Neural networks-the duelists. Artificial embryos. Artificial intelligence in the cloud. What other technologies MIT Technology Review has called the most important in 2018? Annual compilation of technologies that can change the world of the future, going with 2001. People often ask what a “breakthrough” in technology? On this question the answer may not always be obvious because some of the technology is just not widespread, however, other commercially unrealizable. But they somehow have an impact on our lives.
This year a new technique of artificial intelligence called GAN gives the imagination machines; artificial embryos, despite the fears of the sceptics, to review the creation of life and open a huge window for studies of the first moments of human life; experimental setup in the center of the petrochemical industry of Texas is trying to create a completely clean electricity from natural gas is probably the main energy source of the future. Go.
3D printing metal
Although 3D printing has been used for quite some time, she remained mostly in the field of hobby and designers, producing single prototypes. And print the object out of anything, not counting the plastic, for example, metal was expensive and painfully slow.
However, now it becomes cheap and simple enough to be potentially practical method for the production of parts. If it is widely adopted, it could change the way we produce many products of mass production (sorry for the tautology).
In the short term, producers would not need to store a large inventory, they could just print the object, any replacement part for an aging car, if someone is needed.
In the long term, large plants producing a small range of parts in high volume, could be replaced by smaller plants, but with a range more suitable to the changing needs of our clients.
3D printing method allows you to create lighter and stronger parts and to reproduce complicated shapes that would be impossible to create using traditional methods of metal processing. Also it would be possible to devote more attention to the microstructure of metals. In 2017, scientists from the National laboratory Lawrence Livermore stated that it had developed a method of 3D printing, allowing you to create parts from stainless steel, which is twice as strong as traditional.
Also in 2017, 3D printing Markforged is a small startup located near Boston, presented the first 3D metal printer for just $ 100,000.
Another Boston startup Desktop Metal began to ship the first machines for metal prototyping in December 2017. He plans to start selling more cars, especially for manufacturers that will work 100 times faster than the old methods of metal printing.
Printing of metal parts also becomes easier. Desktop Metal now offers software that creates a design ready for 3D printing. Users set the program specifications of the object that they want to print, and produces a computer model suitable for printing.
General Electric, which has long been a proponent of 3D printing in their aviation product, working on a test version of its new metallocenter, which quickly works to create great parts. The company plans to begin selling this device in 2018.
In a breakthrough that could change our view of the creation of life, embryologists at the University of Cambridge in the UK have grown to be quite realistic in view of the mouse embryos using only stem cells. No eggs. No sperm. Just take the cells from other embryos.
Scientists gently placed the cells in a three dimensional forest and spellbound watched as they began to communicate and to form well distinguishable form of mouse embryo of the age of a few days.
“We know that stem cells possess magical powers and great potential. We didn’t think that they will be able so beautiful or even perfect to organize themselves,” said steering group of scientists Magdalena Zernicka-Goetz.
Zernicka-Goetz says that its “synthetic” embryos probably would not have been able to develop in mice. Nevertheless, it is a hint that soon we can see mammals, born without egg at all.
However, the goal Zernicka-Goetz puts. She wants to study how cells of the early embryo begin to take their role. The next step will be the creation of an artificial human embryo from stem cells. This work is being conducted at the University of Michigan and Rockefeller University.
Synthetic human embryos would be a boon to scholars, as it allowed to observe all events at an early stage of fetal development. And because these embryos come from stem cells, easily manipulated, a laboratory could use the whole range of tools, such as editing of genes for study of their growth.
Artificial embryos, however, forced to ask ethical questions. What if they become indistinguishable from the real thing? How long they can be grown in the lab before they start to feel pain? We need to address these issues before the scientific race to go too far.
The city with feelings
A variety of schemes of smart cities faced delays, cancellations, impossible plans or expensive. A new project in Toronto called the Quayside needs to change this already established scheme of things, rethinking it in terms of the urban environment from scratch and build it around the latest digital technology.
Sidewalk Labs, which is owned by Google and is located in new York, working with the canadian government, working on this high-tech project designed to industrial waterfront Toronto.
One of the goals of the project is to base decisions on design, policy and technology on information obtained from extensive network of sensors that collect data about everything from air quality and finishing level of noise and human activity.
According to the plan, all vehicles should be common and independent. Under the ground, will scurry robots doing grunt work like delivering the mail. Sidewalk Labs says open access to the software and systems which creates that other companies could create services on top of them, as people are creating applications for mobile phones.
The company intends to closely monitor public infrastructure, and this, of course, raises questions of privacy and data. However, Sidewalk Labs believes that the joint work of local governments and free community will allow you to get rid of these feelings.
“We’re trying to do from the Quayside not only an ambitious project, but quite humble,” says RIT of Aggarwal, Manager of planning urban systems in Sidewalk Labs. This humility can help Quayside avoid the pitfalls that plagued previous initiatives of smart cities.
AI for everyone
Artificial intelligence still was toy big tech companies — Amazon, Baidu, Google, Microsoft, and several startups. For many other companies in AI systems too expensive and complex for full implementation.
Solution? The machine learning tools, built on a cloud Foundation that will lead the AI to a wider audience. Currently Amazon dominates the cloud AI with its AWS. Google competes with it with the help of TensorFlow, open library of the AI, which can be used to create another ON machine learning. Not so long ago Google announced Cloud AutoML, a set of pre-trained systems that simplify the use of AI.
Microsoft, which has its own cloud platform Azure, teaming up with Amazon to create a Gluon, the open library of deep learning. Gluon will help in the construction of neural networks — a key technology AI, which roughly mimics the learning process in the human brain — they will be as easy to do as the smartphone application.
It is not clear which of these companies will become the leader in the field of cloud services that offer AI. But the winner this promises a huge opportunity for business. Especially if the AI revolution will spread to different parts of the global economy.
Currently, AI is used mostly in the technology sector, where performance improves the old and creates new products and services. But other types of business and industry want to benefit from the use of artificial intelligence. The segments such as medicine, manufacturing, energy, easily converted if fully implemented the latest technology in the field of artificial intelligence.
The majority of companies still don’t know how to use a cloud-based AI. So Amazon and Google forge work and consulting services. As soon as AI will be available to all, will start a revolution.
Neural networks-the duelists
Artificial intelligence has learned very well to identify objects, show him a million pictures, and it can be amazingly accurate to find a pedestrian crossing the street. But the AI manages to create a bad image of pedestrians. If he was capable of it, he wove the tapestry of realistic but synthetic images of pedestrians in different settings. Self-driving cars could use this information for training, not even leaving the road.
The problem is that creating something new requires imagination. And this still remains the prerogative of man.
Ian Goodfellow in 2014, said the decision. The approach, known as generative adversarial network, or GAN, takes two of the neural network is a simplified mathematical model of the human brain that lie at the core of modern machine learning and puts them against each other in a digital game of cat and mouse.
Both neural networks are trained on the same dataset. One neural network — generator — generates variations of images that she has seen — for example, the image of the pedestrian with his third hand. The second is the discriminator to determine whether it has reviewed example is similar to the picture that she saw, or a fake produced by the generator — that is, whether three-handed man to be real?
Over time, the alternator will learn to create a good image so that the discriminator will not be able to distinguish the fake from the original. Thus, the generator learns to recognize and then create realistic images of pedestrians.
This technology has become one of the most promising advances in AI over the last decade.
GAN was used to create realistic hearing speeches and photo-realistic fake images. In one example, scientists from the NVIDIA instructed GAN to create hundreds of trustworthy individuals of non-existent people. Another group was able to make quite a convincing fake image, like the work of van Gogh. GAN can alter the image in different ways — to transform the sun road in snow, horses zebras.
The results are not always perfect. But as images or sounds are realistic, some experts believe that the neural network begin to understand the internal structure of the world. Perhaps they conceived the imagination.
Headphones Babel fish
In the cult science fiction classic “the Hitchhiker’s guide to the galaxy”, you could put in ear yellow Babel fish and get instant translation. In the real world, Google is a temporary solution: a pair of earphone Buds Pixel for $ 159. They work with smartphones and Google Pixel.Translator, providing almost instantaneous translations.
One person wears the headset and the other holds the phone. Wearing the headset speaks its language — for example, in English and the application translates and plays loud translation on the phone. The person holding the phone, responds; this response is translated and replayed in the headphones.
Google Translate can already hold a conversation, and its applications for iOS and Android allow the two users to talk to: they automatically detect languages and translate them. However, background noise difficult to understand the application of what people say, and also the application determines when people begin to speak and when to stop.
Pixel Buds solves these problems, because the owner presses and holds the finger on the right earbud during a call. The interaction between phone and speaker allows you to maintain eye contact because you do not have to be distracted by the phone.
Pixel Buds look weird and bad are ears. They are also difficult to configure with your phone. But hardware problem is the least of all. Main — concept. And fish is not needed.
Natural gas with zero emissions
The world is probably stuck on natural gas as one of main energy sources in the future. Cheap and affordable energy natural gas currently provides 30% of electricity in the United States and 22% of the electricity in the world. And although it is cleaner coal, carbon emissions from it are still significant.
Experimental plant near Houston, the hub of the oil and refining industry, the US is testing technology that could make clean energy from natural gas a reality. A company with a 50-megawatt project Net Power believes that it can produce energy as cheaply as conventional stations, natural gas, and, importantly, to capture all the carbon dioxide, which is released in the process.
If so, the world will produce carbon-free energy from fossil fuels at a good price. Such gas stations it would be possible to turn on and off depending on demand, avoiding the need to supply renewable energy. Net Power combines several energy companies, including 8 Rivers Capital, Exelon Generation, and CB&I. the Company has already started initial testing and provide results in the coming months.
Station places the carbon dioxide released in the combustion process of natural gas under high pressure and temperature using supercritical CO2 as a “working fluid” that drives a special turbine. A large part of the carbon dioxide can be recycled continuously; the rest will be captured quickly and easily.
A key part of cost reduction depends on the sale of carbon dioxide. Today its main use is to extract oil from oil wells. This is a limited market and certainly not the most environmentally friendly. But Net Power plans eventually to see increased demand for carbon dioxide in the production of cement, plastics and other materials based on carbon.
Technology Net Power will not solve all problems with natural gas, especially in terms of production. But while we use natural gas, we can try to make it as clean as possible. Of all the technologies that are being developed in the field of clean energy technology Net Power is one of the most important.
Perfect integrity online
Real inviolability of personal data and online life can become possible, thanks to a new tool that, for example, can prove that you are 18 years old without revealing your date of birth, or that your Bank account has enough money without revealing your account balance. This limits the risk of identity theft.
What is this tool? Cryptographic Protocol called zero knowledge proof (zero-knowledge proof). Although researchers have long developed, great interest was manifested last year, thanks to the growing cryptocurrency market.
A large part of the practical application of zero knowledge proof was done by the team zcash for digital currency, which appeared at the end of 2016. The developers used a technique called zcash for zk-SNARK, giving users the ability to make anonymous transactions.
In Bitcoin and most other public blockchains that would be impossible, because the transactions are visible to everyone. Although these transactions are theoretically anonymous, they can be combined with other data to identify users. Buterin Vitalik, founder Ethereum, the second most popular blockchain network мирt, described the zk-SNARK as “an absolutely revolutionary technology.”
Banks could use payment systems without revealing personal information. Last year, JPMorgan Chase added a zk-SNARK in its own payment system on the basis of the blockchain.
And yet, despite all the advantages, the zk-SNARK is complex and slow algorithm. However, alternatives are being actively developed and one day will replace the existing options.
Once all newborn babies will receive a card with the DNA. It will be painted, what are the chances of getting a heart attack or cancer, get hooked on cigarettes or to become a Prodigy.
The science that made it possible, unexpectedly, thanks to the powerful genetic research. Some of them participated more than a million people. It turned out that the most common disorders, but also personality traits like intelligence are a result of one or more genes, caught up in that or in the wrong place at the right time. Using data from a large genetic research, scientists create a “polygenic”risk assessment.
Although new DNA tests provide probabilities, not diagnoses, they can significantly help physicians. For example, a woman with an increased risk of breast cancer is more to do mammograms and checkups.
Pharmaceutical companies can also use risk assessment to create preventive drugs for diseases like Alzheimer’s or heart attack. Collecting volunteers who are likely to get sick, they can determine the quality of the drugs.
Still, the forecasts are far from perfect. In addition, polygenic evaluation still questionable, since predicting traits, and not only the possibility of diseases. For example, they are now with 10% probability predict the intelligence factor of an individual. But as parents and educators take advantage of this information?
A quantum leap in materials
The prospect of a powerful new quantum computers are full of uncertainties. They will be able to perform calculations, an elusive forces in modern machines, but we are not yet fully realized, what can you do with this power.
One of the possible applications: the exact design of the molecules.
Chemists already dream of creating new proteins for more effective drugs, new electrolytes for best battery life, compounds that are able to directly convert sunlight into liquid fuel, and more efficient solar cells.
We have no such, because the molecules are incredibly hard to simulate on a classical computer. Try to mimic the behavior of electrons in a relatively simple molecule and you will face difficulties that go far beyond the capabilities of modern computers.
But for quantum computers are quite capable of it, because instead of digital bits that represent ones and zeros, they use “qubits”, which are themselves quantum systems. Recently, scientists from IBM used a quantum computer with seven qubits for simulation of small molecules of three atoms. Very soon it will be possible to accurate modeling of larger and more interesting molecules, as well as the best quantum algorithms.