Super Data Science: ML & AI Podcast with Jon Krohn 972: In Case You Missed It in February 2026
45 snips
Mar 6, 2026 
Praveen Murugesan, VP of Engineering at Samsara, on quantum computing for routing and more autonomous ops. Antje Barth, Amazon technical staff on developer-facing AI tooling and UI automation workflows. Tom Griffiths, cognitive scientist and author exploring human cognition shaping AI design. Will Falcon, CEO of Lightning AI and PyTorch Lightning creator on turning open source into a revenue-generating company.
AI Snips
Chapters
Transcript
Episode notes
Open Source Project Forced An Unexpected Startup
- Will Falcon almost didn't found a company and considered shutting down PyTorch Lightning because it was distracting from his research.
- VCs sent 10 term sheets within days after the project took off, forcing him to choose startup life over returning fully to academia.
Math Meets Engineering Powers Pre training Breakthroughs
- Will Falcon contrasts theoretical training-algorithm research with engineering tradeoffs needed for large-scale pre-training.
- FAIR combined deep math (loss functions, regularizers) with engineering (distributed strategies, fault tolerance) to advance pre-training approaches.
Human Intelligence Is Adaptation Under Tight Limits
- Tom Griffiths frames human intelligence as rational adaptation under tight limits like lifespan, compute, and communication bandwidth.
- Because humans face those constraints we learn from few examples and use conventions (language, science) to pool knowledge, unlike unconstrained AI.