Larry Birnbaum, Professor of Computer Science at Northwestern University and co-founder of Narrative Science, joins the AI Lab to explain why today’s large language models work at all—and why their success has surprised even longtime AI researchers.

In this conversation, Larry and Ryan explore how decades of AI research led to the current moment, why massive models somehow generalize instead of just memorizing, and what this means for businesses adopting AI at scale.

We also dive into:

• Why today’s AI boom feels sudden. but has decades of groundwork behind it

• The overfitting paradox: why massive models generalize instead of just memorizing

• How transformers and attention changed context, reference, and meaning in languag

• Why AI delivers the most value at scale and through personalization (work that wouldn’t

otherwise exist)

• Bias vs. inductive bias and how agreeable AI systems can reflect a user’s worldview back at them

• What AI breaks in education, and why fundamentals still matter for students and engineers

• The interpretability problem: we know how LLMs learn, but not what they’ve learned or how they decide

Larry shares hard-earned perspective from a career spanning academic AI research and real-world deployment, including lessons from building one of the earliest automated narrative platforms and why understanding AI’s limits is just as important as celebrating its breakthroughs.

Guest Bio:

Larry Birnbaum is a Professor of Computer Science at Northwestern University whose research focuses on artificial intelligence, natural language processing, and human–AI collaboration. He is the co-founder of Narrative Science, an early pioneer in automated storytelling from data, and has spent decades studying how machines can generate language that is coherent, contextual, and useful at scale.

Resources

→ Connect with Larry Birnbaum on LinkedIn: https://www.linkedin.com/in/larry-birnbaum-1393b4147/