Understanding AI through the algorithms they compute
IBM, Monday, August 18th, 2025
A decades-old approach to measuring algorithmic complexity could provide a window into better understanding how AI systems compute.
Midway through the 20th century, the study of computer algorithms took a major leap forward with the development of computability theory and models like Turing machines. Invented by Alan Turing to formalize what it means for a problem to be computable, Turing machines are simple abstract models of computation that use algorithms - step-by-step instructions used to solve a problem - to process information.
An algorithm can be as simple as following a recipe to bake a cake, or as complex as Dijkstra's algorithm, which finds the shortest path between nodes in a graph. These theoretical breakthroughs laid the groundwork for the systematic study of algorithms we see today.