Super Data Science: ML & AI Podcast with Jon Krohn 907: Neuroscience, AI and the Limitations of LLMs, with Dr. Zohar Bronfman
101 snips
Jul 22, 2025 Dr. Zohar Bronfman, Co-founder and CEO of Pecan AI, holds dual PhDs in computational neuroscience and philosophy. In an engaging chat, he argues that LLMs fall short of achieving true AGI, highlighting the importance of understanding decision-making through a neuroscientific lens. Bronfman shares insights on why predictive models are superior for businesses over generative ones and discusses the philosophical nuances of consciousness that machines can't grasp. He also touches on animal intelligence and the creative divide between humans and AI.
AI Snips
Chapters
Transcript
Episode notes
LLMs Lack True General Intelligence
- LLMs lack a unified value system and do not possess genuine consciousness or goals. - This limits their ability for true general intelligence and transfer learning across domains.
Bumblebees Demonstrate Transfer Learning
- Bumblebees can cross-modally transfer knowledge by recognizing objects by sight that they only touched before. - This ability shows even small brains mastered forms of associative learning AI still cannot replicate.
Generative vs. Transformational Creativity
- Generative AI exhibits combinatorial creativity by combining existing ideas. - It lacks transformational creativity, which involves forming fundamentally new perspectives or paradigms.