Google AI Created It’s Own Child AI, That Is Superior Than Man Made Models

If I want to bet over a company which will have an absolute dominance in AI, then with no doubt it will be Google. Whether it is offering collections of code that best the world’s board game champions or the new Google Duplex which is a completely automated system that places calls like humans.

Now researchers at Google Brain has announced an artificial intelligence (AI) system that has created its own “child”. The creation called  AutoML, an AI that’s capable of generating its own AI. The machine is named as NASNet.

NASNet is capable of outperforming all of its human-made counterparts. Its creations have gone from analyzing words to disseminating complex imagery and it can recognize objects, such as people, cars, handbags and traffic lights, in real-time in a video.

The machine learns to do so through reinforcement learning. Here AutoML acts as a controller neural network which we can perhaps call the “parent” that develops a child AI network for a specific task. A particular task is given to the child and the feedback given to the parent.

This feedback is then used to tell the controller how to improve its proposals for the next round. The process is then repeated thousands of times, generating new architectures, testing them, and giving that feedback to the controller to learn from.

Google designed NASNet, as a system can do object detection. After creation, it later tested on the ImageNet image classification and COCO object detection data sets, which the Google researchers call “two of the most respected large-scale academic data sets in computer vision”.

NASNet outperformed all other computer vision systems in this test. Eventually, NASNet was 82.7 percent accurate at predicting images correctly. This is 1.2 percent better than any previously published results, and the system is also 4 percent more efficient, with a 43.1 percent mean Average Precision (mAP). Additionally, a less computationally demanding version of NASNet outperformed the best similarly sized models for mobile platforms by 3.1 percent.

The Google researchers acknowledge that NASNet could prove useful for a lot of applications and have open-sourced the AI for inference on image classification and object detection.

“We hope that the larger machine learning community will be able to build on these models to discuss multitudes of computer vision problems we have not yet imagined,” said the researchers, who have open-sourced NASNet so it could be used for computer vision applications.”

While there are many possible uses of AutoML and NASNet, there are also ethical issues related to AI. For instance, what if AutoML creates AI systems at such a speed that the society simply cannot keep up with them or what if the AI parent passes down unwanted biases to its child.

To keep all these things in human control, it is very important to carry out more strict regulations and enhanced ethical standards to prevent the use of AI for malicious purposes.

More in AI :

Artificial Intelligence (AI) Helps Stroke Patients To Walk Again

New AI Detects Guns, Identifies Shooters And Alert Law Enforcement

This New Artificial Intelligence System Can Spot Fake Online Reviews

This New AI Jellyfish Robots Will Help Us To Find Untold Secret Of Ocean

Meet “Will” World’s First AI Digital Teacher, Teaching Kids About Energy