Add idea


«    May 2019    »










DateDate: 18-05-2019, 05:58

Modern smartphones, laptops and other mobile devices often have a good performance margin, but usually after a year of active use, the batteries noticeably lose in capacity. Professor Atsuo Yamada from the University of Tokyo has taken up the challenge. He and his team found a way to not only extend battery life, but also make them self-healing.
The most popular type of battery today is lithium-ion, and sodium can replace it. For both types, charging cycles reduce capacity over time. The fact is that inside the battery there are metal layers. During charging and discharging, they decompose and cracks form in them, which reduces the ability to save energy. These defects are due to the fact that the material is held by van der Waals force, which is not able to withstand the load on the material during charging and battery operation.
Japanese engineers have found a way to eliminate the deformation of the metal layers inside the batteries. They clearly demonstrated that when used in the production of oxygen oxide batteries with a redox layer (Na2RuO3), degradation practically does not occur. During charging and discharging, not only is the degradation of the battery reduced, but also the properties of self-healing appear. This is because the new material is held by a force known as Coulomb attraction.
“This means that batteries can have a much longer service life. Increasing the energy density of batteries is of paramount importance for the implementation of electrification of transport, said Professor Atsuo Yamada.

DateDate: 17-05-2019, 05:54

Understanding and translation of lively speech is much more complicated than just processing written text. Initially, the speech must be interpreted by the machine and deciphered in the text, which will then be translated into the desired language. The company is currently working to accelerate and accelerate this process.
Google decided not to stand still and did not develop a more natural method of converting oral speech, called Translatotron. This method, unlike the old one, does not require intermediate speech processing.
In fact, Google has already made this feature available for your application a live speech right while talking online. As I said, these functions typically use a cascade data processing method to perform their functionality. Interpretation of speech that can occur during conversion, and may even ignore words that do not need to be translated. Translatotron can keep the original voice of the speaker in the language selected for translation, the dialogue dialogue is more natural.
In fact, this experiment is still under development. All these results are still excellent. This means that the translatotron can imitate the original voice, which sounds very timid, unnatural and slow. overcoming language barriers for many users. A source

DateDate: 16-05-2019, 05:54

The scope of artificial intelligence is becoming wider every year. Neural networks are able not only to process photos, but also to paint pictures themselves, or even turn videos into computer game levels. According to the data of the new research, until the appearance of a machine with “human” thinking, it was not very long. Scientists have discovered a computer's ability not only to logical, but also to abstract operations - which he learned independently.
The difference between human and computer thinking lies not only in the speed of computation, but also in the ability to think abstractly. In order for a computer to be able to count something, it must first determine what it will count: for example, whole numbers or certain objects specified by the program. The latter task is complicated by the fact that many features of objects (shape, color, and size) may be similar in completely different objects. A man, looking at four cats, four apples and the number 4, intuitively knows that they all have a common feature - the abstract concept of "four". That was the difference between the biological and electronic brain - until recently.
The scientific journal of the American Association for the Advancement of Science Science Advance published the results of a study in which artificial intelligence independently developed abstract thinking. During the experiment, a neural network designed for the simple detection of visual objects, was able to spontaneously develop a “human” approach to the calculation of objects.
The researchers found that certain parts of the network suddenly “tuned in” to abstract numbers, showing signs of neuron interaction in the real brain. AI "understood" that the image of four apples looks like the image of four cats - because the number of objects in both cases coincided.
According to the developers, their plans did not include the training of artificial intelligence in this method of calculation, and now they will have to figure out how he developed an unusual ability in himself. In the future, the use of similar algorithms can significantly expand the scope of neural networks.

DateDate: 15-05-2019, 05:55

It is reported by Hightech +, reports Sotsportal.
- Single malt Ai Whiskey from the Swedish company Mackmyra debuts on the market this fall. The price of a unique bottle is not called yet. But the creators claim that this is the first such drink, the recipe of which was created not by experts, but by the neural network, the author noted.
The algorithm has reworked all the techniques available to Swedes, taking into account sales volumes and the availability of certain barrels in the warehouse, in which alcohol "matures".
As noted by Popular Mechanics, experts, people for each new recipe take months, if not years of work.
And the nuances of taste that are important for lovers of strong drink determine little things like varieties of grain or the age of barrels in which whiskey is kept.
The task of generating new recipes AI, working on the cloud platform Microsoft Azure, coped even too well: the model gave out 70 million recipes that can be implemented, having received a delicious flavored drink at the exit.
- Mackmyra claims that human intervention was required at this stage. Experts first chose the most incredible recipes, to which people would not have thought of, and then - the most promising option in terms of taste. After all, this part of the AI whiskey can only be estimated theoretically, - the author is sure.
Subtleties of the technology proposed by the algorithm, the Swedes do not describe. But a lot of attention is paid to the description of taste: single-malt AI Whiskey - bright golden-yellow color, with herbal notes of anise, ginger and white pepper, citrus, spicy and dry aftertaste.
Mackmyra and Microsoft claim that their collaboration is the first time that AI creates a recipe for a complex consumer product. However, as Popular Mechanics reminds, previously at least three companies - including the market leader Carlsberg - experimented with creating AI beer recipes.

DateDate: 14-05-2019, 05:56

The device received a body with variable geometry and four screws.
Engineers from Ben-Gurion University created a robot that is also a wheel device and a quadcopter. The housing with variable geometry and four screws allows it to take off in order to overcome high obstacles, writes “High-tech”.
The device consists of several parts. In the center are the battery and the board for management. The two side built-in electric motors that drive the wheels and screws. The robot is able to change the angles between the side parts, which allows it to switch between wheels and screws, depending on the type of task.
Technical details are not disclosed yet. It is known that the drone can carry loads weighing up to 400 grams. He can change the types of movement almost instantly.