Chesley Bonestell, “The Exploration of Mars” (1953), oil on board, 143/8 x 28 inches, gift of William Estler, Smithsonian National Air and Space Museum. Reproduced courtesy of Bonestell LLC.

The End of Insight?

Ed Finn launches the Spring 2017 “Issues in Science and Technology”

Researchers across a variety of fields use algorithms to investigate a wide range of questions and problems, from modeling genetic expression to understanding weather patterns. Sometimes the algorithms produce insights that are verifiable, but that humans cannot fully understand. What are we to make of these algorithmic insights? Can a viable research enterprise be built on calculations we don’t comprehend, identifying causalities we can’t otherwise see? What if our black-box machines are wrong in some critical but unknowable way?

Artificial intelligence, deep learning, and other bleeding-edge technologies seem to offer the possibility of achieving the very old—and very human—dream of unlocking the mysteries of the universe through rational calculation. But the complexities of our algorithms and models threaten to exceed our grasp. Their computational sophistication may already be undermining our capacity for insight, and with it, our ability to act knowledgeably in the world.

Drawing from his essay in the Spring 2017 Issues in Science and Technology and from his new book, What Algorithms Want: Imagination in the Age of Computing, Ed Finn will explore how computer-generated knowledge is affecting scientific research, and how humans can become better users and architects of these powerful algorithms.