Quantum computing, artificial intelligence and blockchain

The next 5 to 10 years will be the key to a new round of global technological revolution and industrial transformation from being ready to explode to a mass explosion. period. With the rapid development of a new round of global technological revolution, disruptive technological innovations are surging, the most eye-catching of which include quantum computing, artificial intelligence and blockchain. Are there any connections between these disruptive technologies and traditional Chinese culture? How does it relate to basic science (e.g. mathematics, physics)? How to objectively understand these cutting-edge technologies? This special issue publishes a report by Zhang Shousheng, an American theoretical physicist and foreign academician of the Chinese Academy of Sciences, who won the International Science and Technology Cooperation Award of the People's Republic of China in January 2018.

Currently, quantum computing, artificial intelligence and blockchain are the three most important basic technologies in the entire information technology industry. In the future, if information technology is to truly develop by leaps and bounds, we must pay attention to basic science, which requires both physics and mathematics, because physics and mathematics are closely related to the information technology revolution.

The discovery of "angel particles" has changed the dilemma of quantum computer research and development

Before talking about quantum computing, let's first talk about the scientific discovery stories related to "angel particles". There are many interesting modern times Scientific discoveries are all related to changes in philosophical concepts, including the deep-rooted ancient philosophical concepts of the Chinese nation. For example, it seems that the world has always been a world of positive and negative opposites. If there are positive numbers, there must be negative numbers, and if there are yin, there must be yang. Where there is good, there must be evil. This opposing view of the world is also present in the physical world of elementary particles.

There was a very great theoretical physicist in history, Dirac, who described Einstein's theory. Special relativity and quantum mechanics were unified. In the process of unification, he performed a very simple mathematical operation and opened a root. When opening the root, there will always be two solutions, positive and negative. Most people may only care about " "Positive solution", not caring about the "negative solution". Dirac explained the "negative solution" as all particles must have antiparticles, and predicted that all particles must have antiparticles.

In 1928, physics Antiparticles were not discovered in the world, and everyone questioned him very much, saying that his equation was definitely wrong. He insisted that his equation was correct. Five years later, he was very lucky, and it turned out that he was in the rays of cosmic radiation. Physicists found the antiparticle of the electron, which is the positive particle, and named it the Dirac sea.

After that, elementary particle physics found protons, antiprotons, neutrons, and antineutrons. And it has been applied. For example, positrons have been widely used in the medical field. There is a medical test called PET, which can image using positrons and negative electrons. The best way to detect Alzheimer's disease is to do PET.

Today, Chinese people are very concerned about scientific development. What is the biggest driving force for scientific development? I think theoretical physicists in history, such as Newton, were under the apple tree. , the falling apple inspired him, and he discovered gravity. When Einstein was riding in the elevator, he felt that the up and down motion of the elevator was very similar to the effect of gravity, thus creating the great theory of general relativity.

In addition, the development of science should not be superstitious about authority. After Dirac became a very famous theoretical physicist, scientists firmly believed that if there are particles in the world, there must be antiparticles. But another great theoretical physicist, Ma. Jorana, out of curiosity, asked whether there are some particles in the world that do not have antiparticles. He invented the Majorana equation, which wonderfully describes that there is a particle that does not have antiparticles, or it itself is My own antiparticle.

Later, the entire physics community was looking for the two coveted particles. One particle was called the "God particle", which was found in an accelerator in Europe in 2012. The physicist who predicted it Higgs won the Nobel Prize, and another one is the "Majorana fermion"

I work in theoretical physics, and the job of a theoretical physicist is generally to make predictions and let experiments take place. Physicists come to test. In 2010, my experimental group predicted that Majorana fermions could be found in a combined device.

But we still need to find a signal that proves the existence of such a particle.

The other day I thought that the Majorana particle has only one side and no opposite side, so in a sense it is half the size of a normal particle. Our theory group made a bold prediction: Since Majorana particles are different from ordinary particles, in a sense, they are only half of ordinary particles. Therefore, its conductivity will be different. The usual particle conductivity is an integer multiple of 0, 1, 2, and 3, which will inevitably lead to a conductivity step of half an integer multiple. We predict it will have a step of 0.5 or 1/2. Later, our theoretical group worked closely with the experimental group and made experimental observations. Indeed, at 0.5, you can see that it is the original pattern of the experiment. A step appeared at 0.5, proving that Majorana The existence of fermions. We named it "Angel Particle" and everyone liked the name very much.

What is the relationship between "angel particles" and the development of information technology?

Currently computers have been divided into two categories, classical computers and quantum computers. Some problems are easy to solve by classical computers, such as multiplying two large numbers. Classic computers can calculate very quickly. But see if a number can be divided into the product of two other numbers. For example, 15 can be written as 3 times 5. If the number is relatively small, you can also calculate it yourself. But if you are given a very large number, it would take a long time for a classical computer to calculate whether the number is the product of two numbers, because the algorithm it uses is an exhaustive method, which divides all numbers that may be divided into one. After dividing the two numbers, you can finally confirm whether this is the product of two numbers. Classical computers are very slow to calculate.

Classical computers can only use exhaustive methods to finally calculate an answer. But the quantum world is a very magical world, a parallel world. For example, in a famous experiment, if I release a particle, such as a photon, it has two holes, either on the left or on the right. But there is an authentic parallelism in the quantum world. A basic particle passes through two holes at the same time at a certain moment. If it is either left or right, the image is not the displayed image.

The quantum world itself is parallel. If you use the quantum world to do calculations, you can calculate in seconds and calculate all possibilities at once, because the quantum world has its original parallelism, which is the most basic concept of quantum computing. But it is very difficult to actually build this quantum computer. For example, the most basic unit of a classical computer is the bit, which means that information is either 0 or 1. All information can be expressed with 0 or 1. This is classical Computer concept. But in the quantum world, a particle passes through the left hole and the right hole at the same time, and is in a certain state of superposition. A qubit cannot tell whether it is 0 or 1. It is in a state of superposition of 0 and 1. Let's listen to an analogy: Schr?dinger's cat is in a superposition of dead and alive. This is a very wonderful phenomenon. But because of this basic phenomenon, it means that a quantum bit itself is unstable. If you look around, you will know that it is either on the left or on the right, either 0 or 1, whatever. A piece of noise will cause great interference to the qubit.

Recently, quantum computers have become a matter of special concern to famous companies around the world and the United States. Google, Microsoft, IBM, and Intel are all making investments, but they cannot fundamentally solve this problem because a qubit is very unstable. Yes, if one day we are told that we have made 50 qubits, the key question is how many useful bits are there. If there is only one useful bit, 10, 20 or even 40 are often needed under the framework of this kind of quantum computing. 50 error correction bits are required to serve it, making quantum computing difficult to truly implement.

But the discovery of angel particles has fundamentally changed the dilemma of quantum computer research and development. This is a process from quantitative change to qualitative change. The qubit itself has the ability to correct errors, which is how I usually split a qubit into two angel particles. Normal particles have two sides, but angel particles have only one side, so angel particles are usually only half of a particle. So usually one qubit can use two angel particles to store it.

Once two particles are used to store it, they are in a distant place and they are entangled with each other. In the classical world, noises are not entangled with each other. In this case, noise cannot be used to destroy the quanta stored by angel particles, so this is a revolutionary change.

So, I gave a speech at the American Physical Society not long ago, saying that angel particles are an exciting discovery. They can be used to make quantum computers with as many bits as they need. There is no need to add error correction bits, and they have built-in correction. Wrong function, this will play a rapid role in the development of quantum computers.

The day robots can make scientific discoveries, that day intelligent machines will surpass humans

As a basic concept, artificial intelligence has been proposed in the 1960s. Today's artificial intelligence can develop by leaps and bounds, mainly due to the summary of many new technologies. According to the iteration of Moore's Law, it can double every 18 months. If you use quantum computing, it will not just double Moore's Law every 18 months, but completely change from quantitative to qualitative change. Human computing power has continued to grow over the years. The birth of the Internet and the Internet of Things has generated a large amount of data. Intelligent algorithms are changing by leaps and bounds. Big data can help machines learn. However, the basis of artificial intelligence is all kinds of data. No matter how good the algorithm is, no matter how powerful the computer is, it cannot become artificial intelligence without data.

Artificial intelligence, although I can see that it is advancing by leaps and bounds, I feel that it is still in a very early stage. Why do you say this? To make a simple analogy, for example, we once saw birds flying, and people also wanted to fly very much. However, in the early days, learning to fly was just simple bionics. Tie wings on human arms. This is simple bionics, but the real state of flying is Since humans have understood the first principle of flight - aerodynamics, and with the physical principles and mathematical equations, they can design the best aircraft. Today's aircraft fly high, fast and well, but they are not like birds. , this is a very core point.

Most of the current artificial intelligence is simply imitating human neurons, but what we should think more about is that there is an opportunity for a major breakthrough in basic science. We must truly understand that wisdom and intelligence. The basic principles of artificial intelligence can truly make fundamental changes in artificial intelligence.

What kind of basis can we use to truly measure whether artificial intelligence reaches human standards? Some people may have heard of the Turing test. The Turing test means that a person talks to a machine, but it is not known whether the other person is a person or a machine. During the entire conversation, if you spend a day and can't feel it at all, it means that the robot seems to have reached the level of a human. Although Turing was a great computer scientist, I do not agree with this method of judgment. Many human emotions are not rational emotions. It may not be easy for a rational machine to learn from the brain of an irrational human being.

So I would like to propose a new judgment method. When will intelligent robots truly have intelligence that surpasses humans? I think the greatest thing about human beings is that we can make scientific discoveries. One day robots can really make scientific discoveries, that day machines will surpass humans.

Recently I wrote an article on artificial intelligence, which will be published in the Journal of the American Academy of Sciences. It will be mentioned in the article that mankind’s greatest scientific discoveries include relativity, quantum mechanics, etc., in chemistry. The greatest discovery there is the discovery of the periodic table of elements. Can an intelligent machine automatically discover the periodic table of elements without any guidance? Can it help humans discover new drugs? Can machine learning be used to discover new materials? These are the standards by which the level of artificial intelligence is judged.

To realize the mutual coexistence and development of blockchain and artificial intelligence, they will be the most valuable

In today’s world, individuals will generate a lot of data, and their genes Data, medical data, educational data, behavioral data, etc. are particularly needed for the development of artificial intelligence. A lot of data is held in central institutions and has not achieved true decentralization. The emergence of blockchain can create a decentralized data market.

I describe the entire concept of blockchain in one sentence, called "In Math we trust". This concept is based on mathematics.

The most basic thing in the entire blockchain and the entire information technology field is basic mathematics, which can protect personal privacy in the data market and make reasonable statistical calculations. For example, there is a very magical calculation method called zero-knowledge proof, which can prove to you that my data is very valuable, but it does not tell you where the truly private data is.

With blockchain, the data market can make society more equitable. The biggest injustice in modern society is that people tend to discriminate against some minorities. But what is most needed in the machine learning process is the data owned by the minority. If the accuracy of machine learning reaches 90% today and increases from 90% to 99%, what it needs is not the data that has been learned, but different data from before. It is often a small amount of data that is most valuable for machine learning. Once our data is built on the basis of blockchain, coupled with these wonderful mathematical algorithms, we will be able to have a healthy data market. In this world, blockchain and artificial intelligence will be the most valuable if they achieve the concept of mutual coexistence.

As for the entire blockchain, the public’s understanding of it is not the most fundamental first-principles understanding. In terms of the most basic physics principles, reaching mutual consciousness is like everyone agreeing on the same "ledger". It is equivalent to that in physics, magnets are originally messy, but in the ferromagnetic state they point in the same direction. It's the same.

Achieving ultimate consciousness also exists in the natural world. This phenomenon is called entropy reduction. To achieve ultimate consciousness, if everyone is moving in the same direction, the entropy of this state is far smaller than the entropy of chaos. It is very difficult to achieve this ultimate consciousness because entropy is always increasing.

To achieve a consensus system on the blockchain, an algorithm is used, which requires energy consumption. This thing sounds unreasonable, why does the account consume energy, but from the second theorem of physics, this is a very reasonable thing, because reaching consciousness itself is a decrease in entropy, but the entropy of the entire world is certain To increase, so while achieving ultimate consciousness, other entropy must be eliminated. This non-centralized mechanism is very similar to the natural world in which magnets change from a chaotic state to an ordered ferromagnetic state. It is also an inevitable trend to consume energy and pay a price.

So in the ideal information world, in the future, everyone will have all their own data and it will be stored in a completely decentralized manner, so that hackers will not be able to hack everyone's data. Then using some encrypted algorithms on the blockchain can truly protect personal privacy and make good calculations, and things like the theft of many personal data in Facebook will not happen.

The problems we have to solve today in quantum computing, artificial intelligence, and blockchain technology are all problems for mankind. Chinese scientists will face great opportunities. In addition to doing a good job in applied science and technology, they must also There should be truly original basic scientific breakthroughs, such as the physical and mathematical principles introduced above, although these things sound relatively abstract, such as the principle of entropy increase and positrons and negative electrons. The wonder of the world lies in the fact that basic science can provide broad new development prospects for the entire information technology industry.