Interview
Gary Marcus: Nature vs Nurture is a False Dichotomy | AI Podcast Clips
- Anticipates a perception among field practitioners that integrating innate mechanisms with machine learning may be incorrectly viewed as "cheating."
- Proposes that children possess innate concepts of space, time, other agents, places, and mental algebra lacking specific causal notions.
- Posits that mechanisms representing space and time can communicate, reflecting the relational nature of human reasoning.
- Suggests evolutionary discovery of superior ideas requires a large initial pool of ideas for a successful path to emerge.
- Indicates that starting with richer innate structures or libraries could accelerate progress speed.
- Expects cumulative evolution involving related points to enable the process to proceed at an increased rate.
- Suggests that understanding the implementation of dogonation could provide insights beneficial for AI development.
- Proposes that biomimicry of biological problem-solving methods can serve as a shortcut to accelerate AI development.
- Draws a parallel to material science, noting that while biomimicry is common there, a similar analog should be applied to AI.
- Recommends examining cognitive science fields, including psychology, neuroscience, and linguistics, to gain further insight into biological systems.