The basicsAI glossary
Knowledge cutoff
The date at which a model's training data stops: anything published afterwards does not exist for it. Because training and then testing take months, it often falls well before the model's release. Hence outdated facts presented as current, which real-time web search only partly corrects.
In other words
A traveller back from a long expedition without signal: they talk confidently about news that has moved on since they left.
Alyriaz articles that mention it
- How does an AI search the web?
When an AI cites web sources, it follows a four-step chain, each of which can fail. We explain the mechanism and what to check.
- AI or search engine: which one to use, and when?
Many people have swapped web search for AI, or the other way round, without distinguishing what each does well. We offer a clear split, with the cases where one is clearly superior to the other.
- AI and your money: what it can do, and what you should never ask it to do
More and more people are asking an AI about their personal finances. We distinguish between uses where it brings real value and those where it exposes them to real risk, without jargon and without moralising.
- Why an AI doesn’t know yesterday’s news
The knowledge cutoff date marks the end of a model's training data. We explain what it implies, why it produces subtle errors, and how real-time search changes the game without solving everything.
- How does an AI "learn"? A model's journey, from raw chaos to the polished assistant
How does a model go from a pile of raw data to an assistant capable of answering you politely? We describe the three main stages of AI training (pre-training, fine-tuning, RLHF) with a simple analogy, to understand where its strengths and biases come from.