The AI Field Guideby Stephen Hunt's SFcrowsnest
The AI Field Guide / K
3 terms, explained without the techno-murk.
A method that judges a new item by looking at the most similar known items nearby.
Imagine moving to a new street and guessing its character from the nearest few homes. The 'k' is simply how many neighbours the method consults. The result depends heavily on how similarity and distance are measured.
A new flower is classified by checking the labels of the five most similar measured flowers.
The latest point in time covered by a model's built-in training knowledge.
A model may know nothing reliable about events after its cutoff unless it can search or use another current source. Even information before the cutoff can be incomplete or wrong.
A model with an older cutoff needs live search to answer who currently holds an office.
A structured map of things and the relationships between them.
Instead of storing isolated facts, a knowledge graph links them: a person works for a company, the company owns a product and the product belongs to a category. It resembles a labelled web of index cards connected by string, making relationships easier for software to follow.
A museum graph connects artists to artworks, styles, dates, places and related exhibitions.