Post
IBM patented Euler's 200 year old math technique by repackaging a well-established mathematical concept of generalized continued fractions with new, flashy branding like "CoFrNets" and "ladders."
They implement this in PyTorch, using the reciprocal as a nonlinearity, and achieve mediocre results (61% accuracy on a simple dataset).
https://leetarxiv.substack.com/p/ibm-patented-eulers-fractions
As a mathematician, I say:
No mathematical principle should ever be patented or require a license for its usage.
Imagine, if that happened to FFT!
--
Note: "FFT" means "Fast Fourier Transformation", which is a pretty much genius method for efficient signal processing, basically used everywhere, where electromagnetic waves occur.
@mina I'm against the whole concept intellectual property as it serves to silo human knowledge, but it's especially heinous when it comes to mathematics
The only positive thing I can imagine about copyrighted/patented scientific evidence is that it will not just prove that the Science Spectrum Theory is right and that falsiability is just a part of the spectrum of science, but also that those pro-Corporations studies/research are indeed what we claim and say they're about.
And I'd say: Nearly all mathematicians would agree on that.
A space for Bonfire maintainers and contributors to communicate
