Exploring the Misguided Notion that ‘Merely Doing It on A Computer’ Negates Eligibility

The U.S. Supreme Court’s Alice decision alleges that “…merely requiring generic computer implementation fails to transform that abstract idea into a patent-eligible invention.” And the Patent Eligibility Restoration Act (PERA) of 2023 alleges that “adding a non-essential reference to a computer by merely stating, ‘do it on a computer’ shall not establish such eligibility.” Clearly, it is assumed that “merely” doing something on a computer or “merely” saying “do it on a computer” is not a desirable thing in the eyes of some; a computer supposedly invalidates the inventive effort and “merely” doing something on a computer is undeserving of even consideration of a patent.

But no person of ordinary skill at this stage can say “just do it on a computer” and make it work. Not yet, anyway. We all have to go through the arduous process of programming. The “just do it on a computer” allegation is a myth. However, that is about to change.

Novel Cryptography

Here is a real-life and recent computer implemented invention that I made, related to machine cryptography. I use it because it is not hypothetical. And because I developed and implemented it on a computer.  It is inspired by computer implemented radix-n addition. In the Advanced Encryption Standard (AES), (the world’s most widely used encryption standard), one component is a step called AddRoundKey(). This is a processing of four blocks of 32-bits. It is performed during encryption and may be reversed during decryption. It is a computationally simple operation, a carry-less addition. I treat a 32-bit word as four eight-bit words that are processed as a radix-256 addition using the carry-less addition with a novel carry. The carry function is modified in an unpredictable way. Decryption may be a subtraction with a corresponding modified borrow function. The result looks like a random number, with randomness increasing due to unpredictable carry propagation. It is no longer what we understand to be an addition. It is a composite computer operation. It works, I programmed it, and it has aspects that dramatically increase security of encryption like AES-GCM and ChaCha20.

In the context of PERA, I make my claims “practical” by requiring a minimum performance like an equivalent of processing at least 1 million bytes per second and limiting it to exchange of data between computers.

An opponent likely will argue that this is “just” a mathematical discovery. Unusual, but mathematical none the less. And what I did was “merely” executing the abstract discovery on a computer. So, what is it? Did I make it practical on a computer as protected by PERA or did I just “merely do it” on a computer, as alleged by PERA?

Doing it “merely … on a computer” has two meanings. The first one is implementing something we already know on a computer. The second one is to ask a computer to do something that for some reason we cannot or do not want to do ourselves.

A further example. I need a primitive polynomial of degree 100 over a large base field GF(256), which itself is an extension field. A mind-numbing exercise. Turns out it may be achieved with the help of the public Magma Calculator. But it still requires non-trivial programming steps.

Currently, you cannot “merely” instruct ChatGPT or other AI to do any of the above. I tried. But recommended instructions regress rapidly to almost programming it yourself. AI keeps repeating the wrong code even after being instructed to change. Not unlike a GPS system that persists in guiding you back to an undesirable route. In fact, AI’s coding is quite buggy. Errors are made by AI in basic syntax and in concepts and solutions are often, especially in my field, more complicated than needed. It forces a very incremental way of programming that is tiresome to an experienced programmer.

But, one can already see, how ChatGPT or Bard, with some extra capabilities, may improve and be able to deliver executable code by “merely” asking for it in general terms. In which case an invention was “merely done on a computer.” One may then allege that you have really no idea what you are effectively doing and that you are “merely using a computer” to get a result.  This is what SCOTUS says: “Thus, if a patent’s recitation of a computer amounts to a mere instruction to “implemen[t]” an abstract idea “on . . . a computer,” Mayo, supra, at ___ (slip op., at 16), that addition cannot impart patent eligibility.

The “mere” thus refers not only to a “known” idea, but to any “abstract” idea. We expect that PERA will get rid of the SCOTUS judicial exceptions, but PERA’s “merely” clause brings it right back in. Despite “Patentability shall not be negated by the manner in which the invention was made” there continue to be efforts to shut out not only computer implemented inventions, but also computer assisted inventions.

Doing it on a Computer Facilitates Inventions!

In the future, AI will assist in generating inventions based on what you may call an inventor’s “brain fart.” AI will democratize inventing. What I did as described above and with considerable efforts and based on knowledge of Number Theory may then be done by anyone having some “idea.” Just try it and if viable, AI will generate the code. Some subject knowledge will be required, but not to the extent to do it yourself. Not unlike designing a digital filter, where you specify its parameters such as frequency range, stop and pass bands and the like, but a computer computes the actual filter coefficients. But several steps more advanced.

The Internet promoted democratizing inventions. Many of those inventions were stopped being patent eligible as “mere business methods”, ignoring the fact that integrated transactions over multiple networks were nothing short of miraculous before ERP systems (and cryptographic security) came to be. We likely will see these “stopping” efforts again against AI assisted inventions that are submitted for patents.

As AI may democratize inventing, our patent system is showing a reversing anti-democratizing trend, favoring big infringers and working against small inventors. This is particularly troublesome as we attempt to out-innovate China. It is telling that AI and its launch as a public product is again (like the Internet) an American phenomenon. Not a Chinese one, not a Japanese one and not even a European one. It is a strength of the American inventor to apply new technologies, materials and concepts into new inventions. No matter how daunting. We need to keep that alive for the small inventors.

I believe PERA to be a needed step to improve our patent system. However, this “merely stating … do it on a computer” language stands out as a problem to me. Inventing by using a computer may—no, will—become a lot easier. And not just in code generation, but in drug development and other areas where modeling and simulation are essential and self-correcting feedback will lead to self-improvement of the AI application.

I believe AI still will need prompting by an inventor to get started. The “prompts” will become much more advanced as many suggestions for improvement will come from AI itself.  That is why future efforts of reducing an idea to practice may become in many cases “merely doing it on a computer.” Which may diminish human efforts and lead to inventions on a level we never imagined. Which I believe is a good thing.

But …! It may entice SCOTUS to reinstate patent ineligibility when asked to do so on this matter. Not as a judicial exception this time, like in the past when this Court started defining exceptions to patent eligibility on its own, culminating in the damaging Alice v. CLS Bank decision, but this time because Congress intends “merely computer implemented” to be ineligible. The current text of PERA disparages “merely” doing it on a computer, without defining what that means. And in view of AI, the issue of “merely doing it on a computer” will become more urgent.

Image Source: Deposit Photos
Author: donscarpo
Image ID: 9469910