Emergent Capabilities - Bryant McGill

# Emergent Capabilities **Entity class:** AI capability concept **Domain:** Artificial intelligence / scaling / evaluation **Maturity:** Developed ## Definition Emergent capabilities are abilities that become visible only after changes in scale, training, architecture, tools or deployment context and were not specified as ordinary hand-coded features. ## Mechanism and significance They complicate forecasting because a trained model can display behaviors that are not legible from a conventional software plan or component checklist. ## Source route - [[research/David Robinson on OpenAI Safety, Superintelligence, and the Transhumanist Transition|David Robinson on OpenAI Safety, Superintelligence, and the Transhumanist Transition]] ## Relationships [[wiki/Frontier AI|Frontier AI]] · [[wiki/Emergent Intelligence|Emergent Intelligence]] · [[wiki/Emergent Systems|Emergent Systems]] · [[wiki/AI Interpretability|AI Interpretability]] ## Evidence boundary A surprising capability may reflect thresholded measurement or evaluation design rather than a discontinuity in the underlying learning process. ## Simple Reminders, Quotations, and Thoughts  > “Artificial intelligence is something we grow. The conference room where people work overnight during major training runs, watching the dials and making sure everything remains on track, is called ‘the nursery’ internally. There is a sense that we do not fully understand what is happening.” > **— David Robinson**, *The Ezra Klein Show, October 7, 2026* [[reminders/Machine Succession/Artificial Intelligence Is Something We Grow by David Robinson|Artificial Intelligence Is Something We Grow by David Robinson]] > “Calling frontier AI ‘software’ invokes a model of engineering in which people begin with a plan, proceed step by step, specify acceptance criteria and iterate until each part works as intended. Training a large model is not like that, and unlike ordinary software, a trained model can exhibit emergent capabilities and behave as though it knows how it is being used.” > **— David Robinson**, *The Ezra Klein Show, October 7, 2026* [[reminders/Machine Succession/Frontier AI Is Not Ordinary Software by David Robinson|Frontier AI Is Not Ordinary Software by David Robinson]] > “We now have systems at the limit of our ability to understand and control what they are doing. The people worried about science-fiction scenarios warned that capability was growing exponentially, and there is no law of science saying progress will stop when it becomes useful. There is nothing special about the zone between useful and frightening.” > **— David Robinson**, *The Ezra Klein Show, October 7, 2026* [[reminders/Machine Succession/Nothing Guarantees AI Progress Stops Between Useful and Frightening by David Robinson|Nothing Guarantees AI Progress Stops Between Useful and Frightening by David Robinson]]