News

Advances in Quantum Computing Portend a Fantastic AI Future

Breakthroughs hint we live in a multiverse.

• By Pure AI Editors
• 01/08/2025

The field of quantum computing continues to make steady progress towards the goal of commercially viable systems that will revolutionize computer science and change the world in fantastic, unpredictable ways. In the last few months of 2024, researchers unveiled a significant advance in quantum hardware (a new chip from Google), and an advance in quantum software (a new algorithm from the University of Hamburg).

The capabilities of current AI systems, such as the OpenAI ChatGPT natural language system and the Sora text-to-video system, are limited by the amount of computer memory needed for the billions of parameters they use, and the GPU processing power needed to train the systems and to compute a response during the inference phase.

When quantum computers become practical, both the memory and speed limitations of AI systems will be obliterated, enabling unimaginably powerful systems with superhuman abilities. Some speculate this will signal the Singularity -- the point in time when technological growth becomes uncontrollable and irreversible, resulting in unpredictable consequences.

For good or bad, quantum computing offers the potential to vastly improve the performance of AI algorithms, particularly in areas like:

• Optimization Problems: Many AI algorithms, such as those used in machine learning, rely on optimization (e.g., minimizing error in training models). Quantum algorithms like the Quantum Approximate Optimization Algorithm (QAOA) could significantly speed up these processes.
• Speeding Up Computations: Tasks such as matrix operations and solving systems of linear equations, which are foundational in AI, could benefit from quantum speedups through algorithms like the Harrow-Hassidim-Lloyd (HHL) algorithm.
• Sampling Efficiency: Quantum computers can sample from complex probability distributions much faster than classical computers, which is valuable for probabilistic models like Bayesian networks.

The Willow Quantum Computing Chip
In December 2024, Google Research unveiled a new quantum computer chip called Willow. The Willow chip demonstrates real-time error correction and performance that points the way to the creation of a practical quantum computer. Willow was fabricated at a facility in Santa Barbara, California -- one of only a few facilities in the world built from the start for manufacturing quantum computing chips.

One of the biggest challenges in quantum computing is dealing with physical errors. The fundamental components of quantum computing, called qubits (quantum bits), rapidly exchange information with their environment, making it difficult to protect the encoded information needed to complete a computation. When more qubits are used, more errors occur, and a quantum system becomes useless. The Willow chip has 150 qubits.

The Willow chip uses a new design that, somewhat remarkably, reduces errors as the number of qubits increases. Google tested different sized arrays of qubits, starting from a grid of 3x3 encoded qubits, then to a grid of 5x5 qubits, and then to a grid of 7x7 qubits. For each succeeding grid, the error rate was cut in half.

To test the Willow chip, it was applied to a quantum computing benchmark problem called the random circuit sampling (RCS) problem. Willow completed the computation in under 300 seconds. Google estimated that the same calculation, if performed on one of the world's fastest non-quantum supercomputers, would require approximately 10,000,000,000,000,000,000,000,000 years -- vastly longer than the estimated age of the universe, which is about 13,700,000,000 years.

The RCS benchmark is not a practical problem. It's just a combinatorial problem that's difficult to solve. To get a feel for the problem, consider this quantum computing program, written using the Google "cirq" (circuits-quantum) library and language.

These four cirq statements create two simulated qubits:

import cirq
print("Begin create simple quantum circuit ")
a = cirq.NamedQubit("0")
b = cirq.NamedQubit("1")

These six statements create a simulated quantum circuit:

mom0 = cirq.Moment([cirq.H(a), cirq.H(b)])
mom1 = cirq.CNOT(a,b)
mom2 = cirq.measure(b, key="m")
circuit = cirq.Circuit(mom0, mom1, mom2)
print("Circuit: ")
print(circuit)

And these seven statements execute the simulated circuit:

sim = cirq.Simulator()
result = sim.run(circuit, repetitions=10)
print("run 1: ")
print(result)
result = sim.run(circuit, repetitions=10)
print("run 2: ")
print(result)

The output of the simulation is:

Circuit:
0: ───H───@────────────
          │
1: ───H───X───M('m')───

run 1:
m=1100001101

run 2:
m=0110001010

Loosely speaking, the point is that, because of quantum superpostion, each qubit is simultaneously both 0 and 1, and the quantum circuit gives a different result on each run. The RCS problem is related to this idea, and samples from all possible results of a random circuit.

A New Quantum Algorithm from the University of Hamburg
The Willow chip was tested on the somewhat obscure random circuit sampling problem primarily because common combinatorial optimization problems require more qubits than the Willow chip has. For example, the goal of the Traveling Salesman Problem (TSP) is to find the shortest path that visits all cities in the problem.

Suppose you have n = 5 cities: Anaheim, Boston, Chicago, Detroit, Eugene. If you start at any city, there are factorial(5) = 5 * 4 * 3 * 2 * 1 = 120 different paths, each of which looks like (Anaheim, Eugene, Chicago, Boston, Detroit). But if there are even just 20 cities, there are factorial(20) = 2,432,902,008,176,640,000 different paths.

This would be a good test of a quantum chip, but encoding the Traveling Salesman Problem for n = 10 cities on a quantum computer would require over 5,000 physical qubits -- far more than the largest usable quantum chip currently has.

Researchers from the University of Hamburg demonstrated a way to solve the Traveling Salesman Problem for up to n = 9 cities with an algorithm that uses just one qubit. The technique is detailed in the 2024 paper "Solving The Travelling Salesman Problem Using A Single Qubit" by K. Goswami, G. Veereshi, P. Schmelcher, and R. Mukherjee. The technique is very tricky but points the way for using quantum computers with a limited number of qubits to solve previously intractable problems.

What Does It All Mean?
The Pure AI editors asked Dr. James McCaffrey, from Microsoft Research, for comments. McCaffrey replied, "It's clear that the self-correcting ability of the Willow quantum computing chip represents an important step forward. My colleagues and I often discuss the progress of quantum computing towards a practical, commercial system. None of us knows if the progress will consist of many small steps, such as the Willow chip and the single-bit TSP algorithm, or if progress will be accelerated by unexpected large leaps."

McCaffrey added, "Because nobody knows what the evolution path of quantum computing will look like, nobody can predict when practical quantum computing will become a reality. It could be in a matter of months, or it might take years."

Furthermore, he said, "My colleagues and I were interested that the researchers who developed the Willow quantum chip speculated that the speed of the Willow chip suggests that quantum computation occurs in many parallel universes, and therefore that we live in a multiverse. This is a rather mind-boggling idea that has profound philosophical implications."