Can AI Have Common Sense? #AI #CommonSense #LSE

Why It Matters

Re-centering AI on common sense is crucial for safety, trust, and real-world reliability—preventing harmful edge-case behavior and ensuring machines act in socially intelligible ways. Without this shift, AI may continue to optimize tasks at the expense of human norms and community values.

Summary

A speaker uses a pizza-delivery robot thought experiment to argue that AI lacks common sense and so can behave dangerously when tasks omit everyday human norms. They define common sense three ways: multisensory integration, everyday background reasoning about what exists and matters, and social/community norms like rituals. The talk notes that early AI research in the 1950s aimed to encode common sense but over the past 50 years industry shifted toward performance and automation, sidelining those goals. The speaker urges a renewed focus on building systems that combine sensors, reasoning, and social awareness to restore practical common sense in AI.

Original Description

From pizza delivery robots to funeral rites, what happens when AI has no common sense?

Professor Martin Bauer from LSE's Department of Psychological & Behavioural Science, explores why the machines we build are missing something fundamentally human. Decades after common sense was side-lined as AI's founding ambition, he considers the scenario of bringing it back. But what would a truly common sense machine actually look like?