Artificial intelligence and Machine Learning are the two hottest terms in many industries in the recent years. The most recent technological advancement in this domain is the development of AI which in turn creates an AI better than any human-made AI.
In January 2017 Google Brain started developing AI software that can build more AIs. Later in May this year, the researchers developed AutoML, a machine learning algorithm that is capable of creating its own AIs eventually reducing the dependency on humans. More recently, they decided to throw AutoML with its biggest challenge to date, i.e to create a “child” that outperformed all its human-made counterparts.
Google Brain’s team used an approach called reinforcement learning to automate the design of machine learning models. AutoML acts as a controller neural network that can create a “child” network to execute a specific task.
For this particular child AI, which the researchers called NASNet, it’s task was to recognize objects in a real-time video feed, like people, cars, traffic lights, handbags, or backpacks. Then its performance will be evaluated by AutoML’s controller neural net and then re-trains the child by giving the feedback until it enhances its performance and gets to a superior version of NASNet.
After going through the improvising process endlessly, the NASNet was then subjected to test on the ImageNet image classification and COCO object detection datasets – both of which known as“ the most respected large-scale academic data sets in computer vision”. And according to the results, NASNet outperformed all other computer vision systems.
NASNet was able to predict 82.7% of the images it was shown on ImageNet’s validation set which is 1.2% higher than the previously published results, according to the researchers. Also, the system was 4 percent more efficient, with a 43.1 percent mean Average Precision (mAP). Additionally, a less computationally demanding version of NASNet outperformed mobile platforms by 3.1%.
The Google researchers also acknowledged that the improvised version of NASNet could be used for many computer vision applications. Also, they have open-sourced NASNet for inference on image classification and for object detection in the Slim and Object Detection TensorFlow repositories.
“We hope that the larger machine learning community will be able to build on these models to address multitudes of computer vision problems we have not yet imagined,” the researchers wrote in their blog post.
No matter how useful the NASNet and AutoML maybe, we do not know if the society can keep up with the systems that AutoML creates. To keep thing under control, it is very important to implement strict regulations and enhanced ethical standards to prevent the use of AI for malicious purposes. And various governments and companies are focusing on creating moral and ethical implications of AI.
