Combining Quantum Computing with Artificial Intelligence

>by Roman Tsyupryk

>July 4, 2026

This text is just my thoughts out loud. I'm only a human being trying to analyze current information and imagine what might happen in the future. My thoughts could be completely wrong or might be just "noise" or they could be food for brainstorming about "what if..." scenarios.

A journey from the smallest layer of reality to the future of machine intelligence.

1. The Foundation: Quantum Physics

To understand quantum computing and its relationship with AI, we must start at the very bottom — the most fundamental layer of reality: quantum physics.

Quantum physics is the science of extremely small things — atoms, electrons, photons. At that scale, reality behaves in ways that completely contradict everyday experience. Five ideas define it:

Energy comes in packets. Classical physics assumed energy flows smoothly, like water. Quantum physics discovered it comes in discrete chunks called quanta — like a staircase instead of a ramp. You can stand on a step, but not between steps.

Wave-particle duality. Light and electrons behave like waves and like particles, depending on how you observe them. In the famous double-slit experiment, electrons fired at two openings create a wave interference pattern — but the moment you watch which slit they pass through, the pattern disappears and they act like particles. The act of observation changes reality.

The uncertainty principle. You cannot simultaneously know both the exact position and exact momentum (mass × velocity) of a particle. The more precisely you measure one, the less precisely you can know the other. This is not a flaw in our instruments — it is a fundamental property of reality itself. (Formally: σₓ · σₚ ≥ ℏ/2)

Superposition. A particle exists in multiple states at once until it is measured. Schrödinger's famous thought experiment captures this: a cat in a sealed box is theoretically both alive and dead until you open the box and look.

Entanglement. Two particles can be "linked" so that measuring one instantly determines the state of the other, regardless of the distance between them. Einstein called this "spooky action at a distance" — he disliked it, but decades of experiments have proven it real.

Quantum physics is mathematically precise and experimentally verified to extraordinary accuracy. Yet as Richard Feynman said: "If you think you understand quantum mechanics, you don't understand quantum mechanics."

2. What is Quantum Computing?

A classical computer stores data as bits — switches that are either 0 or 1. Like a coin lying flat: heads or tails.

A quantum computer stores data as qubits. A qubit can be 0, 1, or both at the same time — until you observe it. Like a spinning coin: both heads and tails until it lands.

This single difference, combined with entanglement and interference, gives quantum computers extraordinary power for certain types of problems:

| Task                          | Classical Computer              | Quantum Computer (future, fault-tolerant)                    |
|---                            |---                               |---                                                            |
| Breaking RSA encryption       | Billions of years                | Under a week, per current best estimates (Shor's algorithm)  |
| Simulating drug molecules     | Intractable for large molecules  | Feasible                                                      |
| Searching unsorted databases  | Linear scan O(N)                 | Quadratic speedup O(√N)                                       |
| Certain optimization problems | Very slow                        | Faster for specific problem types                             |

The RSA estimate is a moving target, not a settled number: Gidney & Ekerå's original 2019 estimate (arXiv:1905.09749) needed ~20 million noisy qubits in 8 hours; Gidney's 2025 refinement (arXiv:2505.15917, Google Quantum AI) cut that to under 1 million qubits, in under a week. No machine anywhere near this scale exists yet, and the qubit count has dropped by more than an order of magnitude in six years — treat any specific qubit count or timeline for "breaking RSA" as provisional.

The key insight is that a quantum computer does not try solutions one by one. It explores all paths simultaneously, then uses interference — amplifying correct answers, canceling wrong ones — to arrive at the solution. Like sending a wave through a maze: the wave naturally amplifies at the exit.

3. How Quantum Computing Works

The computation proceeds in five steps:

Step 1 — Store information in qubits. Physical systems that obey quantum physics serve as qubits: the spin of an electron, the polarization of a photon, the energy level of a superconducting circuit.

Step 2 — Create superposition. A quantum gate (the Hadamard gate) puts qubits into superposition. With n qubits, the computer represents 2ⁿ states simultaneously. With 300 qubits, more states than atoms in the observable universe. (This is a genuine fact about the state space, but a common source of hype: you never read out all 2ⁿ states — measurement collapses the system to one classical answer. The real practical wall is much closer: exact classical simulation of a quantum circuit already becomes intractable around 45–50 qubits, which is why even today's modest quantum processors can do things no supercomputer can fully verify.)

Step 3 — Entangle qubits. Qubits are linked so they influence each other. This is what makes the simultaneous states coordinated rather than independent.

Step 4 — Apply quantum gates. A sequence of gates (the quantum circuit) manipulates the qubits. The art of quantum programming is designing a circuit where wrong answers cancel each other out and the right answer is amplified.

Step 5 — Measure. The superposition collapses and each qubit returns a definite 0 or 1. Because interference has amplified the correct answer, it appears with high probability. The circuit is run many times; the most frequent result is taken as the answer.

4. The Emergence Problem: Why Quantum Does Not Help Everything

A natural question arises: if quantum physics governs everything, shouldn't a quantum computer be better at everything?

The answer is no — and the reason reveals something profound about how reality is organized.

Just because something is built from quantum particles does not mean it behaves quantum-ly at a higher scale. This is called emergence: higher-level systems develop their own rules, independent of the lower-level details.

Water is made of H₂O molecules, but rivers are described by fluid dynamics — not molecular physics. Your laptop's CPU is made of billions of quantum particles, but at the logic gate level those particles behave classically, because thermal noise and scale wash out the quantum weirdness.

Classical logic (AND, OR, NOT) is a perfect abstraction that emerges from quantum physics at the transistor level. It does not need quantum behavior — it deliberately eliminates it. So for sequential logic problems (run this loop, load this file, render this pixel), classical computation is already optimal. A quantum computer offers no advantage because the problem itself has no quantum structure.

Quantum advantage appears only when the problem has quantum structure:

Factoring large numbers has a hidden periodic structure that quantum Fourier transforms find instantly
Molecule simulation is naturally quantum because molecules are quantum systems
Certain optimization problems benefit from quantum tunneling, which can escape local minima that trap classical algorithms — though quantum speedup for general optimization is problem-dependent and not universally guaranteed

5. Simulating Reality: From Molecules to Worlds

This is where quantum computing becomes genuinely transformative — and philosophically fascinating.

In 1981, physicist Richard Feynman gave a conference keynote that launched the entire field (published the following year as a paper):

"Nature isn't classical, dammit, and if you want to make a simulation of nature, you'd better make it quantum mechanical." — Richard Feynman, "Simulating Physics with Computers," 1982

A classical computer cannot efficiently simulate quantum systems because the mathematics grows exponentially. A quantum computer can, because it is a quantum system itself.

In theory, if you can simulate molecules, you can simulate larger and larger systems — proteins, cells, ecosystems, planets. Each step up requires exponentially more qubits. This is the wall:

| What you simulate       | Approximate atoms | Qubits needed (quantum sim) | Context                                                       |
|---                      |---                 |---                          |---                                                             |
| Small molecule (H₂O)    | ~3                 | ~10–50                      | Achievable today                                              |
| Protein                 | ~10,000            | ~millions                   | Far beyond current hardware                                   |
| A grain of sand         | ~10¹⁹              | ~10¹⁹                       | ~100 million × all stars in the Milky Way (~10¹¹)             |
| A planet (e.g. Earth)   | ~10⁵⁰              | ~10⁵⁰                       | Still less than atoms in the universe (~10⁸⁰–10⁸⁵)            |

Scale reference: The largest physical-qubit processors today are in the ~1,000–2,000 range (e.g. IBM's Condor at 1,121, Atom Computing above 1,200), with larger modular chips rolling out through 2026. But physical qubits are noisy; what actually matters for hard problems is logical (error-corrected) qubits, and the frontier there is far smaller — dozens, not thousands (e.g. QuEra reported 96 verified logical qubits in early 2026). Stars in the observable universe number ~10²²–10²⁴. Atoms in the observable universe: ~10⁸⁰–10⁸⁵. The number of qubits needed for a quantum simulation scales roughly with the number of atoms in the system.

To perfectly simulate a world at the quantum level, you would need a quantum computer bigger than the world itself. This is not a technology problem — it is a fundamental physical limit called the Bekenstein bound: you cannot store more information about a region of space than the region itself contains.

However, simulation does not require perfect resolution. Scientists already run ensemble forecasting — thousands of parallel simulations with slightly different starting conditions — to map ranges of possible futures. Quantum computers would make this far more powerful.

Yet a deeper obstacle exists: chaos. Tiny differences in starting conditions grow exponentially over time (the butterfly effect). After a certain horizon, simulations diverge completely regardless of computing power. The universe is sensitive to tiny changes that no measurement can ever capture perfectly.

Stephen Wolfram identified an even deeper limit: computational irreducibility. For many systems, there is no shortcut. The only way to know where a system will be in 10,000 years is to simulate all 10,000 years, step by step, in real time. You cannot outrun the universe's own computation.

Why does this matter for AI? Because simulation is where the two fields most naturally meet. AI is already used to approximate physical systems that are too expensive to simulate directly — protein structure prediction, weather models, materials screening. Quantum computers attack the same problems from the other side, by simulating the physics exactly. The most likely near-term breakthroughs are hybrids: quantum simulation generating precise training data at small scales, and AI learning to generalize from it where exact simulation hits the wall.

6. AI as Algorithm, Quantum as Compute

This brings us to the relationship between quantum computing and artificial intelligence — and a clean conceptual framework:

Algorithm  =  WHAT to do  (the logic, the math, the steps)
Compute    =  HOW to do it  (the machine, the hardware, the speed)

AI is the algorithm. Neural networks, transformers, and machine learning models are fundamentally mathematical procedures — matrix multiplications, gradient descents, probability distributions. They define what the computer should compute.

Quantum computing is a type of compute. It is the hardware layer — the machine that executes instructions. It sits beneath the algorithm.

┌─────────────────────────────────────┐
│            GOAL                     │  ← recognize patterns, predict outcomes
├─────────────────────────────────────┤
│           ALGORITHM                 │  ← AI / Machine Learning
├─────────────────────────────────────┤
│            COMPUTE                  │  ← Classical GPU  OR  Quantum processor
└─────────────────────────────────────┘

Current AI runs on classical compute — and runs well. The algorithm was designed for classical hardware. Quantum computing can accelerate certain AI operations (optimization, linear algebra, sampling), but it is not a direct replacement. Like a diesel engine: you cannot run it on gasoline. The engine must be redesigned for the new fuel.

This redesigned engine is the field of Quantum Machine Learning (QML) — genuinely quantum algorithms that run on quantum hardware and perform AI tasks. These are not classical neural networks moved to a quantum chip; they are new architectures designed from the ground up using quantum principles. The field today rests on a few concrete building blocks:

Variational quantum circuits (VQCs) — the closest quantum analogue of a neural network. A quantum circuit with tunable gate parameters is run on quantum hardware, measured, and then a classical optimizer adjusts the parameters — a hybrid quantum-classical training loop.
Quantum kernel methods — using a quantum circuit to map data into an exponentially large quantum state space, where patterns that are hard to separate classically may become linearly separable. The quantum computer computes the similarity measure; a classical algorithm (like a support vector machine) does the rest.
Quantum sampling — quantum hardware can natively draw samples from probability distributions that are provably hard to sample classically, which is useful for generative models.

The field also has a well-known open problem: barren plateaus. As variational circuits grow, their training gradients tend to vanish exponentially, making large QML models extremely hard to train. Whether QML delivers a practical advantage over classical ML — and for which tasks — is still an open research question, not a settled fact. That honesty matters: QML today is roughly where classical neural networks were in the 1980s — promising architecture, unproven at scale.

7. The Deep Question: Is Intelligence Quantum?

The most profound open question connecting these fields is whether intelligence itself — specifically consciousness — has quantum foundations.

Physicist Roger Penrose and anesthesiologist Stuart Hameroff proposed the Orchestrated Objective Reduction (Orch-OR) theory: human consciousness arises from quantum processes inside neurons, specifically within tiny protein structures called microtubules. If true:

Biological intelligence is a quantum system
Artificial General Intelligence (AGI) that truly replicates human consciousness would require quantum hardware
The statement "consciousness is a quantum system — simulating it with quantum hardware is natural" would be precisely correct

Most neuroscientists and physicists consider this highly controversial. The primary objection — made rigorously by physicist Max Tegmark in a 2000 paper in Physical Review E — is that quantum states in microtubules would decohere in roughly 10⁻¹³ seconds due to the warm, wet, noisy environment of the brain. Neural computation, by contrast, happens on millisecond timescales (~10⁻³ seconds) — ten orders of magnitude slower. Penrose and Hameroff disputed Tegmark's model, and their collaborators (Hagan, Hameroff & Tuszynski, 2002) recalculated decoherence times as 10⁻⁵–10⁻⁴ seconds under corrected assumptions — still short of the ~10⁻³ s neural timescale, though the same paper argues that additional biological effects (ordered water, actin gelation) could push this to 10⁻²–10⁻¹ s, which would match or exceed it. That extended claim rests on assumptions most physicists consider unproven. A 2025 review (Wiest, Neuroscience of Consciousness) argues newer experimental evidence — anesthetic effects on microtubules, room-temperature superradiance, cross-neuron resonance data — is more supportive of Orch-OR than commonly acknowledged, so this is an active, not closed, debate. (Note: quantum coherence itself is not impossible in biology — photosynthesis shows confirmed room-temperature quantum effects — but Orch-OR requires coherence to persist far longer than most physicists think is possible in neurons.) The mainstream scientific position remains skeptical, though less settled than it is often portrayed.

If the answer is yes, quantum computing is not merely an accelerator for AI — it is a prerequisite for true machine intelligence.

See also: The Quantum Mind — a short story exploring what the world looks like if Orch-OR turns out to be correct.

8. What This Means for the Future

Three trajectories are plausible:

Near term (5–15 years): Quantum computers may reach early fault-tolerant operation with thousands of stable logical qubits — this is the optimistic end of expert roadmaps, and timelines have slipped before. If they do, they accelerate specific AI-adjacent tasks — drug discovery, materials science, financial optimization, cryptography. Classical AI remains dominant for general tasks either way.

Medium term (15–30 years): Quantum Machine Learning matures. Quantum algorithms are developed that offer genuine advantages in training large models. AI + quantum simulation enables breakthroughs in biology, climate science, and chemistry that are impossible with classical hardware.

Long term (unknown): If consciousness is quantum, artificial general intelligence may require quantum hardware as a foundation — not just an accelerator. The boundary between simulation and reality becomes philosophically interesting.

Conclusion

The relationship between quantum computing and AI is not simple replacement or direct equivalence. It is a relationship between layers:

Quantum physics is the foundational layer of reality
Quantum computing harnesses that layer for computation
AI is an algorithmic layer that sits above hardware
Quantum ML is what happens when the algorithmic and quantum hardware layers are co-designed

The most important insight from this exploration is subtle: just because quantum physics underlies everything does not mean quantum computing helps with everything. Emergence means that higher layers develop their own rules. Classical AI running on classical compute is already the right tool for most tasks — because those tasks live at a classical layer of abstraction.

But for the problems that live at the quantum layer — simulating molecules, breaking encryption, modeling physical systems, and perhaps one day modeling consciousness itself — quantum computing is not merely faster. It is the natural language of the problem.

The marriage of quantum computing and AI is not yet complete. But the conversation has begun.

Sources

Written as an exploration of quantum physics, quantum computing, and artificial intelligence — their foundations, their relationships, and their shared future.

Part of the Still Becoming series. These two articles are just a small part of a larger Still Becoming series exploring how different technologies could complement each other.

→ Next in the series: The Quantum Mind