Clicked Gallery

What is training (in AI)?

Highlighted from a real engineering doc. Explained by Clicked.

Used in a sentence

Engineering Notes · AI Systems

Training the model consumed several months of compute before a single question was ever asked of it.

The reader highlighted one word in the docs. Clicked explained the technical term “training” in simple terms:

Explained in three depths

Same facts, different vibe — Slang mode 😎

The Clicked way

●○○

Overview

Training is how a model gets built. It is shown enormous amounts of material, usually text but often images and code as well, and its internal settings are adjusted until its predictions stop improving. Further stages then shape how it responds rather than what it knows. The settings are frozen for release only after all of that, and nothing you type in a conversation changes them.
●○○

Overview

Every model you have ever used arrived pre-baked. Months of electricity went into shifting its internals a hair at a time until the output stopped being gibberish, and then it shipped. Labs can go back and adjust it afterwards, and for open models so can anyone with serious hardware. What definitely does not adjust it is you, typing. 😎

A quick take — often all you need.

●●○

Detail

A model begins as millions of numbers set at random, producing nonsense. Training hides the next word in a passage, compares the model's guess against the real answer, and nudges the numbers slightly toward being less wrong, a method called stochastic gradient descent. Repeat that a few trillion times and the numbers settle into something that predicts language well. Why not program the rules of language directly instead? Because nobody can state them: explaining why "strong tea" sounds right and "powerful tea" does not would take a lifetime, and that is one phrase out of millions. Learning from examples sidesteps the need to write any rule down. Three stages follow one another: pretraining on the bulk of the material, fine-tuning on a narrower set of examples, and reinforcement learning from human feedback, where people rank answers. All three adjust the same numbers. Pretraining is the expensive stage and finishes long before release, which is why every model has a knowledge cutoff. Fine-tuning is cheaper and can happen later, and with open-weight models anyone can do it.
●●○

Detail

Training is the bit where the thing actually gets made, and the mechanism is simple enough. Hide a word, let it guess, see how badly it missed, shift every number a hair toward better. Do that a few trillion times across a data centre and months of electricity, and noise turns into fluent language. Why not hand it the grammar and skip all this? Because try putting into words why "a quick little walk" is fine but "a little quick walk" is not, then do it several million more times, for every language. That list has never been finished by anyone, so the machine goes and finds the patterns itself. Which is also why "I trained ChatGPT to write in my style" is not quite what happened. You gave it instructions. The app may even save them for next time, but the model itself is untouched. Actual training leaves a permanent mark on the numbers and costs roughly what a large building costs. 😎

Want more? One click digs deeper.

●●●

Analogy

Learning to catch a ball. Nobody hands you the equations for a parabola, and reading them would not help much anyway. You put your hands out, miss, and something adjusts very slightly. A few thousand catches later you are good at it, and you still could not write down the rule you learned.
Deepest level reached.
●●●

Analogy

Getting good at a video game. You did not read the physics documentation. Your character died two hundred times and your hands quietly figured it out. Ask afterwards what you learned and you can barely put it into words, which is embarrassing at parties and also exactly how this works. 😎
Deepest level reached.

Unfamiliar concept? A real-world example makes it click — fresh analogies on tap.

AI explanations may contain errors · Not professional advice

Formal definition — The same term, explained the usual way

Training is the process of iteratively adjusting a model's parameters to minimise a loss function computed over a dataset, typically by stochastic gradient descent with backpropagation. Pretraining is followed in most production systems by further stages such as instruction tuning and reinforcement learning from human feedback, all of which occur before deployment. Inference, by contrast, applies the resulting fixed parameters to new inputs without modifying them.

Want Clicked to explain terms like “training” directly in your browser — including on PDFs?

Add to Chrome — Free

50 free Explanations · No credit card required