In this podcast, Gary Marcus, author of the book, Rebooting AI: Building Artificial Intelligence We Can Trust, talks about his extensive background in science, studying the possibilities and pitfalls of artificial intelligence.
Marcus has a long history with science, starting early as a kid who was fascinated by all things scientific. His work today often focuses on natural and artificial intelligence. Marcus was Founder and CEO of Geometric Intelligence, a machine learning company acquired by Uber.
Marcus discusses his extensive background in science and his keen interest in human cognition and AI. He explains the variations and differences in intelligence, from the ability to accomplish one or a few tasks to general intelligence, which seeks to create machines that can not only handle tasks but begin to figure things out on their own, without further input. The scientist discusses human intelligence and innate abilities. Additionally, Marcus provides his thoughts on self-driving software and how it simply has not been debugged. He cites examples of self-driving vehicle accidents in which the software failed to make critical decisions that led to human injury and deaths.
Expanding on his thoughts further, Marcus describes how some essential information will be ‘learned’ while others must be built-in. And Marcus states that AI needs to have a ‘rough draft’ of the psychology built-in so it can function in the world in which it exists.