The MIT Technology Review unveiled 20 Breakthrough Technologies in 2018. The technologies selected this year include artificial intelligence technology to generate a confrontation network (GAN), artificial embryos, and clean energy based on natural gas. Technology, etc., the authoritative list of emerging global technology fields has been 17 years old.

Practical 3D metal printer

Reason for selection: New equipment makes 3D printed metal parts a practical technology for the first time

Technological breakthrough: 3D metal printers enable low-cost, fast metal object printing

Significant: The ability to print large, complex metal objects on demand will revolutionize manufacturing

Leading researchers: Markforged, Desktop Metal, GE, etc.

Maturity: now

 

As costs become lower and easier to use, this technology is expected to become a practical technology for component production. If it is widely used, it will likely change the way we mass produce products.

In the short term, with this technology, manufacturers will no longer need to maintain a large inventory, they can print a part on demand. For example, when a customer needs to replace a part with an old car, he can immediately provide it to him.

In the long run, large factories that produce a certain number of parts on a large scale will be replaced by small workshops with rich product lines. These small workshops will be able to print a wide range of parts at any time according to the needs of customers.

The advantage of this technology is that it can produce lighter, stronger metal parts and complex-shaped parts that cannot be manufactured using traditional metalworking methods. It can even precisely control the microstructure of the metal during the manufacturing process.

Professional interpretation

In 2012 and 2013, 3D printing received enthusiastic attention from the media. From the outside world, although the follow-up development has returned to the normal level, the evolution of technology has not stopped, especially the opportunity to change the traditional manufacturing production mode of 3D metal printing. Bigger.

According to the relevant standards set by the F42 Technical Committee of the American Society for Testing and Materials, the additive manufacturing, also known as 3D printing, is divided into seven major technical methods. Currently, there are four technologies for printing in "metal". For metal powder bed melting (PBF, Powder BedFusion), laser metal deposition (LMD, Laser Metal Deposition), adhesive spray molding (Binder Jetting), and layered solid manufacturing (LOM, Laminated Object Manufacturing).

Mainly actively researching the application industry of 3D metal printing, including aerospace, medical materials and top sports cars, mainly high demand and high demand for customization. And what about the future development? Many of the equipment exhibited by Formnext 2017 have been mass-produced, or in the direction of mass production, showing that commercial operation of metal 3D printing is feasible, but compared to traditional casting or forging methods, 3D metal printing There are still several obstacles. First, the cost of machinery and metal powder is still high. Second, although 3D metal printing has come to work with four laser nozzles, the speed is still slow from the user's point of view.

Perfect online privacy

Reason for selection: A tool originally developed for the cryptocurrency transaction process that now allows you to avoid revealing any non-essential information while surfing the Internet.

Technological breakthroughs: Computer scientists are perfecting an encryption tool that can be verified without revealing non-essential information.

Significant: If you need to disclose personal information to do something online, this approach allows you to easily avoid the risk of privacy breaches or identity theft.

Main investigators: Zcash, JPMorgan Chase, Holland International Group, etc.

Maturity: now

Thanks to the emergence of a new tool, true Internet privacy is finally achievable. For example, the tool allows you to prove that you are 18 years old without revealing your date of birth, or you can prove that you have enough money in your bank to complete a financial transaction without revealing your bank balance or other details. This greatly reduces the risk of privacy breaches or identity theft. This tool is a new cryptographic protocol called "zero-knowledge proof".

Researchers have been researching for decades, but until last year people’s interest in zero-knowledge verification began to soar, in part, thanks to the growing enthusiasm for cryptocurrencies, and most cryptocurrencies The reality that the organization has. At the same time, to a large extent, it also benefited from the electronic currency established in late 2016 – Zcash applied zero-knowledge verification to practice. Zcash's developers use a method called zk-SNARK (Concise Non-Interactive Zero Knowledge Verification) to allow users to conduct anonymous transactions. Often, this is not possible in Bitcoin and other public blockchain systems, and transactions in Bitcoin and other public blockchain systems are transparent to everyone.

Although in theory, these transactions are anonymous, they can be tracked or even identified by combining with other data. Vitalik Buterin, founder of Ethereum, the world's second-largest blockchain network, called zk-SNARK a "technology that revolutionized the rules of the game." For banks, this allows blockchains to be used in payment systems while protecting customer privacy.

Last year, JPMorgan Chase added zk-SNARK to its own blockchain-based payment system. However, although zk-SNARK promises all kinds of benefits, it is computationally intensive and slow. At the same time, zk-SNARK needs a "trust installation", and the generated key can damage the entire system if it falls into the wrong hands. However, researchers are working hard on alternatives, hoping to deploy zero-knowledge verification more efficiently, without the need for the above keys.

Professional interpretation

Zhu Mingjie, founder and CEO of Yuxin: If there is a system that can openly and transparently guarantee the privacy of users, then the system will be attractive enough, especially in the blockchain. The transaction is open to the whole network, and the blockchain system based on zero-knowledge proof can realize the privacy protection of information in a completely open and transparent form, which is undoubtedly of great practical significance. Both Zcash and JP Morgan's blockchain systems are based on this implementation.

Now, more blockchain systems will or are integrating zero-knowledge proof technology. There are some mature applications in the blockchain, and the transaction-related private data needs to be kept secret to any third party. If exposed in the supply chain system, it will have huge consequences. For systems that do not expose information to potential third parties, zero-knowledge proof is undoubtedly a “just-needed”. With the rapid development of the blockchain, the zero-knowledge proof technology is expected to be widely used in the near future and become the cornerstone of the next-generation value Internet.

Zero carbon natural gas power generation

Reason for selection: a new engineering approach to natural gas power plants that recycles carbon dioxide

Technological breakthrough: A power plant can capture carbon emissions from natural gas combustion in a cheap and efficient way, avoiding greenhouse gas emissions

Significant: Natural gas power generation supplies nearly 32% of the electricity to the United States, and its carbon emissions also reach 30% of the total carbon emissions of the power sector.

Principal Investigator: 8 RiversCapital, Exelon Power Company, CB&I, etc.

Maturity: 3-5 years

 

For the foreseeable future, we may have to use natural gas as one of the main sources of power generation. The electricity generated by off-the-shelf and cheap natural gas accounts for 30% of the total US electricity generation and 22% of the world's electricity generation. Although natural gas is much cleaner than coal, it still produces a lot of carbon emissions.

A leading-edge power plant has emerged outside of Houston, the center of the US refinery, and they are testing a technology that can achieve clean natural gas energy. The company has a 50 megawatt project, they are Net Power. The company believes they can capture all the carbon dioxide released during natural gas power generation while generating electricity at a low cost, at least as much as the cost of a standard natural gas power plant.

If this is really achievable, it means that zero carbon energy can be obtained from fossil fuels at a reasonable price. Such natural gas power generation will certainly improve the energy supply situation because it is neither as expensive as nuclear energy nor as unstable as renewable energy.

Net Power is a product of 8 Rivers Capital, Exelon Power and CB&I Energy. The company's power plant is already in trial operation and has begun initial testing, and they intend to announce the results of the initial assessment in the coming months.

The power plant places the carbon dioxide produced by burning natural gas into a high-pressure, high-temperature environment and uses synthetic supercritical carbon dioxide as a "working fluid" to drive a special turbine. Among them, most of the carbon dioxide can be continuously reused, and the rest can not be used to capture in a low-cost way.

The key to reducing costs is to sell some of the carbon dioxide. Currently, carbon dioxide is mainly used to assist in the extraction of crude oil. This market has limited capacity and is not environmentally friendly. However, in the end, Net Power hopes that the demand for carbon dioxide in other industries will rise, such as cement manufacturing, plastics manufacturing and other carbon-based materials industries.

Professional interpretation

Chen Chengmeng, associate researcher at the Shanxi Institute of Coal Chemistry, Chinese Academy of Sciences: China's current natural gas accounts for about 3% of total electricity supply, and is expected to account for 6.7% by 2020. In order to make the thermal power plant fueled by natural gas and coal more clean and environmentally friendly, the existing technology system is usually implemented by further adding environmental protection devices such as CO2 adsorption, desulfurization and denitration, and ash reduction.

 

Artificial embryo

Reason for selection: Scientists have begun to make embryos from stem cells

Technological breakthroughs: Without the use of egg cells or sperm cells, researchers can develop embryo-like structures from stem cells alone, providing a new way to create artificial life.

Important: Artificial embryos will provide researchers with a more convenient tool for studying the mysterious origins of human life, but the technology is triggering new bioethical controversies

Principal Investigators: Cambridge University, University of Michigan, Rockefeller University, Chinese Academy of Sciences, etc.

Maturity: now

 

Embryologists at the University of Cambridge in the UK have used stem cells to create a realistic mouse embryo in a breakthrough study that redefines how to create artificial life. The embryo is not derived from the combination of egg cells and sperm, but only cells obtained from another embryo.

The researchers carefully placed the cells on a three-dimensional scaffold, and the cells then began to associate with each other and were arranged in the shape of bullets unique to the mouse embryos of several days, and the researchers were attracted by the scene. "We know that stem cells have extremely powerful potentials that can show near-magic abilities. However, we don't realize that they can achieve self-organization so perfectly," said team lead Magdelena Zernicka-Goetz.

Zernicka-Goetz claims that her "synthetic" embryos may not develop into mice. Nonetheless, they also mean that we can soon bred mammals without eggs.

But this is not the ultimate goal of Zernicka-Goetz. She wants to study how the cells of early embryos begin to differentiate into their special effects. The next step in the research, she said, is the use of human embryonic stem cells to generate artificial embryos, an ongoing study at the University of Michigan and Rockefeller University.

Synthetic human embryos will be the gospel of scientists, allowing them to sort out the processes that embryos have experienced in their early development. Moreover, since these embryos are developed from easy-to-operate stem cells, laboratories will be able to use various tools, such as gene editing techniques, to study them as they grow.

Professional interpretation:

Li Linxian, assistant professor of the Karolinska Institute of Sweden: The first three-dimensional co-culture of two stem cells in an in vitro culture dish simulates the early process of embryonic development, providing a possible alternative for the early development of embryos. The value of research on artificial embryos is still very obvious, for example, in basic research on early embryo development. In the basic research of early embryo development, it is often necessary to use a genome editing tool such as CRISPR to compare the difficulty of operating stem cells and embryos in a culture dish, and the operation of stem cells is easier. Using genome editing techniques such as CRISPR for stem cells is more convenient than embryos.

 

Antagonistic neural network

Reason for selection: Two AI systems gain imagination by playing the "cat and mouse" game

Technological breakthroughs: Two AI systems can create super-real original images or sounds by confronting each other. Until then, machines have never had this capability.

Significant: This gives the machine a imaginative ability, so it may make them less dependent on humans, but they also turn them into an amazing digital counterfeit tool.

Main investigators: Nvidia, Institute of Automation, Chinese Academy of Sciences, Google Brain, DeepMind, Baidu, Alibaba, Tencent, Shangtang Technology, Yitu Technology, Yun Cong Technology, Defiance Technology, etc.

Maturity: now

 

The ability of artificial intelligence to recognize objects has grown stronger: by showing it a million pictures, it can tell you with a surprising accuracy about which pedestrian is crossing the road. But it is almost impossible for AI to generate a picture of a pedestrian alone. If it can achieve this, it will be able to create a large number of seemingly realistic composite images, putting pedestrians in various environments. Automated driving systems may use these pictures for training without leaving home.

But the problem is that creating something from scratch requires imagination, and this is the ability that artificial intelligence technology has been difficult to achieve.

It wasn't until 2014, when Ian Goodfellow, a PhD student at the University of Montreal, had an academic debate with a friend at the bar, he suddenly thought of the answer to this question. This means called "confrontational generation network" (GAN) uses two neural networks (a simplified human brain mathematical model that is the cornerstone of modern machine learning), and then lets both of them in the digital version of "cat and mouse "The game kills each other.

Both networks use the same data set for training. One of the neural networks is called the generation network. Its task is to generate new pictures according to the pictures you have seen, such as a pedestrian with one arm. The other neural network is called the discriminant network. Its task is to judge whether the picture it sees is similar to the picture at the time of training, or whether it is a "fake" created by the generated model. For example, to judge the three arms. Is it possible for people to be true?

Slowly, the ability to generate a network to create a picture is so strong that it cannot be discerned by the network. Basically, after training, the generation network learned to recognize and create a picture of a pedestrian that looks very real.

This technology has become the most promising artificial intelligence breakthrough in the past decade, helping machines to produce and even deceive human results.

Currently, GAN has been used to create speech that sounds very real, as well as very realistic fake pictures. Take a well-known example. Researchers from the chip company Nvidia used a star photo to train a GAN system, which produced hundreds of face photos that didn’t exist but looked real. . Another research team generated Van Gogh oil paintings that looked very realistic. After further training, GAN can make various modifications to the picture, such as putting a layer of snow on a clean road or turning a horse into a zebra.

But the results of GAN are not perfect: they may produce a bicycle with two sets of handles, or a face with a misaligned eyebrow. But because some pictures and sounds are too realistic, some experts believe that GAN has begun to understand the underlying structure of the world they have seen and heard. And that means that as artificial intelligence begins to gain imagination, they may also begin to understand what it sees in the world.

After inventing GAN, Ian Goodfellow was praised by Daniel, the chief scientist of Facebook, Huang Renxun, founder of NVIDIA, and Wu Enda, founder of Landing.ai, and attracted many institutions and companies to start research. In China, academic institutions are committed to the further improvement and optimization of GAN theory. For example, researchers at the Institute of Automation of Chinese Academy of Sciences are inspired by the human visual recognition process, and proposed a dual path GAN (TP-GAN) for frontal face image synthesis. The Shangtang-Hong Kong Zhongda United Laboratory published a number of GAN-related research results at the International Academic Conference.

The Chinese business community is more inclined to apply technology to services. There are countless examples. For example, Baidu uses GAN to build a speech recognition framework. The combination of GAN and traditional deep learning framework has gained a lot in the field of speech synthesis. Great progress. A paper published by Alibaba's Urban Brain Project team at the ACM MM 2017 conference used GAN to generate training data sets for license plate recognition.

Professional interpretation:

Shang Tang, Professor of the Joint Laboratory of the Chinese University of Hong Kong, Li Hongsheng: GAN may have an impact on computer graphics in the future. It has been developing GAN for more than three years. Although it has been a new technology in the field of artificial intelligence that has been developed for 60 years, it has already There are various variants or advanced versions, and there are still many possibilities in the future with the input of many researchers and companies. For example, the opportunity to progress from a two-dimensional image to a three-dimensional video, etc., may have an impact or challenge on graphics in the far future.

Artificial intelligence for everyone

Reason for selection: Bringing machine learning tools to the cloud will help spread AI more widely

Technology breakthrough: Cloud-based artificial intelligence is reducing the difficulty and price of this technology

Significant: At present, the application of artificial intelligence is ruled by a few companies. But once it is combined with cloud technology, it will be accessible to many people, thus achieving explosive growth in the economy.

The main researchers include: Amazon, Google, Microsoft, Baidu, Tencent, Alibaba, Keda Xunfei, Fourth Paradigm, etc.

Maturity: now

 

Cloud-based machine learning tools are bringing artificial intelligence to a wider community. Today, Amazon's AWS subsidiary dominates the cloud AI market. Google is trying to challenge its position with TensorFlow, an open source artificial intelligence framework that can develop machine learning systems. Cloud AutoML, which Google recently announced, is also a pre-trained system that makes artificial intelligence easier to use.

Microsoft, which joined the cloud service war with the Azure platform, chose to work with Amazon to launch an open source deep learning framework, Gluon. In theory, Gluon can make creating a neural network—an important artificial intelligence technology that attempts to replicate the way people learn about the brain—becomes as simple as developing a mobile APP.

Although we don't know which company will be the leader in the artificial intelligence cloud service market, the winner will definitely get huge business opportunities.

If the artificial intelligence revolution spreads to every corner of the economy, machine learning tools will become a necessity.

Most of today's artificial intelligence technology is only used in the technology industry, bringing efficiency and a variety of new products and services to the field. But other companies and industries have struggled to take advantage of the development of artificial intelligence technology. If artificial intelligence technology can be more comprehensively implemented in industries such as medical, manufacturing, and energy, it will greatly increase the productivity of various industries.

Professional interpretation:

Shen Yichen, co-founder and CEO of Lightelligence: Computational hardware is one of the cores of artificial intelligence. More computational computing hardware can complete neural network training in less time, and is updated by AI processors (such as NVIDIA GPUs). The replacement is very fast, the price is high, and it is more troublesome to replace the hardware. It is not economical for individual users to replace the processor every year, and the cloud computing platform intensively shares limited resources to the public. AI algorithm sharing is also a big advantage of the cloud platform. Some AI algorithms that have been widely used, such as face recognition, speech recognition, image recognition, etc., are very well defined, and the public only needs one algorithm with the best results.

Gene divination

Reason for inclusion: Large-scale genetic research will allow scientists to predict common diseases and personality traits

Technology breakthroughs: Scientists can now use your genomic data to predict your chances of developing heart disease or breast cancer, and even your IQ can be predicted

Significant: DNA-based predictive technology may make a major breakthrough in public health, but it will increase the risk of discrimination

Principal Investigators: Huada Gene, Zhenzhen Bio, WeGene, Helix, etc.

Maturity: now

 

One day, a baby will get a DNA test report when he is born. These reports will provide a baby with a chance of heart disease or cancer, whether it is addictive to tobacco, and whether it is a smarter prediction than the average person. Due to the development of large-scale genetic research (some of which involve more than 1 million people) and scientific advances, such reports will soon turn from concept to reality.

It turns out that the most common diseases and many of the behaviors and characteristics of people, including the level of intelligence, are not the result of one or several genes, but the result of many genes. Using data from ongoing large-scale genetic studies, scientists are creating what they call "multi-gene risk scoring" indicators.

Although the new DNA test only provides probabilistic inference, rather than directly drawing diagnostic conclusions, it can still greatly benefit the development of medicine. For example, if women with high rates of breast cancer have more mammograms and women with lower rates of disease have fewer mammograms, these tests may find more patients with cancer. It also reduces the chance of false alarms. Pharmaceutical companies can also use these scores in clinical trials for preventive drugs for diseases such as Alzheimer's disease or heart disease. By selecting volunteers with a higher risk of illness, they can test the effects of the drug more accurately.

Behavioral geneticist Eric Turkheimer said the new technology is “exciting and worrying” because genetic data can not only benefit us, but it can also be used for other purposes and has a bad influence.

Professional interpretation

MIT-Harvard University's Broad Institute researcher, Tsinghua University visiting scholar Cong Le: genomics research progress combined with large-scale clinical research, enabling scientists to see the future of genes predicting the future.

This field has accelerated in recent years, and has received continuous investment from research institutions such as universities and colleges, startup multinational pharmaceutical companies, and capital markets such as venture capital, allowing researchers to analyze and predict genetic information for humans. The impact of health status, disease risk, and even personal ability such as intelligence will undoubtedly affect many aspects such as medical care, insurance, and education. Although this series of effects has just begun but is developing rapidly, it is unclear whether it is a blessing or a curse.

Genetic information has strong individualized differences and regional differences. Therefore, research and technical results for a group of people in a region may not be applicable to other individuals in the world (such as Asians vs. Europeans and Americans), which will undoubtedly bring more More challenges, but it also means more opportunities. Genetic information has static parts. For example, each person's cells are originally derived from embryonic cells formed by sperm egg binding, and there are also dynamic parts. For example, genetic mutations during development may lead to genetic diseases, or genetic mutations during aging may occur. Leading to cancer and geriatric diseases, it is not enough to do an examination once in a lifetime. The more you do or not, the better and more accurate. You still need a lot of basic and clinical research and development related technologies and data analysis tools.

In addition to the classic DNA genomics information, there are many diseases caused by factors such as genomic modification and RNA expression changes, which we call crown genetics. On the whole, although the development of gene prediction technology has been highly noticed, in terms of the actual development process, we are only seeing the tip of the iceberg of human genetic information.

Sensing city

Reason for selection: Alphabet's Sidewalk Labs plans to create a high-tech community to rethink how it should build and operate a city

Technology breakthrough: One block in Toronto is expected to be the first place in the world to successfully integrate cutting-edge urban design with cutting-edge technology

Significant significance: Smart cities will make urban areas more affordable, livable and environmentally friendly

Principal Investigator: Sidewalk Labs of Alphabet, Toronto Waterfront, Alibaba, etc.

Maturity: The project will be announced in October 2017 and construction is expected to begin in 2019.

 

Today, many smart city plans around the world have been stranded, either by downgrading once ambitious goals or by forcing ordinary residents outside the super rich because of cost of living. And a project in Toronto called Quayside, I hope to redesign a community from the ground up, rebuild it with the latest digital technology, and break the existing failure.

Alphabet's Sidewalk Labs in New York City will work with the Canadian government to bring this high-tech project to the Waterfront industrial district in Toronto.

One of the goals of the project is to make all decisions about design, policy, and information technology based on a vast sensor network. This network will collect a variety of information: air quality, noise levels and people's behavior.

In this plan, all vehicles are self-driving shared vehicles, and the underground will also run the robot responsible for the low-level manual labor. Sidewalk Labs said they plan to open up the software and systems they are designing, allowing other companies to create services on them, similar to developing apps for mobile phones.

The company plans to closely monitor public infrastructure, but this raises concerns about data management and privacy. But Sidewalk Labs believes it can alleviate some of its concerns by working with communities and local governments.

“The most unique thing we did in the Quayside project was that it included not only our huge ambitions, but also a degree of humility,” said RitAggarwala, Sidewalk Labs executive responsible for urban systems planning. And this humility is expected to help Quayside avoid the problems that were often encountered in previous smart city plans.

Currently, several cities in North America are striving to become the next target of Sidewalk Labs. According to WillFleissig, a public sector CEO who manages Quayside, "San Francisco, Denver, Los Angeles, and Boston are all coming to contact us for referrals."

Professional interpretation

Wang Xu, Associate Research Fellow, Center for Eco-Environmental Research, Chinese Academy of Sciences: By 2050, 70% of the world's population will live in cities, and with the rapid expansion of urbanization, the problems of traditional infrastructure and the resulting social, economic and environmental pressures will The increase. Compared to the development and presence of traditional infrastructure, a new paradigm for future urban infrastructure based on sensor construction will build and manage infrastructure in a more integrated and smarter way than the city's energy, transportation and water services. The infrastructure unit or link is managed separately.

However, the current new paradigm of infrastructure, research and application hotspots pay more attention to urban transportation and energy systems, and relatively less investment in other important urban infrastructure systems, such as water and sanitation, and different urban infrastructure. Research attention on the coupling, anti-interference, resilience and sustainability of unit links needs to be improved.

 

Babel fish earplugs

Reason for selection: Although the existing hardware is not so easy to use, Google Pixel Buds shows the prospect of real-time translation

Technical breakthrough: Near real-time translation for multiple languages, and easy to use

Significant significance: In today's increasingly globalized world, language is still a major obstacle to communication.

Main investigators: Google, Keda Xunfei, Baidu, Tencent, Sogou, Tsinghua University, Harbin Institute of Technology, Suzhou University, etc.

Maturity: now

 

In the popular sci-fi classic "Galaxy Roaming Guide", you can hear a real-time translation by stuffing a yellow babe into your ear. In the real world, Google has developed a transitional solution: a pair of earplugs called Pixel Buds worth $159. This pair of earbuds can be translated in real time on the Pixel smartphone via the Google Translate app. One person needs to wear earplugs and the other person holds a mobile phone.

The person wearing the earbuds speaks in their own language - the default is English - and then the Google Translate app translates the words spoken and plays them out loud on the smartphone. After the person holding the phone responds, the answer is translated and then played in the earbuds.

Google Translate has a conversation function before, and its iOS and Android apps can automatically recognize the speaker's language and then automatically translate it. But background noise increases the difficulty of understanding the discourse, and it also makes it difficult for the app to determine when the speaker pauses and when to start translating. Pixel Buds effectively solves these problems because the wearer can tap and lengthen the earbuds on the right while talking. Putting the interactions on the smartphone and earbuds separately allows both parties to control the microphone and help the speaker maintain eye contact, as there is no need to pass the phone back and forth.

Currently, Pixel Buds has been criticized for its design below the industry average. The earbuds look very unintelligible, not very close to the ear, and are difficult to adapt to the phone. However, the hardware is awkward or has something to do. Pixel Buds lets you see the dawn of free communication in near real-time translations across cross-language barriers, and you don't have to stuff a barb fish into your ears.

Professional interpretation

Wei Furu, Senior Researcher/Research Manager, Microsoft Research Asia: End-to-end Neural Network Translation (NMT) based on sequence-sequence has greatly improved the quality and level of machine translation in recent years. One of the biggest breakthroughs and achievements in the field of natural language processing, the advancement of NMT technology has further triggered innovation in related services and hardware.

 

Quantum leap of material

Reason for inclusion: Researchers have recently begun to model simple molecules using quantum computers, and this is just the beginning.

Technological breakthrough: IBM successfully simulated the electronic structure of small molecules using a 7-qubit quantum computer

Significant: With this technology, scientists can understand all aspects of the molecule and develop more effective drugs and new materials that generate or transfer energy more efficiently.

Principal Investigators: IBM, Google, Harvard University Professor Alán Aspuru-Guzik, University of Science and Technology of China, Chinese Academy of Sciences, Zhejiang University, Alibaba, etc.

Maturity: 5 to 10 years

 

The new quantum computer is powerful, but its development path still hangs on a fog: quantum computers have the computing power that today's computers can't match, but we have yet to figure out what this ability can be used for. An infinite application direction is beckoning in vector sub-computers: precise molecular design.

For many years, chemists have dreamed of designing new proteins for more effective drugs, or designing electrolytes in new high-efficiency batteries, magical compounds that directly convert solar energy into liquid fuels, and more efficient solar cells. . However, the material molecules in these technologies are difficult to model and simulate on computers, and the paradox is designed and synthesized. Even the task of simulating the electronic form of a simple molecule can be complicated to defeat an existing computer. However, this is a piece of cake for quantum computers.

Compared with traditional computers, "1" or "0" digital bits (Digital Bits) are used as calculation and storage units, and quantum computers use quantum bits (Qubits) of quantum systems as arithmetic units. Recently, IBM researchers applied a 7-bit quantum computer to simulate a triatomic molecule.

Nowadays, scientists are building quantum computers with more qubits, and quantum algorithms are also elevating. It is also possible to perform accurate simulation calculations of macromolecules that we are more interested in.

In fact, China has also experienced considerable growth in quantum computing. Although the current level of technology cannot be compared with the previous big companies, it is gradually catching up with the cooperation of industry, academia, and the government. The footsteps of the leader.

In May 2017, the Chinese Academy of Sciences announced the official birth of a light quantum computer jointly developed by the Chinese Academy of Sciences, the Chinese Academy of Sciences, the Alibaba Quantum Accounting Laboratory, the Zhejiang University, and the Institute of Physics of the Chinese Academy of Sciences. In addition, on October 11 of the same year, the Chinese Academy of Sciences and Alibaba Cloud launched a quantum computing cloud platform. The commercialization of quantum computing is close at hand, and the speed is not inferior to that of Europe and the United States.

However, quantum computing still has a lot of breakthroughs. First, the accuracy of quantum computing is quite low. Although it is quite suitable for calculations with low precision requirements such as deep learning, it is possible to deal with the general calculation of traditional computers. The power has not been caught. Second, the highly parallel computing environment of quantum computing requires the adaptation of the framework and the targeted optimization of the compiler, which is completely different from the existing computing architecture.