The MAD Podcast with Matt Turck Jeremy Howard on Building 5,000 AI Products with 14 People (Answer AI Deep-Dive)
153 snips
May 15, 2025 
In this discussion, Jeremy Howard, an AI scientist and co-founder of Answer AI and Fast.ai, reveals his journey from philosophy to revolutionizing AI product development with a tiny team. He shares how Answer AI is redefining R&D by focusing on real-world applications instead of chasing AGI. Topics include the rise of open-source models, the efficiency wars in AI, and innovative tools like Solve It and Shell Sage that empower users. Jeremy emphasizes the need for small, mission-driven teams to prioritize societal benefits over profits.
AI Snips
Chapters
Books
Transcript
Episode notes
Test-Time Compute Advantages
- Test-time compute optimization is a key, previously overlooked, method to improve AI performance.
- Enhancing efficiency reduces costs and increases model performance, benefiting more users.
Perspective on AGI and Progress
- AI progress is continuous without clear evidence that artificial superintelligence is imminent.
- The shift to natural language interfaces changes perception but not underlying AI capabilities.
From Philosophy to AI Expert
- Jeremy Howard excelled at math late by mastering exam rubrics rather than early deep study.
- His interest in computer games propelled self-taught expertise leading to exceptional predictive modeling skills.
