Big Brains Are We Making AI Too Human?, with James Evans
13 snips
Jun 12, 2025 
James Evans, a University of Chicago sociologist and data scientist, discusses the limitations of training AI to think like humans. He argues that this approach stifles scientific discovery and advocates for the creation of 'cognitive aliens'—AIs that think differently to unlock breakthroughs. Evans explores how expanding AI’s sensory capabilities can lead to new forms of innovation and emphasizes the need for a shift in AI development to overcome biases in research. He also addresses AI's paradoxical role in both hindering and enhancing scientific inquiry.
AI Snips
Chapters
Transcript
Episode notes
AI Mimics Human Intelligence by Default
- AI is traditionally designed to mimic human intelligence as defined historically by pioneers like John McCarthy. - This human-centric model may limit AI's potential for novel discovery.
AI Narrows Rather Than Expands Science
- AI currently accelerates research mostly in areas with abundant existing data, causing scientific fields to narrow. - This reliance on existing data restricts discovery of new disciplines or novel scientific spaces.
AI Predicts Science but Risks Predictability
- AI models can predict not just who will make scientific discoveries but exactly what those breakthroughs will be. - Such predictive capabilities risk reinforcing predictable science, missing radical, unexpected breakthroughs.