Google has evolved an artificial intelligence system that has developed its own “Child”
There is more, the original AI has trained its creation to such advanced level that it can outperform any AI system built by humans.
Google Artificial Intelligence Creates Its Own “AI CHILD”[irp posts=”1839″ name=”Artificial Intelligence & Augmented Reality (AR) Vs Virtual Reality (VR) – Which is Best for Future”]
Google revealed its AutoML project back in May, with the motive of easing design machine learning by making the process automatic.
At that time, the company said that according to their approach, a controller neutral net can propose a ‘child’ model architecture, which can be then trained and evaluated for a better quality on a specific task.
Last month, the AutoML plans were used to generate NASNet, a “child” AL developed for object detection, which turned out to be better in performance than any other AL machines created by humans.
Just to test it, it was applied to two of the most respected and huge scale academic databases ever made, ImageNet image classification and COCO object detection dataset.[irp posts=”2211″ name=”WhatsApp Empowers Group Admins With New Dubbed Restricted Groups Feature”]
After the tests, the results were calculated and you won’t believe them.
“Child” got a prediction accuracy of 82.7% on ImageNet, 1.2 percent better than all previous results ever recorded.
According to Google, NASNet acquired 43.1% mAP on COCO, better than the last, published state-of-the-art.
The results clearly show how the Google AI own “Child” outperformed any other AI system ever made.
The “Child” of Google AI, NASNet learns through ‘reinforcement learning’ and it reports back to its “Parent” and learns how it can perform better. The AI’s “child” is advanced enough to recognize people, cars, handbags etc.[irp posts=”2205″ name=”Google’s Best Android Applications, Games & Books Of 2017″]
Anyhow, such creations bring ethical problems with them and many people have already warned of the perils of this super-smart AI.
It was only last month when Tesla CEO, Elon Mush reiterated his fears on the rise of machines, where responding to footage a gigantic robot doing back-flips he told how the development of AI needs to be controlled.
He also said that the regulations will stop humanity from being outsmarted by these AIs.
However, this Google AI “child” has proved to be of advanced level. Let’s see what Google has got for us now.