Microsoft Research has launched a new podcast series examining whether today's AI systems possess genuine intelligence, bringing together experts in digital and biological architectures to debate the fundamental question.

The inaugural episode of "The Shape of Things to Come" features Doug Burger, a Microsoft Research leader, alongside Nicolò Fusi from Microsoft Research and Subutai Ahmad from Numenta. The trio compares transformer-based large language models with the human brain's distributed, continuously learning architecture.

Fusi, who specializes in digital transformer architectures and information theory, brings expertise in Bayesian nonparametrics and computational biology to the discussion. Ahmad approaches the question from a biological perspective, focusing on how the human brain processes information.

The conversation explores critical differences between current AI systems and human cognition, including efficiency, representation methods, and sensory-motor grounding. The researchers examine where today's models excel and where they fall short compared to biological intelligence.

Key architectural differences

The discussion probes whether digital intelligence and biological intelligence are fundamentally different systems optimized for different tasks, or whether AI models are on a trajectory to surpass human capabilities.

The researchers analyze the human brain's ability to learn continuously while maintaining previous knowledge, contrasting this with current AI training methods that typically require complete retraining on new datasets.

Efficiency emerges as another crucial factor, with the human brain operating on roughly 20 watts of power while performing complex reasoning tasks that require massive computational resources in current AI systems.

The podcast series aims to help technologists, policymakers, and business leaders better understand AI's trajectory and implications. Future episodes will likely explore additional aspects of AI development and its societal impact.

Burger emphasized that the goal is "to amplify the shared understanding needed to build a future in which the AI transition is a net positive." The series represents Microsoft Research's effort to address fundamental questions about AI capabilities and limitations as the technology rapidly advances.