AI’s Self-Portrait

IBM Research asked AI to create a self-portrait. And the result (shown here) was published in the New York Times on October 19. Creating the image required a lengthy training program. First, the system identified a core visual concept of AI by ingesting over 3,000 articles about AI from the NYT, which it narrowed down to a human and robot shaking hands. Then, it was trained on a data set of over 1,000 images of robot and human hands in order to draw a new image. And, finally, it was trained with samples of NYT cover art so that the final creation would fit the publishers’ visual style. The resulting original art piece portrays AI and human creativity working together, hand in hand.  

Learn more about the process behind the AI self-portrait ->


via IBMblr = IBM Innovation Culture + Tumblr

October 25, 2018 at 07:39PM