Learn AI · Level 3 · Tech and tools · step 3 of 16 · word 2 of 3
What is chain of thought in AI?
Chain of thought is when an AI model works through a problem step by step before giving its answer, which makes it better at maths, logic and planning.
In plain English
People get tricky sums right more often when they write the steps down. Language models are the same. A 2022 Google study found that asking a model to “think step by step” made it much better at word problems.
Today’s reasoning models do this on their own: they write a hidden or visible “thinking” section before they answer.
An example
Ask: “A shirt costs $25 after a 20% discount. What was the full price?” A quick answer might say $30. Working it through, the model notes that $25 is 80% of the price, so the full price is $31.25.
Watch out
The steps a model shows are not always the steps it really took. A neat explanation can still lead to a wrong answer, so check the result, not just how convincing the reasoning looks.
Learn AI path · Level 3 · Tech and tools · step 3 of 16
Next wordContext engineering →See all of level 3