From Planes to Brains: Building AI the Wright Way
What can the invention of a flying machine teach us about how to create a thinking machine? After giving some historical background, we will highlight similarities and differences between the Wright process and how we work at Vicarious. We will discuss how understanding the learning and computational principles employed by the brain is an essential step towards creating AI, just as understanding the aerodynamics of bird-gliding was essential for building a flying machine.
Our approach considers the structure of the brain, world’s data, and a mathematical framework at the same time. For a machine or algorithm to excel at a task, it must make assumptions. We look at the structure of the brain to decipher the assumptions that nature found to be effective through millions of years of evolution. We correlate the properties of the brain to world’s data to tease apart computational principles from the details of biological implementation. We use the mathematical framework to make our assumptions explicit and analyzable. In contrast to simulation approaches that try to imitate the brain in increasing biological detail and scale, our process results in very efficient models that are far easier to build, test and iterate on. We will show some exciting results of applying this process, and describe how we anticipate the future to unfold.