Emergent Behavior
Unexpected capabilities that appear in large AI models but are absent in smaller versions of the same architecture.
Overview
Emergent behaviors are capabilities that appear in large language models at certain scale thresholds but are not present in smaller models. Examples include in-context learning, chain-of-thought reasoning, and few-shot task performance that seem to 'emerge' abruptly as model size increases.
Debate
The existence of true emergence is debated. Some research argues that seemingly sudden capability jumps are artifacts of how we measure performance (discontinuous metrics), and that capabilities actually develop gradually. Regardless, understanding which capabilities scale predictably and which don't is crucial for AI safety — unexpected capabilities at larger scales could include dangerous ones.