CHINA TOPIX

11/21/2024 10:07:16 pm

Make CT Your Homepage

Google Dream Project: Google Working On Developing A Computer Machine That Can Actually Dream!

Google

(Photo : Reuters) Google and Facebook have revealed a set of images that have been developed by the neural network of a computing machine, which is quite similar to the end result of a machine that is dreaming.

Google and Facebook have revealed a set of images that have been developed by the neural network of a computing machine, which is quite similar to the end result of a machine that is dreaming.

Will the system work?

According to Mirror, Google has released a set of pictures that have been created by a machine when it is dreaming. This dreaming is quite the opposite of what the human mind is capable of; however, the search engine giant has managed to empower its machines with artificial neural networks that are able to develop these images.

Like Us on Facebook

Meanwhile, Google labs continues to research the product in greater depth. The development is not sudden as Google first trains the machines by showing them millions of images. Gradually, the machines are asked to recognize images by means of analysis. This teaches them to differentiate and makes them better equipped to develop images as well. 

Google is calling the new technique as Inceptionism. The images that were unveiled by the company recently, were a result of the machines analyzing one layer of the image.

Commenting on the development, Google's Alexander Mordvintsev said that the artificial intelligence in the network also allows it to choose specific features and enhance them in images. With each layer, the features also differ, accordingly to the complexity of the images.

Engadget said that even Facebook showed off a similar neural network that was capable of developing pictures based on the understanding of what the objects looked like. While the primary algorithm uses a random vector to generate the images, the secondary adds the more realistic touch and later the fake ones are rejected by the machine. The output generated in the first round is being rated as good enough as it managed to fool close to 40 percent of human users.

With further improvements and passage of time, this feature is set to improve and become better. When compared to what Google has done, Facebook is using a completely opposite technique. Though there's still a lot of improvement to come in this concept, the initial results are certainly impressive.

Real Time Analytics