Is evolutionary psychology just a bunch of "just so" stories? (with Geoffrey Miller)

It's a field that seeks to understand human nature by examining the survival and reproductive challenges faced by our ancestors in prehistory, integrating insights from other primates, anthropology, and behavior genetics.

This resistance is largely political, stemming from concerns about genetic determinism, the naturalistic fallacy (equating 'natural' with 'good'), and the misconception that evolved traits cannot be changed.

They formulate testable hypotheses about the design features of adaptations, gather empirical data, and are willing to reject or refine hypotheses when data do not support them, similar to other empirical sciences.

While useful for understanding trade-offs, environmental responses, and individual differences in cognitive abilities and personality, caution is needed as humans are the last bipedal hominid, and other great apes don't fully represent our evolutionary path.

It's suggestive evidence but not definitive; a universal behavior could arise from practical reasoning and learning applied to common environmental challenges, rather than a specific innate instinct.

Not necessarily; evolutionary biology also explains how behavior changes adaptively in response to different environmental conditions and trade-offs, meaning cultural variations can be highly informative about underlying psychological constants.

There's strong evidence for an evolved aversion to close genetic inbreeding, observed across thousands of species, including plants. However, cultural definitions of incest can vary at the margins (e.g., first-cousin marriage).

Evolutionary psychology suggests emotions like disgust and anger have strong evolved functions (e.g., pathogen avoidance, threat signaling), providing a valuable framework for analysis, even if cultural context influences their expression and specific manifestations.

Understanding which traits the opposite sex finds attractive can help individuals focus on cultivating actionable, improvable traits (e.g., fashion, fitness, humor, conversation skills) rather than unchangeable physical attributes or wealth.

The core difficulty lies in the immense diversity and inherent conflicts within human values, including religious and political beliefs, making it potentially impossible to align an AI simultaneously with all of them without creating new conflicts or alienating large segments of humanity.

The rapid acceleration of AI progress, coupled with the profound difficulty of ensuring AI alignment with diverse and often conflicting human values, suggests a need for a moratorium to prioritize safety and reduce extinction risks.

This mindset is naive because developers are subject to leverage, incentives, and blackmail, and their personal values may not align with broader humanity. Furthermore, controlling highly intelligent, fast, and self-replicating AI systems is fundamentally much harder than controlling humans, even children.

The most important research involves reminding machine learning researchers about the full diversity, complexity, and richness of human values, highlighting that AI alignment is far harder and will take much longer than currently assumed.

Narrow AI is designed to solve specific problems (like Google Maps or medical diagnostics) and is generally beneficial. AGI is a general intelligence capable of performing any intellectual task a human can, posing significant alignment and existential risks.