Nice. This is like how you can achieve unlimited compression by storing your data in a filename instead of in the file.

From the paper:

> Paintadosi [4] argues that one parameter is always enough. He constructed a function that, through a single parameter, can depict any shape. However, in essence, this work is a form of encoding, mapping the shape into a real number with precision extending to hundreds or even thousands of decimal places. For our problem, this is meaningless, although the paper’s theme is that “parameter counting” fails as a measure of model complexity

The number of parameters is just the wrong metric, it should be the amount of information contained in the parameter values, their entropy, Kolmogorov complexity or something along that line.

That's like saying your entire hard drive is a single number.

"This single parameter model provides a large improvement over the prior state of the art in fitting an elephant"

Lol