Account for the HLMI–FAOL forecast gap

Determine the factors responsible for the large and persistent discrepancy between survey respondents’ predicted arrival times for High‑Level Machine Intelligence (defined as unaided machines accomplishing every task better and more cheaply than human workers) and Full Automation of Labor (defined as all occupations being fully automatable) reported in the 2023 Expert Survey on Progress in AI and earlier editions; specifically, assess whether this gap is a framing effect or arises from differences in interpretation, definitions, or question context.

Background

The survey elicited timelines for two closely related milestones: High‑Level Machine Intelligence (HLMI) defined over tasks, and Full Automation of Labor (FAOL) defined over occupations. Across the 2016, 2022, and 2023 surveys, the median (50%) prediction for FAOL consistently lagged HLMI by more than sixty years, a surprising divergence given their conceptual similarity.

The authors note potential explanations, including framing effects, definitional nuances between tasks and occupations, differences in instructions (e.g., HLMI prompts assume human scientific activity continues without major disruption), and order/context effects from preceding questions in the FAOL block. However, they explicitly state that they do not know what accounts for this gap.