As predicted in the last report, in 2017 the development and testing of unmanned trucks engaged in many companies: Uber, Waymo, Daimler. And truck companies Embark has set a record, driving the U.S. from coast to coast. Gene therapy, which journalists called one of the most promising technologies of 2017, has become gradually introduced into medicine. The United States Department of health approved three gene drug: two of them are struggling with cancer, and the third with blind.
3D printing of metals
Although three-dimensional printing has been around for several decades, for a long time it remained the prerogative of the experimenters and designers who are creating disposable prototypes. However, in recent years the situation is changing: more and more companies are developing inexpensive 3D printers focused on the production of large metal parts. For example, in 2017 the American startup Markforged has released its first printer for 3D printing of metal cost less than $100 thousand And the company Desktop Metal developed software that independently designs model details and allows you to speed up the production 100 times. In the same year, the company General Electric, which has long been testing three-dimensional printing for aircraft construction, showed a prototype 3D printer designed to print parts with a diameter up to 1 meter. The company plans to start selling the printer in 2018, simultaneously developing printer to create larger parts.
This year researchers from the University of Cambridge for the first time raised the embryo stem cells without the use of egg and sperm. The experiment was conducted on mice, but theoretically it is possible to replicate on human cells.
The scientists used embryonic stem cells and extraembryonal trophoblast mouse stem cell that forms the placenta, as well as extracellular matrix — three-dimensional framework on which cells outside the body to grow and form tissues. 96 hours of the experiment, the cells formed a structure with the same characteristics as the mouse embryo 96 hours after fertilization. “And the embryo, and extraembryonic cells interact with each other and organized in a structure that looks and behaves like an embryo. It has anatomically correct departments that are developing in the right place at the right time,” explains chief researcher of the project Magdalena Zernicka-Goetz.
After 96 hours of embryo development stopped because he had no source of oxygen and nutrients. In subsequent experiments, the scientists are going to put on the 3D frame also stem cells of the yolk SAC, which, according to preliminary estimates, can extend the development of the embryo prior to the formation of bodies.
In October, subsidiary of the holding company Alphabet called Sidewalk Labs said plans to build in the coastal the coastal part of the Toronto high-tech Quayside quarter — a “sentient city”, which will be covered by a network of sensors to collect detailed data on the operation of the city from the air to the level of noise. Based on these data, Sidewalk Labs is going to develop software to optimize the operation of the city — for example, to control traffic. All software created for the Quayside, the company promises to publish in open access so that other developers can use it in their projects of smart cities.
According to the plan, all vehicles will be Autonomous and available for sharing. Routine work such as mail delivery will be done by robots. Quayside also will test a modular construction, which is expected to help cope with the housing crisis in Canada. Project budget: $1 billion start of construction is scheduled for 2019.
Also last year, bill gates invested $80 million in the construction of smart city in Arizona. He claims that this city will be an ideal platform for the development of the latest technologies, such as unmanned vehicles, “smart” intersections, data centers and high-speed communications network.
“Artificial intelligence for all”
Until recently, development in the field of artificial intelligence has been the privilege of major technology companies such as Amazon, Baidu, Google and Microsoft. However, the development of cloud technologies makes AI development accessible even for start-UPS without serious funding. Microsoft, which has its own cloud platform, Azure, has teamed up with Amazon to create a library of deep learning open source Gluon. It is assumed that with the power of a Gluon to create the neural network will be as easy as a standard app for a smartphone. Google also announced the launch of a cloud-based system for AI-development Cloud AutoML.
Writes MIT, the advent of affordable cloud platforms will enable more common use of AI technology in medicine, energy and industry.
Adversarial neural networks
In 2017, the neural network learned to accurately recognize the image, but also to create in their image new — sometimes indistinguishable from real photographs.
Generative adversarial network (GAN), used for this purpose, consist of two parts: a generator and a discriminator. They are trained on the same data set in the experiment Nvidia, for example, the photographs of Hollywood stars. The generator then learn to create new images in their image, the discriminator checks them for “realism.” “The main idea is to gradually increase the generator and to the discriminator, since the low resolution images, and then, as learning to add new layers that relate to parts with higher resolution,” — explain the representatives of Nvidia.
In the same year, the company has used its development when creating the training set for control systems of an unmanned vehicle. According to the developers, using adversarial neural networks and unsupervised learning, they were able to develop the artificial “imagination” — the ability to imagine how it would look like depicted object when other conditions, for example, when another weather. This means that in the future for training recognition algorithms will not need to collect and manually label large databases: they will be sufficient to provide one original image and the generator independently processes them according to the given conditions.
So the network turned a winter road in summer, and dry in the wet.
Set the left image to the right is generated. Nvidia
In a science fiction novel by Douglas Adams “Hitchhiker’s guide to the galaxy” for instant translation of unknown speech, the characters were implanted in the ears of Babel fish. This year Google released a similar gadget — Pixel Buds headphones for $159, which c is transferred almost simultaneously and you know 40 languages. While they only work in pairs with smartphones Pixel. To speak with a native of another language, the user must wear headphones in and your phone to keep close to the source. Buy Pixel Buds in the US, UK, Australia and Germany.
The device operates on the basis of Google Translate, which is much wiser over the past two years. According to Google, the new version is called Machine Translation Neural admits 55-85% less errors than the previous version of the translator.
Environmentally friendly natural gas
In the US, Net Power built an experimental gas power plant, which is due to new technologies for processing of fuel will not throw CO2 into the atmosphere. For the combustion gas turbine used is 10 times smaller than conventional plants. She is placed in a room with an area of 5.6 square meters, and the area of the whole plant does not exceed the size of a football field.
A large part of the carbon dioxide emitted from natural gas under high pressure and temperature, directed at maintaining the operation of the turbine and the other is compressed to a dense state. It can be buried in the ground, or to sell to other companies — for example, for the production of cement, plastics and other materials out of carbon.
Network pilot power plant will be connected in 2018. It will produce 50 megawatts of power — enough to power 40 thousand homes. It is assumed that the efficiency of the station will be 80%. For comparison, conventional power plants in the United States on average work with an efficiency of 60%.
A real online privacy
True privacy on the Internet, finally, could become possible thanks to a new cryptographic Protocol called the proof with zero disclosure (zero-knowledge proof). With this Protocol the user can, for example, to prove their solvency, without disclosing how much money is in his account. The Protocol to zero disclosure are already using us banks JP Morgan Chase and ING.
Although scientists have been working on this technology for 10 years, she became known only last year, when the zcash for the cryptocurrency. Unlike bitcoin and other cryptocurrencies, zcash for provides users with maximum anonymity: all transactions are confirmed in a distributed registry, but the track addresses of participants impossible. This method is used for cryptography zk-Snark, developed by Israeli scientists in cooperation with MIT. All transactions are confirmed in the network, but keep track of the addresses of the participants of the operation impossible.
Due to a number of genetic studies that scientists have conducted in recent years, theoretically, already at the analysis of human DNA to determine predisposition to most diseases. For example, in California, doctors are testing a program that the DNA of the patient will be able to predict the probability of development of his coronary disease. In addition, this year the company Myriad Genetics has launched an algorithm that can calculate the chance of breast cancer any woman of European descent. These developments are based on a polygenic analysis — that is, the study of thousands of genes of the patient, and not one that significantly improves the accuracy of the estimation. Scientists project that within three years the technology of genetic prediction will be implemented in health care to allow individually for each patient to match the scheme of vaccinations and surveys.
Pharmaceutical companies will also be able to use genetic predictions in the development and testing of their products. Knowing what diseases affected most of the population, they will be able to produce the most popular drugs and testing their effects on the volunteers with the highest chances of the disease.
Material quantum leap
In 2017, researchers have begun to study the prospect of creating new molecules with a quantum computer — a computational device which for data transmission and processing uses the phenomena of quantum superposition and quantum entanglement, and so it can calculate 100 million times faster than a conventional computer.
Scientists from IBM have managed to create a simple triatomic molecule using the quantum computer with the capacity of 7 qubits. MIT suggests that in the near future scientists will be able to create more complex and interesting molecules, because now is the development of quantum computers with many qubits and more complex algorithms. So, this year, Microsoft released free beta version of the developer kit software designed for a future quantum computer. The kit includes a new programming language Q# (Q Sharp), developed by the company specifically for quantum computing and the simulation of a quantum computer with the capacity of 40 qubits, which will allow developers to test their programs on a regular computer or through a cloud computing platform Azure. Microsoft assures that all programs created using this set in the future will be able to work on a real quantum computer, which is the development company.