RT: The Future Prospects of Quantum Computing and Artificial Intelligence

Welcome to 2018’s first edition of Research Translated! For any new Quantum Authority readers, Research Translated is a new, experimental section that translates the latest in quantum computing research into terms you can understand. Love it? Hate it? Let us know.

— -

Introduction

We had a reader who, after perusing our first article on the relationship between quantum computing and AI, asked us for more information about how quantum computing relates to artificial intelligence. So we decided to take a deeper look this week at what’s going on in the world of quantum computing and AI.

The article that we are analyzing this week is titled “Quantum computation, quantum theory and AI”, by Professor Mingsheng Ying. Professor Ying is the Research Director of the Center for Quantum Computation and Intelligent Systems at the University of Technology in Sydney, Australia, and has been active in the quantum computing research community since the 90s.

Although not strictly research, this essay provides a deep technical dive into where Professor Ying sees quantum computing playing into the development of modern artificial intelligence techniques. The technicality of the paper does require some translation for those who are not familiar with quantum physics and differential equations, which is why we are including it as this weeks edition of Research Translated.

So without further ado, let’s start!

Article Summary

Professor Ying divides his essay into 4 parts, each of which builds into the next:

1. An introduction to quantum computing
2. An exploratory passage into current research on quantum computing, with an emphasis on topics that Professor Ying has worked on personally
3. An exploration of what Professor Ying considers to be the strongest candidates for areas where quantum computing could have applications within artificial intelligence
4. A passage on the intersection of current AI and quantum research (i.e. not where one could be developed and help the other, but instead where both could be developed simultaneously)

Here’s our summary for each section, in bullet point form:

Section 1: Introduction to Quantum Computing

This section is a crash course in quantum computing. It is a high-level overview, and it assumes that you have a college math background through matrix and vector algebra.

If you do not have a background in this kind of math, then you’re in luck! The Quantum Authority has already written several articles going over the high-level ideas that Ying introduces without all of the complex math.

We will not re-hash out the introductory material that Ying talks about. If you want a background in the introductory material, then be sure to check out the following quick-reads:

• Technical Overview of Quantum, Round 1
• Technical Overview of Quantum, Round 2

The only information that Ying offers that TQA has not covered already is his emphasis on quantum registers as the only means of measuring the outcome of a quantum computation.

• A register is a basically a place in a computer’s architecture that can hold data. The data can be accessed very quickly from there.
• Every program and piece of architecture in a traditional computer can be traced down to its bits (i.e. its 0s and its 1s). Similarly, everything in a quantum computer can be traced down to its qubits. A quantum register is the same thing as a normal register in concept, but its built with qubits
• Ok, so we now know that a quantum register is simply a piece of the architecture that allows the quantum computer to store data. So all that Professor Ying is saying is that we need a quantum register in order to store the output of the data

And that’s the crash course in quantum computing!

Section 2: Current Applications of Quantum Computing:

Professor Ying makes sure to emphasize that his overview of research in quantum computing is not inclusive of all fields and has an emphasis on topics that are of interest to him or that he has worked on personally.

Models

Ying spends a lot of time talking about the different quantum computing models out there. It is a topic that the writers at the Quantum Authority plan to write more about in the near future. In the meantime, here’s an overview of what he talks about (Note: some of these topics sound scary! Don’t worry though, bear with us and we’ll get you through):

Quantum Turing Machine

• Alan Turing is a British mathematician who is by and large considered to be the father of computer science. If you want to know more about him, check out the “The Imitation Game” movie. It’s good stuff
• Turing devised the model that would eventually lead to a computer. It became known as a “Turing machine”.
• It is important to note that a Turing machine is a logical model, not a physical machine
• At a high level, a Turing machine can be used to solve mathematical functions by switching between different states based on different criteria, which is the basis for what all software programs do.
• For example, you could have two states: A and B. You start off in State A, and only move to State B when the user inputs a number higher than 5. Otherwise, you stay in State A. If the user inputs a 1, you stay in State A. If they input a 7, then you move to State B.
• This framework can be seen in modern software. For example, if you want to transfer money from one bank account to another through your bank’s online web portal and you enter a negative number into the “amount to transfer” box, then the program will not allow you to click send. It is in the “NoTransfer” state. If you enter a positive number though, the program then shifts to the “TransferOK” state.
• I made those state names up, but you get the idea right?
• Remember how the difference between a register and a quantum register was that one based on bits and the other on qubits? It’s the same idea here. A Quantum Turing machine uses qubits instead of bits to determine how the machine shifts in between states
• As the Turing machine is the basis for traditional computing, it makes sense to start with it as a model for quantum computing too

Quantum Circuits

• This is the most popular quantum computing model.
• Before we can describe a quantum circuit, we first need to define what quantum gate is
• Traditional computers use gates in their architecture.
• We know that a binary 0 represents a false, and a binary 1 represents a true, right? Well, what happens if we have a situation where I want to execute a program if two conditions are true?
• For example, say I want to my online banking portal to email me if I have not paid my credit card bill yet AND I have insufficient funds to pay off the bill (thankfully, not my actual financial situation haha). How do I do that? Well, you would write a program that only sends the email if BillDue is true AND if InsufficientFunds is true.
• In this situation, I am effectively using an AND gate here.
• There are lots of different types of gates. AND, OR, NOT, and XOR (exclusive or) are the most popular. We can go into the nitty-gritty of those at a different time
• Quantum gates are the same thing as normal gates but based on qubits instead of bits.
• A quantum circuit is basically a series of quantum gates in a row

Adiabatic Quantum computing

• This model is pretty different because it has no basis in traditional computing models
• Every other model moves in discrete time. All that means is that the model moves at specific times at the same time. For example, a quantum Turing machine would change states only at specific time intervals. This model is a continuous time model, meaning it moves whenever it likes.
• This model basically works by modeling the system in this mathematical function called a Hamiltonian. We don’t need to get into what exactly a Hamiltonian is, but its used in a lot of control theory problems.
• The Hamiltonian then varies over time, which is how the model computes

Measurement-based quantum computing

• This model also has no basis in traditional computation
• As the name implies, the model basically uses measuring qubits to solve problems

Topological quantum computing

• This model uses two-dimensional particles as a way to construct quantum gates
• It was designed as a way to minimize error in quantum computations due to a phenomenon called “decoherence”.
• Still theoretical and not really used

Math and Logic Tools

Categorical quantum logic

• It sounds scary, but it is basically a mathematical notation that lets one easily describe quantum computing at a high level

Lambda calculus

• Also sounds scary. Lambda calculus is the basis for functional programming, an alternative to object-oriented programming. Basically its a way to code functions that take in other functions as input
• Lambda calculus is seen as a logical way to create quantum programming language

Quantum computational logic

• Basically, a way to write out the state of quantum registers
• Theory of computation based on quantum logic
• Qubit logic instead of bit logic

Algorithms

• This section was kind of disappointing tbh. Ying pretty much said that there are no real quantum algorithms out there and that all the work there was theoretical.

Architecture

• Same deal as algorithms. Not much work, all theoretical.

Programming languages

• Ying went over some languages proposed by researchers. They are older and pretty complicated, so we’ll go over those in a different post. Google “QFC programming language if you’re curious”
• Ying’s article was written before Microsoft announced their new quantum programming language, Q#, which is the best example of a quantum language yet.

Section 3: Quantum assisting AI development:

Ying basically posited that quantum could be used to advance AI and that AI could be used to advance quantum. So it’s a two-way street. He came up with a few current places where quantum could really make a difference in the development of AI

Learning

• The only place that quantum has currently affected AI
• AI programs learn by training on data right? There are a couple of ways to do that, but most of them involve showing the computer a piece of data and labeling it. Eventually, the program will recognize different items.

Decisions Problems

• Ying actually argued that any quantum improvements to existing algorithms for AI decision making problems would be marginal. We already have fast algorithms, there’s no reason to make them faster.

Search

• It is widely believed that search will be the first major application of quantum in AI. There was a guy a while back who described some algorithms that could be used, but to date, no one has devised a quantum search algorithm

Game Theory

• Game theory is a form of economic analysis. It is basically a way to quantify how rational people make their decisions based on information presented to them

Section 4: Co-development of Quantum Computing and AI:

Ok, so this section had suggestions on how AI could advance quantum computing, and how quantum computing could help advance AI. We divided those ideas into two sections.

Quantum ideas that could be used to advance AI:

Semantic analysis

• Semantic Analysis is a field within AI to allow a computer to tell the tone of a text. So a computer could tell if a tweet, or a Facebook post, or a speech was generally very positive or very negative.
  Ying notes that some researchers have found similarities between proposed quantum algorithms and current semantic analysis techniques, but he doesn’t believe that these similarities are strong enough to be meaningful

Entanglement of words in natural languages

• Some researchers noted that NLP algorithms display similar nature to entanglement, called “spooky action at a distance”
• Basically it’s a phenomenon in quantum physics where quantum particles change the closer you get to them
• The similarities suggest that quantum computing could help advance these types of algorithms

AI ideas that could be used to advance quantum computing:

Quantum Bayesian Networks

• Ying argues that AI and quantum both use stats, so that’s an avenue that they can both get into. A Bayes model is basically a graph that uses pre-set probabilities to determine outcomes. There have recently been quantum-based Bayesian networks proposed in physics journals.

Recognition and discrimination of quantum states and quantum operations

• Ying argues that since pattern recognition is a big deal in AI (i.e. computer vision techniques) that those same techniques can be used to help a quantum computer determine different quantum states. This has been getting a lot of attention lately

Our take on the article

Honestly, I thought this essay was super interesting. It’s not often that you get a leading expert in the field of quantum computing give both broad and deep summary of where we are in quantum computing and what directions it could go.

As the owner of The Quantum Authority, I am certainly interested in quantum computing. I also have a strong interest in AI and the development of applications that use AI as its background. I agree with Ying’s assessment that AI and quantum can and should be developed in tandem because that will allow both fields to grow at an exponential rate

Final words

I think Ying gives the reader a lot of insight into possible directions that quantum computing as a field can go and also gives a lot of ideas as to how quantum relates to AI.

I, however, was a bit disappointed as to the lack of specifics in the field. However, one should note that one of the reasons Ying had so few specifics is because there are not many specifics out there since this is such a new and emerging field.

Keep an eye out on quantum computing and AI! We certainly will and we’ll be sure to keep searching for new topics in that area to write about for you guys