Nomography of the Machine
Mathematical Models, Parameter Spaces, and Subject Matter Eligibility
An analytical dissection of patent eligibility under 35 U.S.C. § 101 following Alice Corp. v. CLS Bank. We map the boundary where mathematical formulas in transformer architectures transform from unpatentable abstract ideas into patent-eligible technological improvements.
I. The Abstract Idea Exclusion
Under longstanding Supreme Court jurisprudence, laws of nature, natural phenomena, and abstract ideas are excluded from patent protection. When applied to machine learning algorithms, the Alice/Mayo two-step framework requires patent drafters to demonstrate an 'inventive concept' that significantly exceeds generic mathematical matrix multiplication.
Precedential Parallels Engine
Tracing modern AI copyright disputes back to their historical technological counterparts: cameras, VCRs, search crawlers, and software APIs.
Analytic Note: Parallels indicate similarity in underlying legal and technological mechanisms, not factual equivalence. These historical lineages illuminate how courts historically adapted copyright doctrines to technological shocks.
When photography was challenged as purely mechanical recording incapable of human authorship, SCOTUS held that mental conception, staging, lighting, and posing supply the requisite creative spark.
“Photographs are useful, new, harmonious, characteristic, and graceful pictures, so far as they are representatives of original intellectual conceptions of the author.”— 111 U.S. 53 (1884)
Judges are now evaluating whether prompt engineering, ControlNet constraints, and seed manipulation constitute the modern equivalent of Sarony's manual lighting and drapery arrangement.