The processing power needed to train and run AI models. Measured in FLOPs (floating-point operations) for training and tokens-per-second for inference. The AI industry's most constrained resource — access to enough GPUs determines who can train frontier models. NVIDIA dominates the compute market.
Graphics Processing Unit.
Tensor Processing Unit.
The process of teaching an AI model by exposing it to data and adjusting its parameters to minimize errors.
A mathematical function applied to a neuron's output that introduces non-linearity into the network.
An optimization algorithm that combines the best parts of two other methods — AdaGrad and RMSProp.
Artificial General Intelligence.
Browse our complete glossary or subscribe to our newsletter for the latest AI news and insights.