Researchers Enable AI To Use Its Imagination Closer to Humans’ Understanding of the World – California News Times

The new AI system is inspired by humans. When humans see the color of one object, they can easily apply it to other objects by replacing the original color with a new color. Credit: Chris Kim

A team of USC researchers is helping AI imagine what is invisible. This approach could also lead to fairer AI, new drugs, and improved safety for self-driving cars.

Imagine an orange cat. Now imagine the same cat. However, it has black fur. Now, imagine a cat pretending to be along the Great Wall of China. When you do this, the activation of a series of neurons in the brain brings to mind variations of the images presented, based on previous world knowledge.

In other words, as a human, its easy to imagine objects with different attributes. However, despite advances in deep neural networks that rival or exceed human performance on certain tasks, computers still struggle with the very human skill of imagination.

Currently, a USC research team consisting of Professor Laurent Itti of Computer Science and doctoral students Yunhao Ge, Sami Abu-El-Haija, and Gan Xin has used human-like features to date. We have developed an AI that imagines no different objects. attribute.Title treatise Zero-shot synthesis by learning with group supervised learningWas published in the 2021 International Conference on Learning Representation, held on May 7th.

We were inspired by human visual generalization, which attempts to simulate human imagination with machines, said Ge, the lead author of the study.

Human can separate learned knowledge by attributes (shape, pose, position, color, etc.) and recombine them to imagine new objects. Our paper uses neural networks. Im trying to simulate this process.

For example, suppose you want to create an AI system that produces an image of a car. Ideally, the algorithm would provide some images of the car, allowing it to generate different types of cars of any color from different angles, from Porsche to Pontiac to pickup trucks.

This is one of AIs long-standing goals of creating extrapolable models. This means that, given some examples, the model should be able to extract the underlying rules and apply them to a wide range of new examples that we have never seen before. However, machines are most commonly trained on sample features, such as pixels, without considering the attributes of the object.

In this new study, researchers are trying to overcome this limitation by using a concept called unraveling. Unraveling can be used to generate deepfake, for example, by unraveling human facial movements and identities. By doing this, people can synthesize new images and videos that replace the identity of the original person with another person, but maintain the original movement, Ge said.

Similarly, the new approach takes a group of sample images instead of one sample at a time as in traditional algorithms, mines the similarities between them, and controllably entangled expression learning. To realize what is called.

We then recombine this knowledge to achieve what we call new controllable image composition, or imagination. For example, in the Transformers movie, you can take the shape of a Megatron car, the color and pose of a yellow Bumblebee car, and the background of Times Square in New York. As a result, this sample during a training session. Even if he wasnt witnessed, a Bumblebee-colored Megatron car would be driven in Times Square.

This is similar to how humans extrapolate. When humans see the color of one object, they can easily apply it to other objects by replacing the original color with a new color. Using their technology, the group has generated a new dataset containing 1.56 million images that may be useful for future research in this area.

Unraveling is not a new idea, but researchers say their framework is compatible with almost all types of data and knowledge. This opens up new application opportunities. For example, create a fairer AI by unraveling race- and gender-related knowledge and completely removing sensitive attributes from equations.

In the field of medicine, unraveling the functions of medicine from other properties and recombining them to synthesize new medicine may help doctors and biologists discover more useful drugs. You can also create a safer AI by infusing your machine with imagination, for example, by allowing self-driving cars to imagine and avoid dangerous scenarios never seen before during training.

Deep learning has already demonstrated outstanding performance and expectations in many areas, but often this is done by shallow imitation and does not have a deep understanding of the individual attributes that make each object unique. Said Itti. For the first time, this new unraveling approach unleashes the new imagination of AI systems and brings it closer to understanding the human world.

Reference: May 7, 2021, Zero Shot Synthesis by Group Supervised Learning by Yunhao Ge, Sami Abu-El-Haija, Gan Xin, Laurent Itti 2021 International Conference on Learning Representation..Link

Researchers Enable AI To Use Its Imagination Closer to Humans Understanding of the World Source link Researchers Enable AI To Use Its Imagination Closer to Humans Understanding of the World

See the original post:

Researchers Enable AI To Use Its Imagination Closer to Humans' Understanding of the World - California News Times