"One of the most powerful things about this technology is that, like DALL-E, it does what you tell it to do," said Nate Bennett, one of the researchers in the University of Washington lab. "From a single prompt, it can generate an endless number of designs."

The Rise of OpenAI

The San Francisco company is one of the most innovative artificial intelligence laboratories. Here's an overview of recent developments.

For the generation of images, DALL E is based on artificial intelligence researchers refer to as the neural network which is a mathematical system loosely modeled off the neural network in the brain. It is exactly the same system that can recognize the commands you type through your phone, which allows autonomous vehicles to recognize (and beware of) pedestrians and translate the language through services such as Skype.

A neural network develops its skills by studying huge quantities of digital information. By identifying patterns in thousands of images of corgi for instance it is able to identify the corgi. Researchers at DALL-E developed an artificial neural network that searched for patterns while analyzing millions of digital images as well as the captions for text that described what the images depicted. This way, it became aware of the connections between the images as well as the words.

When you write a description of an image for DALL-E the neural network creates the key elements that this image might contain. One of these features could be the curvature of the ear of a teddy bear. Another example could be the line that runs along the edge of the skateboard. A another neural network -- also known as diffusion models creates the necessary pixels to create these characteristics.

The model for diffusion is developed using a set of images where imperfections - or noise - is added gradually to a photo until it turns into a vast collection of randomly pixelized pixels. While it examines these images, the model is able to operate the process reverse. If your feeds it with random images then it filters out the background noise, and then transforms the pixels into an image that is coherent.

In the University of Washington, other research labs, as well as new startups, researchers are employing similar techniques to make new proteins.

Proteins start out as chemical compounds in strings that transform into three-dimensional shapes , which define the way they function. Recently, artificial intelligence laboratories like DeepMind that are which is owned by Alphabet which is which is the parent of Google. that owns Google has demonstrated that neural networks can precisely predict the three-dimensional structure of any protein found in the body by analyzing the small molecules it is made up of and is an incredible technological breakthrough.

 
http://www.dream11today.com/a-i-turns-its-artistry-to-creating-new-human-proteins/