DeepMind trained an AI agent and learned the ability to cut short-term like a mammal.
Recently, DeepMind has released new Nature papers! This time, they trained an AI agent and learned the ability to cut short-range similar to mammals. The purpose of this study was to try to imitate the human brain and use complex Way to navigate around the space. This is an unprecedented exploration and is considered to be a major step forward in understanding the brain. Recently, Deep Nature has released a new Nature paper! This time, they trained an AI agent and learned the ability to “snap closer†like a mammal. The purpose of this study was to try to imitate the human brain and navigate in a complex way in the surrounding space. This is an unprecedented exploration and is considered to be a major step forward in understanding the brain. Spatial perception is still a difficult task for AI that has easily become a Go master. What's going on in the brain as you move along familiar streets, get around obstacles, and find the shortest way to get to your destination? This is a very complicated matter. Scientists have found three kinds of cells in the animal and human brains that are related to road recognition: positional cells, directional cells, and grid cells. Positional cells can be discharged when the subject reaches a specific location, thereby giving memories of past locations; directional cells can sense the direction of advancement; mesh cells are the most mysterious: they can divide the entire space environment into a honeycomb of six The edge grid, like the coordinate system on the map. The Mosuls who discovered the cells of the mesh thus won the 2014 Nobel Prize in Physiology or Medicine. However, are grid cells only providing GPS location services in a space environment? Some scientists have speculated that they will also participate in vector calculations to assist animal planning. The DeepMind team decided to test the above conjecture with an artificial neural network. Artificial neural network is an arithmetic structure that uses multi-layer processing to simulate brain neural network. The team first used a deep learning algorithm to train the neural network to learn the mammal's foraging movement path and use the linear velocity, angular velocity, and other signals to locate in the visual environment. The researchers then discovered that a structure similar to the grid cell activity feature was born automatically! In the previous training, the researchers did not deliberately guide the neural network to produce this structure. The DeepMind team then used reinforcement learning to verify that this grid structure is capable of vector navigation. Reinforcement learning is commonly used to train game AI. Humans tell AI a game's scoring reward and punishment mechanism, but they do not teach game methods. AI evolves itself in the process of repeatedly playing games and striving for higher scores. The researchers combined the previously-emerged grid structure with a larger neural network architecture into an artificial agent and placed it in a virtual reality gaming environment. After intensive learning, the artificial intelligence advanced beyond the average person to the destination in the game maze, reaching the level of professional game players. It can find new routes and shortcuts like mammals. The most critical point is that when the researchers “quietly†the original grid structure, the navigation capability of the artificial agent will become weak, and the distance and direction of the judgment target will be more inaccurate. One of the authors of the paper, Dharshan Kumaran, said: “We have demonstrated that grid cells are far from providing us with GPS positioning signals. It is also a core navigation mechanism by which the brain calculates the shortest distance between two locations.†Hasabis said that to prove that we can build the kind of universal intelligence that we want to do, the existence of the brain is an evidence. So it's reasonable to find inspiration for new algorithms from neuroscience. But we also believe that this kind of enlightenment should be bi-directional, and the insights of artificial intelligence research can also provide inspiration for open problems in neuroscience. “This work is a good example: By building an artificial agent that can navigate in a complex environment, we emphasized the importance of the bio-grid cells in mammalian navigation and broadened this understanding.†Hasabis said. The DeepMind team believes that similar research methods can also be used to explore brain auditory mechanisms and control the limbs. In the more distant future, neuroscientists can even use artificial intelligence instead of mice to do experiments. At the same time, artificial agents have been used to verify that "grid cells" in the mammalian brain support vector-based navigation. There are also two video interviews with researchers that interpret it in more detail. paper The title of this paper is: Vector-based navigation using grid-like representations in artificial agents. There are more than 20 authors, including Hasabis himself.
Electronic Lock Motor product introduction:
Electronic Lock Motor, also known as Intelligent lock motor, Belong to small reducer series, the main structure composed of drive motor (motor) with a reduction gear box, composed of a shaft connection, form integrated motor reducer; Drive motor is purchasing product, reduction gear box is customized according to the project product demand, the maximum power under 10 w, maximum voltage up to 12 v, below 25 mm diameter size.
Functions: it is widely used in various intelligent locks, such as Shared bicycle lock, fingerprint lock, glass lock, safe lock, hotel lock, family lock, inductive smart lock, etc
Features:China Electronic Lock Motor is based on micro precision reducer development, small intelligent micro precision reducer can do 10 mm in diameter, is widely used in safe, locks, smart locks, sensor intelligent lock;
This kind of motor has the characteristics of low noise, low energy consumption, small volume, light quality, high precision, large torque and durability
Method of use: the best stable in horizontal plane, installed on the China Electronic Lock Motor output shaft parts, cannot use a hammer to knock, knock prone to press into the dc Electronic Lock Motor drive, may cause damage to internal components, and cannot be used in the case of blocked.
Operating temperature range:
Electronic Lock Motor should be used at a temperature of -10~60℃.
The figures stated in the catalog specifications are based on use at ordinary room temperature catalog specifications re based on use at ordinary room temperature (approximately20~25℃.
If a electronic Lock Motor is used outside the prescribed temperature range,the grease on the gearhead area will become unable to function normally and the motor will become unable to start.Depending on the temperature conditions ,it may be possible to deal with them by changing the grease of the motor's parts.Please feel free to consult with us about this.
Storage temperature range:
China Electronic Lock Motor should be stored ta a temperature of -15~65℃.
In case of storage outside this range,the grease on the gearhead area will become unable to function normally and the motor will become unable to start.
Service life:
â—Use with a load that exceeds the rated torque
â—Frequent starting
â—Momentary reversals of turning direction
â—Impact loads
â—Long-term continuous operation
â—Forced turning using the output shaft
â—Use in which the permitted overhang load or the permitted thrust load is exceeded
â—A pulse drive ,e.g.,a short break,counter electromotive force,PWM control
â—Use of a voltage that is nonstandard as regards the rated voltage
â—Use outside the prescribed temperature or relative-humidity range,or in a special environment.
â—Please consult with us about these or any other conditions of use that may apply,so that we can be sure that you select the most appropriate model.
when it come to volume production,we're a major player as well .each month,we rurn out 600000 units,all of which are compliant with the rohs directive.Have any questions or special needed, please contact us, we have the engineer group and best sales department to service to you
Looking forward to your inquiry. Welcome to our factory.
Electronic Lock Motor,Motor For Electronic Lock,Electronic Lock Dc Motor,Electronic Lock High Speed Motor Shenzhen Shunchang Motor Co., LTD. , https://www.scgearmotor.com
