Information Technology University of Copenhagen professor Joel Lehman and colleagues have created an artificial intelligence (AI) that can design and generate physical three-dimensional (3-D) objects. Lehman's team has leveraged the image-recognition power of deep neural networks (DNN) to create new artifacts without human input.
Working with the University of Wyoming's Evolving Artificial Intelligence Lab, the team has designed an AI system that uses an evolutionary algorithm to generate a random blueprint. The algorithm models a 3-D image that resembles a misshapen blob of clay and passes a few snapshots on to the DNN to solicit feedback. The DNN compares the snapshots to the images in its database, decides if the object resembles anything it is familiar with, and gives the algorithm feedback. The evolving algorithm then takes the blueprint, mutates it a little and sends it back for more feedback. If the DNN thinks the blob looks worse, it is discarded and the algorithm restarts the process. If the feedback improves, the new version becomes the basis for further mutations.
The algorithm and DNN go back and forth in this fashion millions of times to sculpt a recognizable object.
From Popular Science
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA
No entries found