Quantum Information and AI
Exploring the term and history in the context of new developments in artificial intelligence
Since there is so much talk of quantum computing I thought it would be appropriate to explore a term I came across, namely quantum neural networks. However to do this I must contend with the term ‘quantum information’. As I am proficient in neither you must see this article as a struggle to understand rather than any final answer. You can see the first segment as a summary of articles on the history of quantum computing from NIST.
“In physics and computer science, quantum information is the information of the state of a quantum system. It is the basic entity of study in quantum information theory, and can be manipulated using quantum information processing techniques.”
“Quantum neural networks (QNNs) are neural network models which are based on the principles of quantum mechanics. There are two different approaches to QNN research, one exploiting quantum information processing to improve existing neural network models (sometimes also vice versa), and the other one searching for potential quantum effects in the brain.”
History of Quantum Information
I was reading a post from the National Institute of Standards and Technology in the United States about the History and Future of Quantum Information by Ben P. Stein. He describes quantum information as the marriage of quantum physics and computing theory.
“Google, Intel, Microsoft, IBM, and so forth have quantum teams, and they’re actually trying to build hardware,”
says Chris Monroe, a JQI fellow, professor of physics at the University of Maryland, and co-founder of the quantum information company IONQ.
The field seems to be transitioning from academic labs to industry.
“Quantum mechanics revolutionized the 20th century. It explains the workings of very small objects such as atoms, and things that have very small amounts of energy such as photons, or individual packets of light. It brought about the laser, tiny computer chips and energy-efficient LEDs. It spawned new sectors of the economy such as the semiconductor industry, in which global worldwide sales reached $339 billion in 2016 and grew by 21 percent in the second quarter of 2017 compared to the same period in 2016, according to the Semiconductor Industry Association.”
“Computing is a physical process, and if you think about it in those terms, it opens up all sorts of new insights in both computing and the physical sciences,” says Ron Boisvert, a mathematician at NIST
In 1994 Peter Shor then a mathematician at AT&T Bell Laboratories created an algorithm, designed to run on a future quantum computer, that could quickly crack the encryption schemes used for banking and other sensitive transactions. According to NIST this discovery launched a race that bloomed into a new field.
Quantum Supremacy
Researchers are no longer focused solely on building a quantum computer that could carry out Shor’s algorithm and break encryption codes. For many, an intermediate goal is to achieve “quantum supremacy,” a term coined by Caltech’s John Preskill. Thus when quantum computers can carry out tasks that are not possible or practical for a normal computer this would be reached, often deemed interesting in optimisation problems.
“There’s ample reason to believe that quantum computers would be good at solving optimization problems. For one, nature seems to have no trouble with them. Plants solve them all the time when they turn sunlight into energy using photosynthesis.”
Quantization, Superposition, Decoherrence and Entanglement
“Quantum computers would get their special powers by exploiting the strange rules of the very small…”
One of these rules are quantization. A firefighter experiences a smooth continuous range of gravitational potential energy while sliding down a pole. A tiny particle such as an electron can make abrupt “quantum” jumps between energy levels in an atom.
Superposition. In the ordinary classical world, a skateboarder could be in only one location or position at a time, such as the left side of the ramp (which could represent a data value of 0) or the right side (representing a 1). But if a skateboarder could behave like a quantum object (such as an atom), he or she could be in a “superposition” of 0 and 1, effectively existing in both places at the same time.
It is argued this could be very useful for information technology.
Ordinary bits of data can only be a 0 or 1.
If the ion is in a superposition on the left and right side of the bowl, it can act as a 0 and 1 at the same time.
According to NIST quantum bits, or qubits, can therefore, be used to process larger amounts of information much more quickly than traditional ones.
Destroying superpositions is called decoherence, and it’s a major problem. So, researchers must protect qubits, and isolate them.
Decoherence occurs when a quantum superposition (such as an atom in a 0 and 1 state at the same time) is destroyed by some outside disturbance, such as an experimenter measuring the atom.
Another ingredient for quantum computers is entanglement, and it’s perhaps the strangest quantum property of them all. It’s a sort of quantum marriage between qubits.
In the quantum phenomenon known as entanglement, the properties of two particles are intertwined even if they are separated by great distances from each other.
“Albert Einstein pointed out this “spooky action at a distance,” and it disturbed him greatly, but this behavior has been confirmed time and again, including in a recent landmark NIST experiment […] Entanglement is an extremely useful commodity. Measure the value of one qubit and it determines the value of a second entangled qubit.”
However the more qubits there are, the faster they tend to decohere. NIST claims this is why there are still no quantum computers that can perform truly useful tasks. The system could crash before doing anything worthwhile.
Atomic Clocks and Theory of Entanglement
In the 1990s NIST researchers already had qubits , but called them atomic clocks.
“Atomic clocks tell time with amazing precision. The NIST-F2, the most accurate U.S. atomic clock used for timekeeping, keeps time to an accuracy of less than a millionth of a billionth of a second. That may seem like overkill, but GPS satellites contain atomic clocks and depend on such precision to send time-stamped signals that help us pinpoint our locations anywhere on Earth to within about a meter.”
In the late spring of 1994, discussions of Shor’s result were circulating via e-mail, NIST physicist Charles Clark organized the NIST Workshop on Quantum Computing and Communication.
Artur Ekert, then a physicist at the University of Oxford. Ekert was making the rounds in the community to spread the word about these exciting new ideas. In 1991, Ekert had introduced the theory of entanglement-based quantum key distribution, an essential component of quantum cryptography, which became realized in actual devices later in the decade.
They envisioned using trapped ions in an experimental system.
“In the proposed scheme, researchers would trap a group of individual ions in a line, like birds on a wire. Lasers would manipulate the ions’ energy states so that they each represented 0s and 1s or superpositions of the two. Each ion would communicate with another by rocking back and forth. The rocking would enable each ion to exchange information with its neighbors. In this way, researchers would be able to carry out computations with the qubits.”
So, he and group member Chris Monroe used lasers to control the ions’ energy states in a way that affected their motion. And in December 1995, they announced the first component of a quantum computer.
Quantum Logic Gates
“Logic gates are building blocks for processing information.”
By arranging gates in a circuit, engineers can create something akin to a flowchart that enables computers to carry out many kinds of logical operations, such as mathematical calculations — and perform the kinds of tasks that computers can do.
Using lasers, the researchers could also cause the ion’s internal energy state to go into a superposition of 0 and 1, putting the ion’s motion into a superposition of 0 and 1. This allows the gate to process multiple possibilities simultaneously, unlike ordinary logic gates, so the gate can consider both the “yes” and “no” possibilities of the business deal at the same time.
Multiple qubits entangled with one another could be used to carry out massive numbers of calculations at the same time.
Two-qubit gates can process four possible combinations of 0s and 1s simultaneously.
Three-qubit gates can process eight possible combinations.
“Each additional qubit doubles the number of combinations the gate can process at the same time, so there is an exponential increase with each new qubit.”
Measurement can only however extract a small amount of information about the results of the computations. To partly understand why NIST mentions another concept.
Wave-particle duality: all objects can either act as particles or waves. As a wave, an ion has peaks and valleys in space; the peaks represent where the ion is more likely to be found when measured. Peaks and valleys add together to create a new wave pattern.The waves will combine or “interfere” to form a new wave pattern. By interfering waves scientists create patterns that can be converted back into qubit values and yield useful information.
According to Ronald F. Boisvert quantum computation entails creating an even superposition of all possible answers, and then manipulating that superposition then. As such when one makes a measurement, one gets the right answer with high probability.
Race to Build a Quantum Computer
As mentioned the tiniest disturbances can destroy superpositions in qubits, and entangled quantum systems get more and more fragile as you add more and more qubits to it.
The NIST-Boulder group created a “racetrack” trap for capturing and storing multiple ions and is used for processing quantum information. They built a quantum processor that can run any computer program that requires only two qubits.
“A practical quantum computer would require many more qubits, so one idea is to connect many of these two-qubit processors and get them to work together as a full-fledged computer.”
There is also an attempt to build tiny microscopic “quantum” drums that vibrate by releasing tiny packets of energy known as phonons.
“Companies — from large ones such as Google, IBM, Intel and Microsoft to smaller ones such as Quantum Circuits and IonQ, a startup set up by JQI’s Monroe — also are working to build quantum computers.”
According to NIST the race to achieve quantum supremacy is intense.
- Google’s John Martinis (a former NIST researcher) has announced that the company aims to soon demonstrate a 49-qubit quantum computer, using superconducting components.
- IBM has reported building and testing a 50-qubit system, based on superconducting components.
- At the 2018 CES show in Las Vegas, Intel announced a 49-qubit quantum test chip.
- In Japan, researchers have built a system using photons in optical fiber to create what they call a Quantum Neural Network, intended as a specialized quantum computer good at tackling tasks such as the traveling salesman problem.
It is thought that the first quantum computers will be bulky like the first traditional computers in the 20th century.
In terms of usage there seems to be potential for communication, transmitting information over great distances.
“Researchers seem to agree on one thing: photons, or particles of light, will be used to transmit information from one quantum computer to another. In a recent dramatic demonstration, China sent entangled photons from a satellite to the ground 1,200 kilometers (approximately 750 miles) away.”
Is Encryption Dead with Quantum Computing?
Most often when I hear talk of quantum computing it is in connection to security. However according to NIST all hope is not lost thanks to two developments: quantum cryptography and post-quantum encryption.
Quantum cryptography: the uncertainty principle indicates that measuring any object will disturb it. Quantum encryption relies on the fact that eavesdropping — trying to gain information on a message — is a form of measurement and will disturb a system in a way that can be detected. Quantum cryptography also has applications in secure communication. Researchers in China recently used the technique to make a secure video call to colleagues in Austria who were 2,500 kilometers (approximately 1,600 miles) away.
On the other hand it may not be viable for three reasons:
- It is very expensive
- It is designed to protect from eavesdroppers not attackers
However quantum cryptography systems have been expensive and they can be hacked:
“…it has been shown that even cryptography systems based on quantum properties can be hacked under real-world conditions; for example, when electrical static or noise — governed by the rules of classical physics — inevitably gets introduced in the system.”
Quantum cryptography is designed to protect communications between two trustworthy parties from being intercepted by an eavesdropper.
But what if the person on the other side is a cyberattacker? Eavesdroppers are currently not the biggest concern in cybersecurity.
Post-quantum encryption: ways of encrypting data, using mathematical functions that do not involve factoring. Post-quantum algorithms are based on classical mathematics techniques that predate quantum information. Researchers are already developing post-quantum algorithms, in case quantum computers become a reality.
Applications of Quantum Information Without Computers
Quantum sensors: atomic clocks potentially could be used as quantum sensors. These quantum logic clocks are part of a new generation of ultraprecise timekeeping devices. The new devices are so precise that they can do more than tell time; they can act as sensors of gravity and other properties.
Another is atom gyroscope,” a rotation sensor, by harnessing wave-particle duality.
“We usually think of atoms as tiny balls — in other words, particles. But under the right circumstances, atoms can act as waves. An atom’s wavelength — the distance between successive peaks and valleys — is too small to be noticeable, but if you make the atoms colder, their wavelength is big enough to be detectable in present-day experiments.”
Quantum teleportation: Researchers at the National Institute of Standards and Technology (NIST) have “teleported” or transferred quantum information carried in light particles over 100 kilometers (km) of optical fiber, four times farther than the previous record.
The Second Quantum Revolution
According to NIST it seems we are making a lot of progress, to the point of being in ‘the second quantum revolution’.
The first revolution: enabled inventions such as the laser and transistor, the basic building block of computers, when scientists knew the rules of quantum mechanics and built devices that followed those rules.
The second quantum revolution: is all about controlling individual quantum systems, such as charged molecules, to a greater extent than before, enabling even more powerful applications of quantum information. The method borrows a quantum logic approach from an experimental NIST atomic clock.
Joint Institute for Laboratory Astrophysics’ (JILA) three-dimensional (3-D) quantum gas atomic clock, an early example of the second quantum resolution, controls the interactions between strontium atoms to an exquisite degree to enable atomic clocks that are more accurate than ever before.
“…the stakes are high — and from nations to corporations, everyone is getting into the game.”
New developments in Artificial Intelligence relating to Quantum
As such we could perhaps see some connections to the field of artificial intelligence. Processing information is important for a lot of companies and quantum mechanics which led to the semi-conductor was the advent for computing as we know it today. Large companies that are interested in developments in Quantum was mentioned, however I will mention one: Google.
Google’s corporate mission is “to organize the world’s information and make it universally accessible and useful.”
In physics, a quantum (plural: quanta) is the minimum amount of any physical entity (physical property) involved in an interaction.
In physics and computer science, quantum information is the information of the state of a quantum system. It is the basic entity of study in quantum information theory, and can be manipulated using quantum information processing techniques.
As such the smallest parts of information is of interest to one of the largest technology companies in the world, and Google.com is the most visited website in the world.
There is a team at Google called Quantum: A research effort from Google AI that: “…aims to build quantum processors and develop novel quantum algorithms to dramatically accelerate computational tasks for machine learning.”
Quantum computing as described by Google AI: “is a new paradigm that will play a big role in accelerating tasks for AI. We want to offer researchers and developers access to open source frameworks and computing power that can operate beyond classical capabilities.”
They have open-source frameworks are specifically designed for developing novel quantum algorithms to help solve near-term applications for practical problems. One tool is Cirq and another is OpenFermion.
They describe near-term applications as:
- Quantum Simulation: the design of new materials and elucidation of complex physics through accurate simulations of chemistry and condensed matter models are among the most promising applications of quantum computing.
- Error mitigation techniques: they work to develop methods on the road to full quantum error correction that have the capability of dramatically reducing noise in current devices.
- Quantum Machine Learning: they are developing hybrid quantum-classical machine learning techniques on near-term quantum devices.
- Quantum Optimization: discrete optimizations in aerospace, automotive, and other industries may benefit from hybrid quantum-classical optimization, for example simulated annealing, quantum assisted optimization algorithm (QAOA) and quantum enhanced population transfer may have utility with today’s processors.
Quantum Machine Learning at Google
They are studying universal quantum circuit learning for classification and clustering of quantum and classical data. They are also interested in generative and discriminative quantum neural networks, that could be used as quantum repeaters and state purification units within quantum communication networks, or for verification of other quantum circuits.
The team at Google AI is exploring Quantum Neural Networks:
“Since its inception, the Google AI Quantum team has pushed to understand the role of quantum computing in machine learning.”
They are often still making progress and the research is in an experimental stage with the beginning of applications.
On September the 20th 2019 according to the Financial Times Google claimed to have built the first quantum computer that can carry out calculations beyond the ability of today’s most powerful supercomputers, a landmark moment that has been hotly anticipated by researchers. 53-qubit design it codenamed Sycamore seems to have done this, but Google did not comment in the article.
Later NewScientist laid the claim on the 26th of September that Google had reached ‘Quantum Supremacy’.
“Google’s quantum computer tackled a task called a random circuit sampling problem. In such a problem, after a series of calculations each qubit outputs a 1 or 0. The aim is to calculate the probability of each possible outcome occurring. Google says Sycamore was able to find the answer in just a few minutes — a task it estimates would take 10,000 years on the most powerful supercomputer. Although that is impressive, there is no practical use for it.”
The full paper of their recent breakthrough called Quantum Supremacy Using a Programmable Superconducting Processor was reportedly pulled and can be read online. However it may be wise to wait to hear from Google to confirm any information in this document. Underneath is the summary:
“The tantalizing promise of quantum computers is that certain computational tasks might be executed exponentially faster on a quantum processor than on a classical processor. A fundamental challenge is to build a high-fidelity processor capable of running quantum algorithms in an exponentially large computational space. Here, we report using a processor with programmable superconducting qubits to create quantum states on 53 qubits, occupying a state space 253∼1016. Measurements from repeated experiments sample the corresponding probability distribution, which we verify using classical simulations. While our processor takes about 200 seconds to sample one instance of the quantum circuit 1 million times, a state-of-the-art supercomputer would require approximately 10,000 years to perform the equivalent task. This dramatic speedup relative to all known classical algorithms provides an experimental realization of quantum supremacy on a computational task and heralds the advent of a much-anticipated computing paradigm.”
It is fascinating to consider what will happen next in the intersection of quantum information and artificial intelligence. It is also hard to tell where it will lead, perhaps a new computing paradigm?
This is day 132 of #500daysofAI. I write one new article about or related to artificial intelligence every day for 500 days.
