Training
Also known as: Model Training, AI Training
In one sentence
The process of feeding large amounts of data to an AI system so it learns patterns, relationships, and rules, enabling it to make predictions or generate output.
Explain like I'm 12
Like practising flashcards thousands of times until you just know the answers. AI looks at millions of examples — text, images, numbers — over and over until it figures out the patterns by itself.
In context
Training a large language model like GPT-4 involves processing trillions of words from books, websites, and code over weeks using thousands of GPUs, costing tens of millions of dollars. Training a spam filter is simpler — showing it thousands of labelled spam and legitimate emails until it can classify new ones accurately. There are different stages: pre-training (learning general knowledge from massive data), fine-tuning (specialising on specific tasks), and RLHF (learning from human preferences). Each stage builds on the last.
See also
Related Guides
Learn more about Training in these guides:
Distributed Training for Large Models
AdvancedScale AI training across multiple GPUs and machines. Learn data parallelism, model parallelism, and pipeline parallelism strategies.
8 min readTraining Efficient Models: Doing More with Less
AdvancedLearn techniques for training AI models efficiently. From data efficiency to compute optimization—practical approaches for reducing training costs and time.
10 min readSynthetic Data Generation for AI Training
AdvancedGenerate training data with AI: create examples, augment datasets, and bootstrap models when real data is scarce or sensitive.
7 min read