26 thoughts on “AI and Inventorship – USPTO Request for Comments

  1. 2

    I think, in reality, that we aren’t ready for this. That the inventor still should be the person or persons that has agency over the AI.

    Revisit this is 10 years when the field has advanced. Right now AI isn’t that advanced.

    1. 2.1

      “Agency over”…?

      You mean, like a slave owner had agency over their slaves?

      Sorry but no — ‘agency over’ has already been deemed insufficient.

    2. 2.2

      The time to think about this is before it exists in a commercially significant form. At this stage, no one knows who stands to win or lose from AI inventorship. Once it becomes clear who will be the financial winner or loser from an AI inventorship dispute, critical faculties will become warped by personal considerations.

      1. 2.2.1

        It is too bad that Greg has chosen “to not see my posts,” as he would realize that he is echoing my position on the matter that I expressed when the DABUS case first broke out on the scene.

        His loss of a chance to learn.

    1. 1.1

      … by the by, questions also came up along the lines that I have previously pointed out: effect on State of the Art vis a vis that non-human fictitious juristic person known as Person Having Ordinary Skill In The Art.

      This point is HERE and NOW — regardless of any discussion of Singularity or General Artificial Intelligence (which has often confused some in regards to the discussions of AI as inventor).

      1. 1.1.1

        Why is that any more of a problem than it’s ever been? Tools have always progressed, and tools have always made what was impossible (or nearly so) quite possible.

        Now PHOSITA has AI to help them invent. What really changes?


            … this has to do with understanding the difference between merely using a tool (while retaining human intent) and having the tool provide something outside of human intent.


              “the difference between merely using a tool (while retaining human intent)”


              So an inventor using a microscope or gas chronograph needs to know what they were looking for (“intent”) if they find, and then apply a useful new quality of a material?



                Is it deliberate?

                You haven’t been following along with what it means for an AI to be the entity that does the inventing, have you?

                Let’s see if I can make this so Ow for you:

                See the Naruto case.

                Equate AI with the simian (not the camera).


          Now PHOSITA has AI to help them invent.

          First, you appear to not recognize that PHOSITA is not a real person.

          Second, your use of “help THEM invent” misstates AI as a mere tool rather than the entity DOING the inventing.

          I suggest that you talk with your patent attorney about these things that you want to use so colloquially.


            Without going to the argument of whether AI can invent, how do we square the round hole of process within the USPTO if we assume AI can be named as an inventor?

            As an employer, do I “hire” AI? It is a consultant who signs a consultant agreement that transfers its property rights to my company? Who signs that agreement? How or from whom do I get an inventor signature? How do I investigate whether the AI believes it contributed to the scope of the invention? Can I subpoena the AI such that it “remembers” its inventive contribution to the invention?

            Has anyone those about these practical aspects once we decide on whether AI should be named as an inventor?


              Naming AI as an inventor and being able to obtain a patent on an invention to which an AI is an inventor (either single inventor — per DABUS, or co-inventor — no case law yet) are two very different things.

              Given DABUS (and the equivalent of the Simian in the Naruto case), people are not going to like what happens in AI as co-inventor situations.


              I do not wish to say that there are no problems with listing AI as an inventor, but none of those things that you list are really insuperable problems. We already have processes by which the heirs of an inventor or the colleagues of a dead inventor can apply for a patent after an inventor has died. The dead inventor is no more able to sign documents or give depositions, but we allow others to sign on the deceased’s behalf and just live without the deposition testimony. If we want to recognize AI inventorship, we can do the same for AI as we do for dead humans.


                Greg’s “solution” runs smack right into the Lockean nature embedded in the US Sovereign’s patent system.

                Which is why the Naruto case is instructive.


                The analogy is in error. In the case of the dead inventor, the inventor is incapacitated and cannot perform the functions of signing papers. A legal representative of the estate performs the duties as if s/he were the inventor, e.g., signing papers. Same goes for inventors declared legally insane etc.

                Why would AI need a legal representative? It is still functioning as an AI, and presumably still inventing. It is not dead nor insane. What exception would we have to make for this type of “being” – one that can invent, but one that doesn’t have capacity to secure its inventions?

                I am trying to be practical and write a clause in section 409 of the MPEP that is consistent.

                1. Why would AI need a legal representative?… It is not dead nor insane.

                  If we want to assimilate AI into the category of “inventor,” I do not think that it would be too difficult to fit it into the category of “insane,” or even “dead.” It has many of the same features of the “insane” (lack of awareness of its surroundings, lack of interest in the ordinary affairs that occupy most living humans’ interest, ability to be manipulated easily into believing untruths, etc). You are dismissing the analogy too quickly. It works better than you are giving it credit.

                2. Put another way, if a human inventor acted the same way as an AI “inventor” does (staying all day every day in the same room, focusing monomaniacally on one single subject to the exclusion of all other activities, relentlessly pouring over data sets—even inaccurate data sets—and then making deductions from those sets to the exclusion of other data in plain sight but nevertheless unobserved), we would call that human “insane,” and act accordingly. If, then, we thought it useful to bring AI into the category of recognized inventors (a point on which I am ambivalent), we already have legal precedents as to how to deal with the problems that you note in # Those legal precedents and mechanisms would require very little adjustment to be useful for the circumstances of an AI “inventor.”

                3. I view MPEP 409 as addressing an inventor who was once sentient and capable of performing all the duties to perfect his/her invention but after the invention was made, became incapacitated for some reason. A valid patent should not be predicated upon the inventor staying alive or sane until grant.

                  In the case of AI, it never was capable of perfecting its invention in the first instance (in the sense of signing or writing an inventors affidavit or being deposed). That why I distinguished your analogy. I appreciate the banter!

                4. I view MPEP 409 as addressing an inventor who was once sentient and capable of performing all the duties to perfect his/her invention…

                  Indeed, I am sure that you are correct about that. MPEP 409 was never meant to address the circumstances of AI. My point, however, is that it would require very little change to MPEP 409 (or any other relevant rules) to make the processes of MPEP 409 work for the circumstances of AI (if we think it desirable to do so, which maybe we should not).

                  The best solution to the problem of AI inventorship would be to allow non-human juristic entities to be listed as inventors in a work-for-hire circumstance, just as we all non-human juristic entities to be listed as “authors” in work-for-hire copyrights. This would not only obviate the need to parse the relative contributions of AI in most circumstances, it would also obviate the need to quibble about whether Mike’s contributions to the team of Mike, Susan, and Jessie (who all work for NEAPCO, Inc.) amount to an inventive contribution. If it works well enough for copyright, it can work for patents as well.

                5. Greg,

                  Your approach does not solve the problem (as I refer to the Lockean nature) of the inchoate right necessarily deriving FIRST in a real human.


              Also, what is attributable to the AI need not depend on any Singularity event for the AI (all of your “do I ask” items), as the AI portion is easy to figure out by eliminating from the end product/process that which could legitimately be claimed as an invention by the human co-inventor(s).

              If no human could legitimately so claim, that part belongs to the AI.

              Note that in the DABUS case, this was the entirety of the invention.


                “If no human could legitimately so claim, that part belongs to the AI”

                Who determines this? If I own the AI and you determine that all parts of the claim were human made, I may take issue with your assessment. I would assert that AI contributed more to the claims than your human inventor. I would argue that you are “writing out” my AI’s contribution to the invention.

                Of course my rights to assert inventorship and ownership of the invention assumes that the AI has/can assign its rights over to me, which is governed by state contract law (correct me if I am wrong here).

                1. These are good questions, and ones that I do expect to see in future cases when the tires of enforcement hit the road.

                  As I said though, the impact of an entity not permitted to have a patent (based on Lockean considerations) as to other human co-inventors will be an interesting wrinkle for the courts (or perhaps Congress might wake up and take care of this before it gets that far — but I am not holding my breath).

                2. As to “assigning rights over” — not available under current law (this too is parallel to the Naruto case).

Comments are closed.