Google made AI Creates AI Far Superior To Human Made
Talk about technology taking on humans. Artificial intelligence is definitely taking over the world, but right now it seems that is overpowering and doing better than the world.
Google, the forerunner of Artificial Intelligence had reported in January of 2017 that Google’s Artificial Intelligence research lab, the Google Brain was working on an Artificial Intelligence software that would be capable of creating and managing other AI programs.
And Google was successful with the project, and the project was released in May of this year. Termed as the AutoML, this software was created with the purpose of eliminating a team of researchers and engineers to come up with a working Artificial Intelligence software. But, how did the software turn out to work?
The Google Brain team challenged the software to create another AI software, a “child”, and the software came out with a program that outdid all other software that was created by humans. The software acted as a controller neural network and thus could create a “child” network to execute and work at a specific task.
The Artificial Intelligence software, otherwise the “child”, was also tested. The “child” was given a task that involved the software to recognize objects in a real-time video feed. The software had to spot out and recognize people, cars, traffic lights, handbags and even backpacks.
But this does not mean that the child has all powers to work and function without any command from the parent software. The child sends the results to the AutoML’s controller neural net, and from there it gets evaluated. The AutoML also goes through the results to enhance the child’s model until is a far more superior model than it is presently.
Superiority
The software was also compared with third-party services and it came out as an Artificial Intelligence software that outdid all other computer vision systems. When compared, the results showed that the software hit an accuracy of 82.7%, and this number was 1.2% higher than all previously published results.
This software, or the “child” software’s capabilities of image classification and object detection is something that the Google Brain Team found to be very practical and useful that they decided to open source the software.
Dangers
It is hoped that the software would be built upon these models and would thus be able to help and solve the various computer vision problems that yet remain unsolved. Though there are many benefits of this arrangement, there are many worries out there too.
Many are concerned about the capability of the software to multiply and create new services. If the speed of creating new services surpasses the number of services that society creates, then that would be a problem. Stricter regulations need to be put in place to prevent something like this from happening, and to prevent the use of Artificial Intelligence for illegal or malicious purposes.
