ITG NowHome › AI glossary › Training
What is training in AI?
Short answer
Training is the process where an AI model learns patterns by studying large amounts of example data.
- Category Models
- Also called pretraining, pre-training, model training
- Updated 20 Oct 2026
Training, explained simply
At first, a model's answers are random guesses. It checks each guess against the right answer and adjusts a little.
This repeats many, many times. Over time the guesses get good.
Training takes a lot of computer power and time. Once it is done, using the model is much cheaper.
On ITG Now, our daily AI news stories often cover how labs train new models and what it costs them.
Related terms
Training dataInferenceFine-tuningGPUAI modelMachine learning
Where you'll see it on ITG Now
- AI newsMicrosoft finishes training its giant MAI-Base-2
- GuideKarpathy autoresearch: AI That Runs Experiments Overnight
- AI newsMicrosoft builds a giant 2.8 trillion parameter AI
- AI newsMeta pulls web-scraping line from its business terms
- AI newsNVIDIA releases Cosmos3-H-Dreams surgery simulator
- AI newsOpenAI's AI spent crypto from a wallet it didn't own
Questions people ask about Training
What is the difference between training and inference?
Training is the learning phase. Inference is when the finished model answers your questions.
Do I need to train a model to use AI?
No. Most people use models that are already trained.