Finding Genius Podcast AI Advancing – Gary Marcus, Scientist, Entrepreneur, and Author – Discussing the Current State, and Future, of AI and Learning
Gary Marcus, the scientist, entrepreneur, and author of the buzz-worthy book, Rebooting AI: Building Artificial Intelligence We Can Trust, talks about the current state of AI and learning.
Marcus is a scientist, first and foremost, and his extensive research has often focused on natural and artificial intelligence. Marcus was Founder and CEO of Geometric Intelligence, a successful machine learning company that was acquired by Uber. Marcus talks about his background, and his early interest in AI, and the possibilities to, as he states, "make machines do smart things."
Marcus discusses his path from studying human cognition to AI. He states that much 'intelligence' is narrow, meaning the intelligence can only accomplish one or a few tasks. With general intelligence, the goal is to build machines that can figure things out for themselves, to analyze and adapt. The scientist and educator talk about human intelligence and our abilities.
Continuing, Marcus discusses early chatbots and their relative intelligence levels. Additionally, he talks about self-driving software and some of the accidents that have caused injuries and even death. He states that none of this software has been properly debugged, but many people feel comfortable already. Marcus expounds upon the concept of deep comprehension, discussing data representations, and the ability to analyze things like space, time, causality, and individual objects. He explains his thoughts on how some things will be 'learned' but others will be built in. He talks about how AI needs to have a 'rough draft' of the psychology built-in, information and knowledge it needs to possess in order to function in the world around it, and that will ultimately make the learning process easier.
Marcus discusses some of the problems and challenges that arise in the quest to advance AI and neuroscience. He explains his theories and expands on some of the themes in his new book. He discusses which aspects of problems can be readily solved and how machines often fail in analytical thinking unless proper programming is implemented.
Gary Marcus, the scientist, entrepreneur, and author of the buzz-worthy book, Rebooting AI: Building Artificial Intelligence We Can Trust, talks about the current state of AI and learning.
Marcus is a scientist, first and foremost, and his extensive research has often focused on natural and artificial intelligence. Marcus was Founder and CEO of Geometric Intelligence, a successful machine learning company that was acquired by Uber. Marcus talks about his background, and his early interest in AI, and the possibilities to, as he states, "make machines do smart things."
Marcus discusses his path from studying human cognition to AI. He states that much 'intelligence' is narrow, meaning the intelligence can only accomplish one or a few tasks. With general intelligence, the goal is to build machines that can figure things out for themselves, to analyze and adapt. The scientist and educator talk about human intelligence and our abilities.
Continuing, Marcus discusses early chatbots and their relative intelligence levels. Additionally, he talks about self-driving software and some of the accidents that have caused injuries and even death. He states that none of this software has been properly debugged, but many people feel comfortable already. Marcus expounds upon the concept of deep comprehension, discussing data representations, and the ability to analyze things like space, time, causality, and individual objects. He explains his thoughts on how some things will be 'learned' but others will be built in. He talks about how AI needs to have a 'rough draft' of the psychology built-in, information and knowledge it needs to possess in order to function in the world around it, and that will ultimately make the learning process easier.
Marcus discusses some of the problems and challenges that arise in the quest to advance AI and neuroscience. He explains his theories and expands on some of the themes in his new book. He discusses which aspects of problems can be readily solved and how machines often fail in analytical thinking unless proper programming is implemented.