AI Google created its own AI, surpassing all analogs

In the spring of this year, Google Brain engineers introduced the artificial intelligence AutoML, capable of creating their own unique AI without the participation of people. Not so long ago it became known that AutoML first created a computer vision system NASNet, significantly superior to all analogs created by man. This system, based on AI, can become a serious help in the development, for example, of autonomous vehicles, as well as in robotics, allowing to bring the vision of robots to a whole new level.

AutoML develops on the so-called training system with reinforcement. In fact, it is a managing neural network, which independently develops completely new neural networks for any specialized tasks. In this case, the main goal of AutoML was to create a system for the most accurate recognition of objects on video in real time. AI independently taught a new neural network, monitoring its mistakes and adjusting its work. The learning process was repeated many thousands of times until the system became operational. Moreover, it surpassed all existing to date similar neural networks, created and trained by people.

According to Google's official statement, the accuracy of NASNet recognition is 82.7%. This is 1.2% better than the previous record set in September this year by specialists from Oxford and Momenta. The neural network also proved to be 4% more efficient than analogues with 43.1% of average accuracy. The simplified version of NASNet, adapted for mobile platforms, surpasses such neural networks by more than 3%. This system in the future can be used to create autonomous cars, because for them, computer vision is incredibly important. Meanwhile, AutoML continues to create new neural networks, and who knows what heights he will be able to achieve in the near future.

hi-news.ru

© All rights reserved. Citing to www.ict.az is necessary upon using news