-
The “Dreams” of Google’s Artificial Intelligence
Have you ever wondered if computers could dream?
Surprisingly, yes, they can.
Google’s innovative Deepdream software is turning AI neural networks inside out to apprehend how computers think.
After a group of artificial brains at Google began producing surreal images from the otherwise standard photographs, the engineers contrasted what they saw to dreamscapes. Their image-generation method was called “Inceptionism”, and the code it powered was termed “DeepDream.”
Wikipedia defines DeepDream to be a computer vision program created by the Google engineer Alexander Mordvintsev, that uses a convolutional neural network to find and enhance patterns in images via algorithmic pareidolia, thus creating a dream-like hallucinogenic appearance within the deliberately over-processed images.
These computer-generated images feature color scrolls, spinning shapes, stretched faces, swirling eyeballs, and awkward patterns of shadow and light. The computer seems to be hallucinating – in an astonishingly human manner.
The project aimed to see how well a neural network identifies varied animals and environments, with the machine explaining what is observed. The results revealed were fascinating. It told a lot about where AI is headed and why it could be more imaginative, ambiguous, and unpredictable than we liked.
The Google artificial neural network is modeled after the central nervous system of animals. It functions similarly to a computer brain. When engineers feed an image to the network, it is examined by the first layer of ‘neurons’. It then communicates with the next layer, which attempts to represent the image. This process is continued for 10 to 30 rounds, with each layer defining and alienating main elements until the picture is deduced. The neural network then informs us about the entity that it has valiantly attempted to identify, usually with little progress. This is the method it uses to recognize images.
Later the Google team tried to reverse the procedure. They hoped to learn more about the qualified features the networks knew and did not know. They gave complete freedom to the network and asked it to interpret and improve an input picture in such a way so as to elicit a specific interpretation.
Next, the researchers discovered something astounding. These neural networks could not only distinguish between a variety of images, but they also had sufficient knowledge to produce images culminating in these unpredicted computational representations.
According to IFL Science, computers have the potential to view images in objects in a way that only an artist can dream of replicating. It can see buildings within clouds, temples in trees, and birds in leaves. Highly detailed elements seem to arise out of nowhere.
This processed image of a cloudy sky proves that Google’s artificial neural network is the champion in finding pictures in a cloudy sky.
This technique of creating images where there are none is called “Inceptionism.” The designers gave the computer free reign over its artwork. The final product was beautiful pictures that were derived from a mechanical mind. The engineers referred to these as “dreams”.
Source:
https://www.analyticsinsight.net/heres-what-the-dreams-of-googles-artificial-intelligence-look-like/ -
0 Replies
Sorry, there were no replies found.
Log in to reply.