What distinguishes AI and non-AI algorithms? Does the media just call every computer algorithm an AI nowadays?

Computer programs map a sequence of input data to an output. For instance, you might give a program a sequence of people data records and a program might give out the mean age of the people.

When the mapping between the input and output is done by hand through programmer created logic, then it falls into traditional programming. Think of sorting. You think through the quicksort algorithm, implement a sequence of steps and then return a sorted algorithm.

Artificial Intelligence

If the mapping between the input and output is not through a defined sequence of steps coded directly by a programmer, but a model that evolves through iterations defined by the programmer, that's AI. For instance, in face recognition, we don’t directly map the sequence of pixels to words through any defined sequence of instructions, but build a model that creates a mapping between an image and a person’s name.

The mapping in a machine learning program can be thought to be a matrix that when multiplied with the input results in the desired output.

A traditional program would return the same output for a given input, but a learning algorithm can get better (or worse!) over time as the weights get altered.