Will the Machine Beat Humanity? At present, Google's AI system has been able to create its own AI even better than what human beings have created. According to reports by Fudzilla and Futurism, Google Brain researchers announced in May that they could build an AI program AI Systems AutoML. They recently decided to further challenge AutoML to create AIs that are better designed than humans. Google researchers use a reinforcement learning approach to automate the design of machine learning models and use AutoML as the controller neural network ) To develop a sub-AI network for a specific task, the researchers called the sub-network AI NASNet, whose mission is to instantly identify objects such as people, cars, traffic signs, handbags, backpacks, etc. AutoML will evaluate NASNet's ability to use this information to improve its sub-AIs and repeat the process thousands of times. Conducted on two of the largest academic datasets in computer vision, the ImageNet Image Classification and COCO Target Detection datasets, When tested, NASNet performed better than all other computer vision systems. NASNet achieved an accuracy of 82 when predicting images on the ImageNet validation set. 7%, 1.2% better than previously published systems, 4% more system efficiency, and 43.1% average mum. In addition, NASNet, which requires less compute power, is also more efficient than similar mobile machines of similar size Learning model is 3.1% high Machine learning is the key capability of many AI systems to perform certain tasks. The concept behind it is simple: the algorithm learns by providing a large amount of data, but the process takes a lot of time and effort.If you can create accurate, Efficient AI systems and other process automation, such as AutoML that create AI's AI system, can do the job for humans eventually, which means that non-experts can also use machine learning and AI technology through AutoML.Currently highly accurate and efficient computer vision The algorithm is highly sought after because of the large number of potential applications.Google researchers said that computer vision algorithms can be used to create advanced robotics driven by AI or to assist visually impaired people to restore sight and to help designers improve self-driving skills. Vehicles can identify objects faster on the path, will be able to respond to them sooner, thereby enhancing the safety of self driving. Google researchers think NASNet Widely used in a variety of applications and this AI has been opened up for inference of image classification and detection of targets, the researchers wrote in a blog that they hope that the larger machine learning community can use these models as a basis for solving what many have not yet thought of A lot of computer vision problems.
