Google AI creates record breaking child A.I

in #ai7 years ago

AutoML is a Google project designed to make the process of designing machine learning models much more accessible. 

Google have developed an A.I that has the capability of creating it's own child A.I, called NASNet. Not only is this incredible in terms of machine learning and automation, but the child A.I outperformed every other human-built AI systems like it. To test the new A.I, Google applied it to the ImageNet image classification and COCO object detection dataset. These are described as two of the most respected large scale academic datasets in computer vision. NASNet achieved a prediction accuracy of 82.7%, which is a 1.2% increase of all previously published results. On Coco, NASNet achieved 43.1% mAP, a 4% increase to any published result. Read more on Google's research blog.

Some readers may be having visions of Skynet right now, and I wouldn't blame you. But on the flip side, this is an amazing step towards the future of A.I and machine learning! 

Sort: