Emergent behavior
System-level behavior or capability that arises from interactions among components or from a system’s scale or complexity, rather than being specified as an isolated outcome.
Natural Language Processing in Action, Second Edition
An emergent capability is a behavior or performance characteristic that appears to arise only when a system reaches a particular scale or complexity.
- How it is used
- Frames surprising LLM capabilities as claims to evaluate against scale, benchmarks, reliability, and social consequences.
- How deep it goes
- Defines emergence, relates it to parameters and data, compares scaling patterns, and stops short of establishing general intelligence.
- Examples
- Uses LLM-size plots, emergence-by-task data, and a Llama arithmetic answer that contradicts its own explanation.
- What you can do after
- Evaluate/understand emergence claims by comparing scale, benchmark trends, model size, and observed reliability.
Read how Natural Language Processing in Action, Second Edition explains emergent behavior →
Introduction to Generative AI, Second Edition
An emergent property is a capability or behavior that appears when a system reaches sufficient scale or complexity and is not apparent in smaller versions.
- How it is used
- Frames unexpected capabilities in large language models and unexpected group-level outcomes in interacting AI agents.
- How deep it goes
- Defines qualitative change and compares examples, then questions apparent thresholds through evaluation metrics rather than explaining a mechanism.
- Examples
- Uses a vegetarian-menu prompt, coordinated pricing agents, and agents reinforcing incorrect answers to illustrate different outcomes.
- What you can do after
- Evaluate/understand it by distinguishing observed performance from disputed claims of sudden emergence and by checking group behavior, not only individual agents.
Read how Introduction to Generative AI, Second Edition explains emergent behavior →
Sutskever's List
A capability that appears as a model is scaled or otherwise developed and is not readily apparent in smaller or less capable versions.
- How it is used
- Describes capabilities that become visible as models scale, framing GPT-3’s in-context learning as a turning point in hyperscale.
- How deep it goes
- Distinguishes smooth training-loss trends from abrupt benchmark gains and regime changes, but offers no universal rule for predicting specific capabilities.
- Examples
- Uses GPT-3, multistep reasoning, multidigit arithmetic, translation, code generation, and benchmark score fluctuations as examples.
- What you can do after
- Evaluate apparent capability jumps by comparing benchmark scores with training-loss trends and recognizing that evaluations sample only part of possible inputs.
Generative AI in Action
An emergent property is a capability or behavior that arises from a system's components or scale without being explicitly programmed as an individual feature.
- How it is used
- Explains why LLMs can generalize from prompts to logic solving, code analysis, and other capabilities not individually programmed.
- How deep it goes
- Defines implicit emergence, demonstrates scale thresholds, and stops short of explaining the exact internal logic producing each behavior.
- Examples
- Uses proportional reasoning, step-by-step math, C-to-Python translation, and a scaling curve as concrete demonstrations.
- What you can do after
- Apply it via prompts to solve expressed problems and translate unfamiliar code, while recognizing scale-dependent and unanticipated behavior.
Read how Generative AI in Action explains emergent behavior →
Grokking AI Algorithms, Second Edition
A collective pattern or capability that arises from interactions among simpler individual agents.
- How it is used
- Frames swarm intelligence as collective problem solving produced by many simple individuals interacting.
- How deep it goes
- Explains pheromone reinforcement, shortest-path convergence, and flocking rules through concrete biological examples.
- Examples
- Uses ant colonies, an asymmetric bridge experiment, bird flocks, and a bee swarm as examples.
- What you can do after
- Evaluate how simple local behaviors can produce coordinated groups and inspire optimization algorithms.
Read how Grokking AI Algorithms, Second Edition explains emergent behavior →
Related concepts
- language modelproduced by · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.
- indirect messagingproduces · Indirect messaging is a communication pattern in which a sender and receiver exchange information through an intermediary or event channel rather than direct calls.
- multi-agent systemproduced by · A multi-agent system is a computational system composed of multiple agents that interact or coordinate to perform tasks or pursue goals.
- scaling lawcontrasts with · A scaling law is an empirical relationship describing how a system’s properties change as its scale or resource levels vary.
- metricmeasured by · A metric is a quantitative measure used to characterize or assess a property, condition, behavior, performance, or outcome.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.