INSEAD Knowledge Podcast Can AI Help Build Smarter Humans?
Sep 30, 2025
Pattie Maes, a renowned computer scientist and MIT Media Lab professor, discusses her pioneering work in AI, emphasizing its role in augmenting human intelligence. She contrasts early recommendation systems with today's large language models, highlighting the importance of trust over mere competence in AI. Pattie raises concerns about AI's biases and cultural limitations, advocating for transparent, multicultural AI models. She envisions personalized AI tutors tailored to individual learning, while warning of potential risks to social connections due to constant AI interaction.
AI Snips
Chapters
Transcript
Episode notes
Early Recommender System Origin Story
- Pattie Maes built Firefly in 1994 to recommend media by matching people's tastes with similar users.
- The company later sold to Microsoft in 1998 after early contracts with Barnes & Noble.
Competence Vs. Trust In Agents
- Foundation models largely solve the competence problem by providing broad world knowledge.
- But agents still need customization and safety checks to avoid costly mistakes.
Treat AI Claims With Healthy Skepticism
- Regulate hype and critically assess AI claims instead of assuming ever-growing competence.
- Do not overtrust models; require audits and transparency before automating critical tasks.