News classification
Contact us
- Add: No. 9, North Fourth Ring Road, Haidian District, Beijing. It mainly includes face recognition, living detection, ID card recognition, bank card recognition, business card recognition, license plate recognition, OCR recognition, and intelligent recognition technology.
- Tel: 13146317170 廖经理
- Fax:
- Email: 398017534@qq.com
Robots are teaching each other, and artificial intelligence AI is evolving
Robots are teaching each other, and artificial intelligence AI is evolving
At a press conference on October 2015, Tesla demonstrated their research and development of Model S automatic driving function, this function allows a car can stop semi autonomous driving, and Tesla CEO Elon Musk also mentioned that each vehicle has to become the "private coach Model S each expert of this car". In fact, each Model S can learn to improve its autonomous characteristics from the driver's body, but more importantly, when a Tesla car learns knowledge from its own driver, it can share knowledge with all other Tesla cars.
As Fred Lambert reported on Electrik shortly afterwards, the Model S owners were aware of how fast the car's driverless function was improved. For example, this Tesla car just started off the wrong way on the highway, which also forced their owners to drive on the right road by manual driving. But in a few short weeks, the owners noticed that their cars were not going to go wrong at the time of autopilot.
The owner of a Tesla Model S car said, "I found it greatly improved this point."
Intelligent systems, such as the intelligent system of the latest generation of machine learning software, not only become more intelligent, but also become more and more intelligent on the basis of intelligence. Understanding the speed of these systems is likely to be a particularly Combat Oriented Project on the innovation of navigation technology.
Ray Kurzweil (Ray Kurzweil) generally describes the gap between human understanding of these two concepts in his so-called "intuitive linear" concept of technological innovation and the current "exponential" change rate. He published the so-called "accelerated reward law" this is a very influential papers for nearly twenty years, a speed related changes in the birth of the theory of evolution between the sharing of knowledge in each of the equipment, they carry out the speed of evolution will be progress from time to time.
Hod, a professor of mechanical engineering and data science at Columbia University, recently told me in an interview: "I think this is probably about the enhancement of the largest exponential trend of artificial intelligence evolution and evolution", Lipson.
"All index technology trends are different", "Lipson added. "But it will have the greatest potential."
According to Lipson's theory, we can call it machine teaching, which represents the improvement of the system's speed of improvement of the index system when the devices can transmit each other and acquire knowledge.
He also said, "sometimes they stop working together, for example, when a machine uses its interlinked way like a hive from another machine. But it has time to be antagonistic, just like stopping the competition from time to time between two systems of chess. "
Lipson thought that the way of developing AI is a big discovery. The partial reason is that it can bypass the demand of human to design data for training.
Data is the fuel of machine learning, but even for machine, some data are difficult to obtain. For example, some data may be risky, slow, scarce or expensive. In this situation, machines can share experiences or add or exchange data for an integrated experience for each other. Facts have proved that this is not a small influence. It is spread to all corners of the country through its own application conditions, so we say its growth is exponential. "
Lipson thought that Google's DeepMind (a project called AlphaGo Zero) has recently got a pause, which is also an excellent example of AI learning without data training. Many people are familiar with AlphaGo, it is a representative application of artificial intelligence and machine learning, after learning the game consists of millions of people go to large range training data set, it has become the world's most dangerous players. However, AlphaGo Zero can even beat AlphaGo by learning the rules of the game and playing games from time to time, and it doesn't enhance data by exercising data. Finally, in order to show you its strength, it started from zero start after eight hours of exercise, and it beat the best chess game software in the world in chess playing.
Now we can imagine that thousands or more of AlphaGo are sharing their knowledge from time to time.
It is not only a value performance in the game only, but we once thought it would have a serious impact on the speed of the performance of the enterprise's progressive equipment.
The new industrial digital twin is an example of GE technology, now only after a machine software can imitate imitate equipment will work in the future, we can see it as a self image of the machine, it can also share data and technical personnel.
For example, a steam turbine with digital twin technology can measure steam temperature, rotor speed, cold start and other data to predict defects, and warn technicians early to avoid costly maintenance after event. Digital twin technology enables them to make predictions through their own performance, but they also depend on the performance of other turbine models that have been put into operation.
As machines begin to learn autonomously in new and strong ways in their surroundings, they can accelerate their development by communicating what they learn from each other. GE steam turbines all over the world can be used in each machine
As Fred Lambert reported on Electrik shortly afterwards, the Model S owners were aware of how fast the car's driverless function was improved. For example, this Tesla car just started off the wrong way on the highway, which also forced their owners to drive on the right road by manual driving. But in a few short weeks, the owners noticed that their cars were not going to go wrong at the time of autopilot.
The owner of a Tesla Model S car said, "I found it greatly improved this point."
Intelligent systems, such as the intelligent system of the latest generation of machine learning software, not only become more intelligent, but also become more and more intelligent on the basis of intelligence. Understanding the speed of these systems is likely to be a particularly Combat Oriented Project on the innovation of navigation technology.
Ray Kurzweil (Ray Kurzweil) generally describes the gap between human understanding of these two concepts in his so-called "intuitive linear" concept of technological innovation and the current "exponential" change rate. He published the so-called "accelerated reward law" this is a very influential papers for nearly twenty years, a speed related changes in the birth of the theory of evolution between the sharing of knowledge in each of the equipment, they carry out the speed of evolution will be progress from time to time.
Hod, a professor of mechanical engineering and data science at Columbia University, recently told me in an interview: "I think this is probably about the enhancement of the largest exponential trend of artificial intelligence evolution and evolution", Lipson.
"All index technology trends are different", "Lipson added. "But it will have the greatest potential."
According to Lipson's theory, we can call it machine teaching, which represents the improvement of the system's speed of improvement of the index system when the devices can transmit each other and acquire knowledge.
He also said, "sometimes they stop working together, for example, when a machine uses its interlinked way like a hive from another machine. But it has time to be antagonistic, just like stopping the competition from time to time between two systems of chess. "
Lipson thought that the way of developing AI is a big discovery. The partial reason is that it can bypass the demand of human to design data for training.
Data is the fuel of machine learning, but even for machine, some data are difficult to obtain. For example, some data may be risky, slow, scarce or expensive. In this situation, machines can share experiences or add or exchange data for an integrated experience for each other. Facts have proved that this is not a small influence. It is spread to all corners of the country through its own application conditions, so we say its growth is exponential. "
Lipson thought that Google's DeepMind (a project called AlphaGo Zero) has recently got a pause, which is also an excellent example of AI learning without data training. Many people are familiar with AlphaGo, it is a representative application of artificial intelligence and machine learning, after learning the game consists of millions of people go to large range training data set, it has become the world's most dangerous players. However, AlphaGo Zero can even beat AlphaGo by learning the rules of the game and playing games from time to time, and it doesn't enhance data by exercising data. Finally, in order to show you its strength, it started from zero start after eight hours of exercise, and it beat the best chess game software in the world in chess playing.
Now we can imagine that thousands or more of AlphaGo are sharing their knowledge from time to time.
It is not only a value performance in the game only, but we once thought it would have a serious impact on the speed of the performance of the enterprise's progressive equipment.
The new industrial digital twin is an example of GE technology, now only after a machine software can imitate imitate equipment will work in the future, we can see it as a self image of the machine, it can also share data and technical personnel.
For example, a steam turbine with digital twin technology can measure steam temperature, rotor speed, cold start and other data to predict defects, and warn technicians early to avoid costly maintenance after event. Digital twin technology enables them to make predictions through their own performance, but they also depend on the performance of other turbine models that have been put into operation.
As machines begin to learn autonomously in new and strong ways in their surroundings, they can accelerate their development by communicating what they learn from each other. GE steam turbines all over the world can be used in each machine