Machine Learning Street Talk (MLST) avatar

Dexa/Machine Learning Street Talk (MLST)

Learn more
Published Dec 16, 2022

#88 Dr. WALID SABA - Why machines will never rule the world [UNPLUGGED]

Dr. Walid Saba delves into the intrinsic limitations of AI and the mathematical conundrums hindering the realization of artificial general intelligence, while examining the challenges AI faces in fully understanding human language and its unique intricacies.
Episode Highlights
Machine Learning Street Talk (MLST) logo

Topics covered

Popular Clips

Episode Highlights

  • Syntax & Semantics

    The discussion highlights the significant progress and remaining challenges in mastering syntax and semantics in AI language models. acknowledges the impressive advancements in large language models, particularly their ability to understand syntax by ingesting vast amounts of text. However, he emphasizes the brittleness of these models in edge cases and the exponential increase in parameters required for marginal improvements in accuracy 1. shares his skepticism about the current capabilities of these models, noting that while they have mastered syntax, they still struggle with semantics and pragmatics 2.

       

    Symbol Grounding

    Symbol grounding remains a critical challenge in AI language models. explains that symbolic systems often define concepts in a cyclical manner, lacking real-world grounding 3. He also discusses the dynamic nature of AI systems, which continuously evolve and adapt, making it difficult to model their behavior mathematically 4. This complexity underscores the need for grounded meaning and embodiment in AI to achieve true language understanding.

       

    Pragmatics

    Pragmatic understanding in language models involves context and meaning, which are crucial for accurate interpretation. emphasizes the importance of abductive reasoning, a unique human capability that involves reasoning to the best explanation based on context 5. adds that abduction is essential for generating and justifying hypotheses, highlighting its role in both scientific discovery and everyday language use 6.

       

    Language Evolution

    The evolution of human language offers insights into AI language models. discusses how humans use non-monotonic reasoning to reinterpret language based on context, a capability that AI models struggle to replicate 7. He also notes that while AI has made strides in simulating language learning, it still falls short of the innate language abilities observed in humans, particularly in children 8.

Related Episodes