Skip to content

GPAI.WIKI

Concepts

The wiki

Short, canonical, heavily cross-linked definitions. Everything else on this site links back into here, so keep entries neutral and keep them current.

sort

C3
Catastrophic forgetting catastrophic interference The abrupt loss of previously learned capability when a network is trained on new data, caused by the same weights being reused for both.
Continual learning lifelong learning · incremental learning Learning from a non-stationary stream of experience without retraining from scratch and without losing what was already learned.
Credit assignment temporal credit assignment · structural credit assignment Determining which of an agent's earlier decisions were responsible for a later outcome.
G1
General-purpose AI GPAI · general-purpose artificial intelligence · foundation model A system whose competence is not bounded to the tasks it was trained on, and which can be directed at new tasks by description rather than by retraining.
S1
Scaling laws neural scaling laws Empirical power-law relationships between model size, dataset size, training compute and loss, which hold across many orders of magnitude and permit extrapolation.
W1
World model learned simulator · forward model · dynamics model A learned predictive model of environment dynamics, used to evaluate candidate actions without executing them.