Google Brain engineers in the spring of this year demonstrated Automl. This artificial intelligence can without the participation of a person to produce his own unique AI. As it turned out recently, Automl was able to create NASNET for the first time, a computer vision system. This technology seriously exceeds all previously created analogues. This system based on artificial intelligence can be an excellent assistant in the development of, say, autonomous cars. It is applicable in robotics – robots will be able to reach a completely new level.
Automl development takes place through a unique training system with reinforcement. We are talking about a manager-manager who independently develops completely new neural networks intended for certain specific tasks. In the case we indicated, Automl is aimed at producing a system that accurately recognizes in real time objects in the video.
Artificial intelligence itself was able to train a new neural network, following errors and adjusting the work. The training process was repeated repeatedly (thousands of times), until the system was suitable for work. It is curious that she was able to bypass any similar neural networks that are currently available, but developed and trained by man.
Google officially stated that NASNET recognizes 82.7%. The result is 1.2 % exceeds the past record, which was set by researchers from Momenta and specialists of Oxford in the early fall of this year. NASNET is 4% more effective than their analogues with an average accuracy of 43.1%.
There is a simplified version of NASNET, which is adapted to mobile platforms. It surpasses analogs a little more than three percent. In the near future it will be possible to use this system for the production of autonomous cars, for which it is important to have computer vision. Automl continues to produce new hereditary neural networks, striving to conquer even greater heights.
Source: Sciencealert.com
Google Brain engineers in the spring of this year demonstrated Automl. This artificial intelligence can without the participation of a person to produce his own unique AI. As it turned out recently, Automl was able to create NASNET for the first time, a computer vision system. This technology seriously exceeds all previously created analogues. This system based on artificial intelligence can be an excellent assistant in the development of, say, autonomous cars. It is applicable in robotics – robots will be able to reach a completely new level.
Automl development takes place through a unique training system with reinforcement. We are talking about a manager-manager who independently develops completely new neural networks intended for certain specific tasks. In the case we indicated, Automl is aimed at producing a system that accurately recognizes in real time objects in the video.
Artificial intelligence itself was able to train a new neural network, following errors and adjusting the work. The training process was repeated repeatedly (thousands of times), until the system was suitable for work. It is curious that she was able to bypass any similar neural networks that are currently available, but developed and trained by man.
Google officially stated that NASNET recognizes 82.7%. The result is 1.2 % exceeds the past record, which was set by researchers from Momenta and specialists of Oxford in the early fall of this year. NASNET is 4% more effective than their analogues with an average accuracy of 43.1%.
There is a simplified version of NASNET, which is adapted to mobile platforms. It surpasses analogs a little more than three percent. In the near future it will be possible to use this system for the production of autonomous cars, for which it is important to have computer vision. Automl continues to produce new hereditary neural networks, striving to conquer even greater heights.
Source: Sciencealert.com