Emergence
Emergence is when a system shows behaviour that none of its individual parts has on its own. The behaviour comes from how the parts interact, not from any single part.
The classic reference is physicist Philip Anderson's 1972 essay "More Is Different", which argued that "at each level of complexity entirely new properties appear" (Anderson, 1972). Knowing how particles behave doesn't tell you how chemistry works, and knowing chemistry doesn't tell you how biology works. Wei et al. summarise the idea as: "Emergence is when quantitative changes in a system result in qualitative changes in behavior" (Wei et al., 2022).
It shows up all over the place:
- In games, Emergent Complexity is when a simple ruleset produces complex interactions, and Emergent Gameplay is when players come up with strategies that the rules never mention.
- In Large Language Models, Emergent Abilities are capabilities that only appear once a model is large enough, like Chain-of-Thought Prompting. Whether these jumps are real or an artefact of how we measure them is still debated.
The common thread is that you can't understand the whole just by studying the parts.
References
P. W. Anderson. More Is Different. Science, 177(4047):393–396, 1972. doi:10.1126/science.177.4047.393. ↩
Jason Wei, Yi Tay, Rishi Bommasani, Colin Raffel, Barret Zoph, Sebastian Borgeaud, Dani Yogatama, Maarten Bosma, Denny Zhou, Donald Metzler, Ed H. Chi, Tatsunori Hashimoto, Oriol Vinyals, Percy Liang, Jeff Dean, and William Fedus. Emergent Abilities of Large Language Models. 2022. doi:10.48550/ARXIV.2206.07682. ↩