Learn more
Join Dexa
Published Aug 14, 2023

Prof. Jürgen Schmidhuber - FATHER OF AI ON ITS DANGERS

AI pioneer Jürgen Schmidhuber delves into the evolution and transformative power of machine learning, stressing key innovations, the promise of recursive self-improvement, and the pivotal role of open source in democratizing AI's future while cautioning against restrictive legislation that could hinder progress.
Episode Highlights
Machine Learning Street Talk (MLST) logo

Topics covered

Popular Clips

Episode Highlights

  • AI Landscape

    shares his thoughts on the current AI landscape, particularly the capabilities of models like GPT-4. He acknowledges the impressive outcomes but emphasizes that these models are not yet close to achieving artificial general intelligence (AGI) 1. Schmidhuber also highlights the rapid advancements in open-source AI, questioning the necessity of companies like OpenAI to lead the field 1.

       

    Self-Improvement

    Schmidhuber delves into the concept of recursive self-improvement, which has been a driving force in his research since the 1980s. He explains how AI systems can learn to improve their own learning algorithms, a process he calls meta-learning 2. This approach aims to create AI that can continually enhance its capabilities, constrained only by the limits of physics and computability 3.

       

    Artificial Curiosity

    Artificial curiosity is another key area of Schmidhuber's work. He describes how AI systems can set their own goals and generate actions that surprise their predictive models, thereby driving learning and innovation 4. This concept, which dates back to his work in the 1990s, has evolved into what we now know as generative adversarial networks (GANs) 4.