For Heads-of · Practitioner
Emergent Capabilities
Abilities that appear in AI models at certain scales but were not explicitly trained for, suggesting complex behaviors arising from scale alone.
- emergent capabilities
- scaling
- unpredictability
- model behavior
Examples
In-context learning (learning from examples in the prompt without prior training), chain-of-thought reasoning, and translation between languages never paired in training data.
Unpredictability
Emergent capabilities are difficult to predict in advance; they appear suddenly as model scale crosses certain thresholds.
Governance Challenge
Emergent capabilities introduce uncertainty: a scaled-up model may have unexpected behaviors that were not present in smaller versions, complicating safety assurance and testing.