L036 · How AI thinks · 2 min
The knowledge cutoff
A model only learned from text up to a certain date, so on its own it does not know what happened after that.
01 · The idea
A model is trained on a huge collection of text gathered up to some point. That point is its knowledge cutoff. Anything newer simply was not in its training data.
Ask about last week's news, today's exchange rate or a new phone model, and a chatbot without web access may give an old answer, guess, or say it does not know.
The cutoff is often months before the model is released, and people keep using a model long after that. So the gap between what it knows and today can be a year or more.
Many chatbots can now search the web to fill the gap. When they do, they usually show sources. If there are no sources, assume the answer comes from older training data.
02 · Think of it like this
It is like someone who has just returned from a long trip with no phone. They know a lot, but nothing about what happened while they were away.
03 · Do it now
04 · Quick check
Words in this lesson
See also
L005How AI thinks · 2 min
Why AI makes things up
AI is built to produce a likely answer, not a true one, so sometimes it confidently invents facts.
L002How AI thinks · 2 min
How machines learn
A machine learns by guessing, checking how wrong it was, and adjusting itself a tiny bit, billions of times.
L001How AI thinks · 1 min
What is AI, really?
AI is software that learned patterns from examples instead of following rules a person wrote.
L039Under the hood · 3 min
How fine-tuning works
Fine-tuning takes a model that already knows language and trains it a bit more on your own examples, so it picks up a style or a task.