applications might demand that exactness.
Research experts expect that strongly correlated quantum matter – e.g. complex electron interactions in advanced materials – is where quantum computers will retain an advantage. AI’s impressive performance so far has been in regimes where either data is abundant or correlations are moderate. Ongoing development of quantum hardware and algorithms could allow precise simulations of strongly correlated systems that AI can only approximate community.materialssquare.com. In essence, AI has raised the bar, covering much of the easy-to-medium difficulty terrain, which “recalibrates” our view of what we truly need quantum computers forgemslabs.io. As Dr. Frank Noé observed, the convergence of AI with classical computing might make the number of problems that strictly require a quantum computer “much smaller” than previously thought gemslabs.io. Quantum computing’s role in simulation may thus narrow to a subset of the most challenging, precision-critical problems.
Summary: In the arena of simulating physical and chemical systems, AI is indeed replacing some anticipated roles of quantum computing – at least delaying the need for quantum machines by solving many problems with classical means. AI offers accessible, cost-effective simulation on GPU clusters today medium.com, whereas quantum hardware is expensive, delicate, and still limited in scale medium.com. For many industrial needs (drug discovery, materials optimization), an AI solution that is “good enough” and available now will beat waiting years for a perfect quantum solution. However, quantum computing remains crucial for the endgame of simulation: as a “native speaker” of quantum mechanics, a sufficiently advanced quantum computer can provide exactitude and handle intrinsically quantum phenomena beyond AI’s grasp medium.com. The likely future is one of convergence: hybrid approaches where classical AI handles large parts of a problem and quantum computing tackles the irreducible quantum core. Rather than rendering quantum computing obsolete, AI is helping clarify where quantum advantages truly lie – focusing quantum efforts on the hardest of the hard problems, while AI handles the rest.
Optimization and Problem-Solving: Different Paths to Solutions
Another key area to examine is optimization – the task of finding best solutions among astronomically many possibilities (for example, the shortest route connecting many cities, optimal allocation of resources, or minimizing a complex cost function). These problems appear in logistics, finance, engineering design, and more. Many such tasks are NP-hard or combinatorially complex, meaning they resist brute-force solution as the problem size grows. Quantum computing has been pursued as a means to accelerate optimization via specialized algorithms and even dedicated hardware, while AI and classical algorithms have developed along a different track of heuristics and learning-based methods.
Quantum approach to optimization: Quantum algorithms like Grover’s search promise quadratic speedups for unstructured search problems, and the Quantum Approximate Optimization Algorithm (QAOA) aims to leverage quantum superposition and interference to guide combinatorial searches toward optimal solutions. Companies like D-Wave have built quantum annealing machines specifically to solve optimization problems by physically mimicking the process of annealing to find low-energy states (which correspond to optimal or near-optimal solutions). The hope has been that such quantum solvers could outpace classical algorithms on certain large optimization tasks like scheduling, portfolio optimization, or vehicle routing. Indeed, there have been tantalizing milestones: in 2023, D-Wave reported that its 5000+ qubit quantum annealer achieved a quantum advantage on a practical problem, simulating a complex spin glass (magnet) system much faster than a classical supercomputer could postquantum.com. Specifically, D-Wave’s prototype solved a materials science optimization/simulation in minutes that was estimated to take a classical supercomputer millions of years postquantum.com. This is touted as the first real-world problem where a quantum optimizer dramatically beat all known classical methods – a significant proof of quantum potential in optimization and simulation of certain physics models.
However, it’s important to note that such successes have been in relatively specialized domains (e.g. finding the lowest-energy configuration of a particular spin glass model) postquantum.com. For more generic optimization problems – say, traveling salesman tours, generic scheduling, or Boolean satisfiability – quantum algorithms are still in their infancy and have not yet demonstrated clear supremacy over classical methods. QAOA, for instance, has been tested at shallow depths on small problems, and while it works, its performance so far has often been matched or exceeded by the best classical heuristics when run on comparably small instances. As of now, no broad NP-hard problem has an empirically demonstrated quantum speedup on practical instance sizes; quantum optimization remains a promising research area but without general-purpose victories yet.
Recommended by LinkedIn
AI and classical progress: On the classical side, AI and advanced algorithms have dramatically improved optimization by introducing smarter heuristics and learning from data. A vivid example is Google’s deep reinforcement learning approach to chip design floorplanning – traditionally a months-long human-engineer task to optimally place components on a silicon die. In 2021, a deep RL model was trained to perform chip floorplanning, and it was able to produce chip layouts in under six hours that were as good as or better than human-designed ones on key metrics like power, performance, and area nature.com. This was a landmark because it solved a decades-old optimization challenge (which had defied automation) using AI, not brute force or quantum tricksnature.com. Similarly, DeepMind’s AlphaZero algorithm famously mastered the combinatorial game of Go (and chess, shogi) via self-play reinforcement learning, outperforming all prior programs – an illustration of AI efficiently searching enormous possibility spaces with learned intuition. And in another striking feat, DeepMind’s AlphaTensor system treated the search for faster matrix multiplication algorithms as a game, and managed to discover new algorithms more efficient than any human-found method in 50 years spectrum.ieee.orgquantamagazine.org. AlphaTensor found dozens of matrix multiplication shortcuts (e.g. a way to multiply two 4×4 matrices in 47 multiplications instead of the human standard 49) that expand the known space of optimal algorithms quantamagazine.org. This is essentially AI performing algorithmic optimization at a meta-level – designing better methods for a computational task spectrum.ieee.org.
These examples show AI’s strength in optimization: it can learn from experience or data to guide the search process in ways classical brute-force never could. For many practical optimization tasks (routing, scheduling, resource allocation), we now have machine learning models and heuristics (like genetic algorithms, neural networks, or swarm intelligence methods) that yield excellent solutions within reasonable time, even if they can’t prove optimality. AI can also generalize – once an AI learns how to optimize chip layouts, it can apply that skill to new chips faster and better, as the Google study showed (the system “utilizes past experience to become better and faster at solving new instances” of floorplanning) nature.com.
Given these advances, one might ask: if AI can solve or at least effectively handle so many optimization problems right now on classical hardware, is there less need for quantum optimizers? In the near term, possibly yes – the bar for quantum to show advantage is higher now. For a quantum computer to be truly valuable in optimization, it must outperform not a naive brute-force, but AI-augmented classical methods that are remarkably good. This is a moving target, because classical algorithms continue to improve. A case in point: some early quantum annealing tests found no clear benefit over classical simulated annealing or other solvers, which themselves keep getting faster on better hardware. The contest is ongoing, but quantum devices will need to solve really hard instances where classical+AI heuristics still struggle, to prove their worth in this field.
Outlook: In optimization, we see both divergence and potential convergence. On one hand, AI and classical computing are covering a vast swath of practical optimization needs already – from industrial design to logistics – reducing the urgency for quantum solutions. On the other hand, quantum computing may eventually offer guaranteed or significant speedups for specific hard cases or larger scales that classical methods can only approximate. It might be that AI will handle 95% of optimization problems efficiently, but the remaining 5% (the hardest, largest ones, or those requiring provable guarantees) could be where quantum shines. There is also a collaborative angle: future quantum-classical hybrid algorithms might use AI to set up or guide a quantum optimizer. For example, an AI could preprocess or divide a problem in a way that a quantum algorithm can solve a crucial piece faster than any classical method, combining strengths. In summary, AI has for now taken the lead in practical optimization, delivering solutions without waiting for quantum hardware. This likely diminishes the near-term importance of quantum computing in most optimization applications. Yet, quantum computing research in optimization is far from redundant – it is pushing toward breakthroughs on problems that remain extremely challenging, keeping alive the possibility that tomorrow’s quantum machine may solve an optimization puzzle in seconds that today’s best AI would still take eons to crack.
AI Designing Algorithms and Systems – A New Paradigm
Beyond specific domains like chemistry or routing, AI’s rapid progress in meta-problem-solving is worth noting. Modern AI can write code, prove simple theorems, and even help design new AI systems. This is significant because it shows AI encroaching on tasks that require creativity and logical planning – areas one might not have expected raw computational brute force (quantum or otherwise) to directly address. For instance, DeepMind’s AlphaCode model achieved roughly average human performance in competitive programming contests, ranking in the top 54% of participants in code challenges by generating and testing massive numbers of candidate programsdeepmind.google. Likewise, generative language models (like OpenAI’s Codex or GPT-4) can produce working code for a given specification, effectively automating portions of software development that were thought to require human-like reasoning. AI tools are now routinely used by developers to generate and debug code, compressing development cycles.
AI is also optimising its own field: neural architecture search (NAS) uses algorithms (like reinforcement learning or evolutionary strategies) to design neural network architectures better than those crafted by human experts nature.com. Google’s AutoML initiative showed that an RL agent could invent a CNN architecture that surpassed hand-designed models on image classification, demonstrating that AI can innovate in AI design. This self-referential improvement – AI designing AI – indicates a virtuous cycle where AI rapidly improves itself, limited primarily by available compute.
Quantum computing, by contrast, was never aimed at such cognitive or creative tasks. A quantum computer doesn’t “think” up a new algorithm or write code; it executes operations faster. Thus, in areas like automated coding, algorithm discovery, or system design, AI is not so much replacing quantum computing as rendering the distinction moot – we wouldn’t have expected quantum computing to contribute here anyway. What is interesting, however, is that AI’s success at these higher-level tasks might reduce the need for certain brute-force computations. If an AI can intelligently search a solution space (whether it’s the space of programs, neural networks, or mathematical proofs), it might find shortcuts that obviate the need for an exhaustive search that even a quantum computer would have struggled with. AlphaTensor’s discovery of faster matrix multiplication algorithms quantamagazine.org is a prime example: a naive approach to improving matrix multiplication could be to try all possibilities (an astronomically large search space), something even a quantum computer couldn’t brute force. But AI guided the search efficiently, revealing new solutions with much less effort quantamagazine.org. Such algorithmic improvements benefit all computing, classical and quantum alike.
In summary, AI’s ability to solve “meta” problems – to generate code, optimize algorithms, and design complex systems – is expanding what we can do with classical computing in ways that quantum computing wasn’t targeting. This doesn’t directly diminish quantum computing’s importance in its core domains, but it does highlight how far classical AI can go. It underscores a theme: many intellectual tasks once assumed to require fundamentally more computation (or a new computing paradigm) have turned out to be solvable with smarter classical techniques. This raises the bar for where quantum advantage is needed. Quantum computing will likely focus on the truly intractable tasks that still remain after we throw advanced AI at the problem.
Convergence and Synergy of AI and Quantum
While we have largely discussed AI and quantum computing in competitive terms, it’s crucial to note that they are also complementary. In fact, a convergence of the two may define the future of high-performance computing. Quantum computers, as they mature, will not operate in isolation – they will be part of hybrid systems alongside classical processors and AI algorithms. We already see early signs of this synergy:
In the long run, rather than AI making quantum computing irrelevant, we are likely to see co-evolution. AI will continue to handle a growing range of tasks on classical platforms, pushing quantum computing to specialize in areas where it has clear, irreplaceable superiority. Those areas (as identified earlier) might include breaking certain cryptographic schemes, simulating truly quantum phenomena with exact fidelity, or solving specific hard optimization problems and mathematical structures. When quantum computers do come online for those tasks, AI will be there to integrate them into workflows – managing hybrid algorithms and compensating for quantum hardware’s weaknesses (like noise and limited qubit counts).
Conclusion
Recent advances in AI have indeed encroached on some problem domains that were once shining targets for quantum computing. AI’s prowess in modeling complex physical systems, optimizing challenging problems, writing code, and even designing algorithms is reducing the immediate need for quantum computing in various applications. In fields like chemistry and materials science, AI-driven simulations are challenging the anticipated dominance of quantum computerscommunity.materialssquare.com, handling “weakly correlated” cases and even many strongly correlated ones with surprising effectiveness gemslabs.iogemslabs.io. In optimization and logical problem-solving, AI algorithms often find excellent solutions without quantum aid, and they improve year by year.
However, concluding that AI will make quantum computing unimportant would be an overstatement. The two technologies operate on different principles and offer different kinds of advantages. Quantum computing still holds the trump card for certain problems of mathematics and physics – most prominently, the ability to factor large numbers and break certain cryptography, which no AI or classical method has achieved cloudsecurityalliance.org. It promises exact solutions where AI must approximate, which will matter for the most demanding applications in science and security. Moreover, as AI takes over more routine and even advanced classical tasks, it helps clarify the niche where quantum computing will truly shine. Far from being made irrelevant, quantum computing may become more focused – zeroing in on a smaller set of high-value problems that remain beyond the reach of even the smartest classical AI gemslabs.io.
Importantly, AI and quantum computing are not mutually exclusive or simply substitutable; they are increasingly seen as complementary components of future computing systems. Each can amplify the other: AI can automate and optimize quantum engineering, while quantum computers can provide boosts to AI once they reach sufficient scale gemslabs.iodeveloper.nvidia.com. The likely scenario is one of convergence: hybrid quantum-classical-AI algorithms tackling problems together, each doing what it does best.
In conclusion, AI’s dramatic progress has certainly shifted perspectives on what can be achieved without quantum computers, perhaps delaying some of quantum computing’s expected impact by solving portions of those problems classically. The “importance” of quantum computing may thus be felt later and in more specialized ways than initially thought. But it is far from diminished in the long term. Just as classical computing did not render mathematics unimportant (it merely changed how we approach it), classical AI’s success does not nullify the profound potential of quantum computation. Instead, it challenges the quantum field to demonstrate its unique value on genuinely unsolvable classical problems. When that happens – be it unbreakable encryption via quantum physics, or exact simulation of a novel quantum material yielding a revolutionary technology – the world will feel the impact of quantum computing. Until then, AI will continue to solve what it can, and for the rest, the quantum quest continues, now with sharper focus. The future will be shaped by both technologies working in tandem, each critical in its own domain. The smartest strategy is not to bet on one to eliminate the other, but to harness the interplay of AI and quantum computing, ensuring that we leverage their convergence while respecting their divergence in capability.
Sources: