60 thoughts on “AI Patenting

  1. 8

    Could always treat the AI’s “inventing” as just a “suggestion/seed of an idea” for the human to then evaluate and then decide to “formalize” into a concrete plan of action etc. for the actual invention to be accomplished.

    1. 8.1

      Thanks 6, but in the express case of DABUS that is just not an option.

      We might as well deal with this directly, rather than trying to avoid the fact that the legal “inventor” might not be (in whole or in part) a real person human entity.

  2. 7

    AI can’t invent. Nor can it host an abstraction.

    The difference between human and non-human must, and will assume a central role in IP law and policy

    Humans invent. AI’s configure. That the owner of an effective configuration may be able to obtain IP protection for it does not seem impossible.

    1. 7.1


      Your preemptive statement – treading in a non-patent law treatment of the term of art of “invent” is of course simply wrong.

    2. 7.2

      I got a belly laugh out of the phrase “can’t host an abstraction”!

      That has to be one of the most telling statements of how all of this 101 stuff is gibberish there can be.

      CAN”T “host an abstraction”, so presumably something can host an abstraction. 🙂

  3. 6

    Let’s keep in mind Paul’s point. That what is lurking there is a big corporation providing a fancy AI tool to its employees and then claiming that the AI tool (corporation) is the inventor of everything.

    1. 6.1

      Paul’s point (while by no means invalid) simply is neither the most important nor the most immediate consideration that should be the focus of discussions of the intersection of AI and patent law.

  4. 5

    David Stein down below writes:

    The core problem here is that the definition of inventors up conflates two entirely unrelated concepts:

    (1) How the invention was made, i.e., who created it, and

    (2) How ownership rights to the patent are initially created.

    I must respectfully disagree.

    First, the “(1)” needs a touch of clarity, as “HOW” and “WHO” are themselves not to be conflated. The “HOW” is a realm regardless of the “WHO.”

    Second, while the mindless (yes, you, MaxDrei) have scoffed, there is indeed a critical — and NOT conflated — aspect of “original ownership” that is at the foundation of US Patent law.

    As I have amply pointed out, the Supreme Court case of Stanford v. Roche (still good law – see link to supremecourt.gov ) discusses the primacy of the US inventor AS initial owner. The America Invents Act expressly did not disturb that reasoning.

    Like it or not – the Lockean nature of US Patent Law cannot be ignored.

      1. 5.1.1

        The how is a realm regardless of The Who.

        Wow, how Deep.

        Substantive, clear; Wrong. No reason to not encourage the yung-ons when they finally, earnestly, try to “engage” with an actual issue. We got it, PFM.

        Unless you meant D Stein was substantive and clear.

        In which case, I agree.


            Can’t complain, Snowflake. We’ve been winding this 20 year war thing down. What have you been up to?


              LOL – are you really trying to get in on some type of credit for the debacle in Afghanistan (as if the manner of that close-out was anything but a total fiasco?) – your use of “we” here is particularly unclear.


                LOL cubed. Your pea brain thinks nobody noticed you dodged the question. Such Hubris.

                Snowflake, what did you have for breakfast?

                1. You really think your question was one that would be “dodged”…? And that any degree of hubris would be involved?


                  In that case, I am up to my usual awesome stuff.

                  As to breakfast, on Football Sundays I typically skip breakfast to treat myself to beer, wings, and pizza for the games.

  5. 4

    Several commentators suggest they don’t think any of this [allowing AI patenting with AI designated as inventor] is a hard problem. But the near impossibility of Congress making such required changes to several parts of the U.S. patent statutes is not just due to dysfunctionality. Prior attempts to allow inventor assignee filing without proper individual inventor designations and oaths have been strongly opposed by national organizations of scientists, inventors, and others. [Not even to mention those politicians mistrusting of large corporations in general.]
    Also, is the ownership of AI inventions always really that clear? What part of an AI invention is due to the AI algorithm programmer’s assignee versus the assignees of those putting in data and/or asking AI to provide a desired output, versus which part is prior art to the other assignees or unpatentable subject matter? What attorney will have a proper power of attorney if there is a conflict?

    1. 4.1

      To put it another way, re the question “What is the purpose of identifying the inventors,” what is missing from the discussion so far, and politically important, is the great importance to many inventors of having their names on patents as inventors for professional prestige, recognition, job mobility and/or advancement, egotism, etc.

      1. 4.1.1

        Paul, is there really a problem? In Europe, inventors are entitled to have their identity withheld from the public part of the PTO file and on the issued patent. Handy when the subject matter attracts the ire of animal rights activists, for example. So why can’t the default position be that the issued patent announces (as ever was) the names of the inventors, but in special cases the issued patent doesn’t?

        For your other issue I have already suggested a fix: just limit the right to challenge ownership of a patent to parties who assert that they have better title. You might then find that rival DABUS owners bring challenges but it will be a long time, won’t it, before an AI starts asserting that it (and not the Applicant or Registered Patent Proprietor) is the real owner.


          I think Paul is suggesting a problem such as a corporation providing a fancy AI tool to their researchers and then claiming all the patent rights to the corporation and saying that the researchers didn’t invent anything but were users of their AI tool.


            Thank you, Night. You (and Paul) flag up an issue that troubles (more or less) all jurisdictions, namely how to manage inventions made by employees. There is a variety of ways to do it but of course they were all devised before the advent of an inventive AI that an employer might like to name as inventor because the AI will never sue the patent-owning employer or decamp to the biggest competitor.

            Back in the 1970’s, the UK was worried that its innovation was flagging. It looked to German innovation and noted its Employees Inventions Act. It copied into the 1977 UK Patents Act the core of the German law. It was not the first to do so. Japan did it too.

            The basic idea was that if an employer doesn’t reward employee inventors, with status and money, and laud their contributions, those contributions dry up, and that, accordingly, in pursuance of the general welfare, the State should therefore encourage employers in that direction.

            How does that basic idea stand up to today’s reality, in the USA? How likely is it, that an employer will suppress the human input to technical innovation and inventorship? It doesn’t look to me like the biggest problem coming over the horizon.



              Had you a mind willing to understand, you would have noticed long ago that there is a very real cultural difference between a seemingly willingness to be a serf in a feudal system and bow to Big Corp (Europe) and the decidedly Yippee Ky Yay M___ F___ individualism and resistance to Corporatocracy that is at the foundation of the US patent system.


                “ Yippee Ky Yay M___ F___ individualism and resistance to Corporatocracy that is at the foundation of the US patent system.”

                LOL what?


              Max, I’ve received the geld for patents in Germany. It did provide quite an incentive for me as you not only get bonuses for inventing but get licensing money that can be significant.

              I fundamentally believe that it is a human rights issue that people are entitled to the fruits of their labor and thought.

              (I think it also fuels innovation too.)


                human rights issue that people are entitled to the fruits of their labor and thought.

                Sounds almost….


                (careful there Night Writer – you may lose MaxDrei on that concept)


                As in Art. 60 of the European Patent Convention then, Night? It begins with the words:

                “The right to a European patent shall belong to the inventor……….”


              Max, I know of at least one major U.S. company that voluntarily tried offering “inventor compensation” schemes for their employed inventors. It led to some fights between employees re inventor designations and suits against the company challenging their award compensation amounts.
              I have also seen cases of U.S. patent attorneys clueless to the fact that if their corporations foreign employees make an invention in a foreign country they are subject to the inventor compensation laws of the country.


                Well, yes, Paul. The classic case is perhaps, that US Corporation who “let go” the Director of R&D shortly after acquiring the company, and soon afterwards heard from the ex-Director’s lawyers on the subject of fair compensation under the Employee Inventor law of Germany. Ouch!

                Do you recall how much compensation the inventor of the “blue diode” eventually secured from the courts in Japan, under its employee inventor statute. Many millions of yen, wasn’t it?

                It is often said that the publicity generated by such cases sensitises employees to the value of the inventions they make, and thereby encouraged better laboratory notebook-keeping and invention reporting. But hey, what do I know about such subjects, never having been in house in a major innovative corporation. It was A.H.’s NS Party that invented Germany’s Employee Inventor Law. Enough said, I suppose.

    2. 4.4

      Several commentators suggest they don’t think

      Do you really think that this is surprising?

      Lots of people prefer to not see the immediate issues arising when a human person can no longer legitimately meet the existing legal definition of “inventor” (or, in other Sovereigns, “devisor”).

      Some think that they can avoid any immediate call for critical thinking by conflating (and thus confusing) separate legal concepts of inventorship and other legal functions such as providing oaths, making assignments, or even contesting inventorship.

      But none of those separate legal concerns impact a first consideration of what it means when no real human person merits the legal definition of “inventor” (in whole or in part).

  6. 3

    >>What is the purpose of identifying the inventors (for this first question?)

    I think this is an issue in terms of what of the disclosure is actually the invention. Often a disclosure includes new things and old things and the claims do as well.

    So when the inventors sign the oath they are saying that they invented what is claimed. This is important as it indicates that this is not an attempt to claim something that existed already or something that someone else invented.

    That has been one of the functions of identifying the inventor and the inventor signing the oath.

    I also think that there needs to be ownership of the AI and that it needs to be traced back to someone that is responsible for whatever the AI does.

    Given these, I don’t think this is a hard problem. The AI can be attributed as the inventor with a different oath of ownership that states exactly what is being claimed as new and the invention.

    I don’t think any of this is a hard problem. I do think that AI is going to invent more and more things. I think “The Poopy Diaper” has no idea what he is talking about when he says that AI is some fixed programmed computer. The Church-Turing Thesis tells us that if a person can do it, then a computer can potentially do it too. The theory tells us that AI can invent whether or not current system invent or not.

    Anyway….all these problems are easy to fix as stated above if the Congress was actually a functioning body capable of fixing problems.

    1. 3.1

      and that it needs to be traced back to someone that is responsible for whatever the AI does

      And this is the sticking point, Night Writer.

      The point being advanced is that the “real human” person that IS responsible is plainly stating “I did not DO the actual thing to which the legal term of inventor would apply.”

      That person is explicitly stating that “responsible for” is just not enough to rise to “inventor,” and it would be a
      for that person to claim otherwise.

      (keep in mind as well that there are other legal considerations – directly of patent law nature – in play as well, not the least of which has to do with the NON-HUMAN juristic person of Person Having Ordinary Skill In The Art).

      Below I have pushed for Malcolm to actually explain his views on a 112 comment. I suspect that there may be overlap between his comment and the possible impact to 103 FOR the possible expanded consideration of an AI machine being taken into account for obviousness determinations when one considers that (as the argument in DABUS goes), AI actually invented something that a human real person did not.

      This ALSO impacts (especially) that “grown-up” art that Malcolm prefers to showcase in his example of AI “sifting through” millions or even billions of possibilities and the machine (without a responsible real human person) comes up with “the answer.”

      This is indeed more than Malcolm’s quip on 112 (whatever it was that he meant by that quip).

  7. 2

    The core problem here is that the definition of inventors up conflates two entirely unrelated concepts:

    (1) How the invention was made, i.e., who created it, and

    (2) How ownership rights to the patent are initially created.

    For people in the “AI cannot be an inventor” camp (including myself), the first issue is a problem. I don’t particularly mean DABUS – I remain deeply skeptical about both the model and its output, based on the scant information released to date. Rather, consider pharmaceuticals in which a machine learning model grinds through a simulation of billions of possible tweaks on a known drug and identifies one with improved effectiveness or reduced side-effects. With $HUGE on the line for the ensuing patent, it is extremely important to get inventorship right – and yet, who is the actual inventor here?

    The question, as I see it, is this: What is the purpose of identifying the inventors (for this first question?) Is it simply to identify the people involved? In the case of pharmaceutical simulations, the right answer may be: no one, and we can dispense with this part of the inventorship requirement. Or is the objective to document the inventive process? The right answer may be: the simulation algorithms. Either way, confining this answer to the names of the people involved is not a rational choice.

    But for the “AI should be able to be an inventor” camp, the second issue is an absolute dealbreaker. Algorithms legally owning property rights is an absurd concept with innumerable obvious issues. One option is to dispense with mandatory inventor-originated ownership altogether, and to just allow the applicant to indicate the original owners (persons or corporations) in the ADS.

    The upshot is that this general problem of “can AI be an inventor” will remain intractable until these two entirely distinct but conflated concepts can be untangled.

    1. 2.1

      David, I agree (of course) but wonder whether the problem with naming an inventor is harder in the USA than elsewhere, because of the invention=conception mindset, a hang-over from your “First to Invent” patent law history. After all, in all other countries it hardly matters what names are given to the PTO as inventors. What does matter, 100%, is (of course) ownership.

      With the arrival of First to File, nobody needs to enquire any more into “conception” or “diligence”. The ONLY thing that matters is the information content of the patent specification filed at the PTO. That document is i) commissioned by the owner and ii) written by a patent lawyer. The “invention” is what that lawyer says it is.

      In other countries, patents issue with no mention of who the inventor was.

      In other countries, the ONLY person who can dispute the validity or enforceability of the granted patent on the ground that it was granted to the wrong person, a, is B, namely a person who asserts that they, and not A, is the true owner of the invention. When an aggrieved DABUS itself makes such an assertion, then I will take seriously the issue whether to name an AI as an inventor. But not before then. For the time being, if there is a dispute whether DABUS (A) or DABUS (B) made the invention, I will suppose that such a dispute will be between the respective owners of DABUS(A) and DABUS(B).

      1. 2.1.1

        If this is mostly a quirk of U.S. law, why have the EPO and the U.K. courts come to the same conclusion as the U.S.? This seems to be a problem for pretty much everyone’s legal systems, not just ours.

      2. 2.1.2

        “ naming an inventor is harder in the USA than elsewhere, because of the invention=conception mindset”

        I don’t know if really that much harder but computers don’t “conceive” of any thing. They are machines that process data according to instructions.

        The conception of an “invention” that was spot out by a programmed computer occurs when a human person conceives, in a fixed and permanent manner, that the computer has not screwed up and outputted something worthless.


          Adding: consider an application filed with 10 million short RNA sequences cranked out by a computer. The sequences cover all possible sequence permutations that were “calculated” to have efficacy against Disease Y when passaged into a cell. The sequences are ranked according to predicted efficacy. But the application has no data showing actual efficacy of anything.

          That application is worthless (or should be) because nobody has conceived of anything. It’s a research plan, at best.

          What inevitably happens later is that the scientist will test the top candidates and discover that some work well (usually the ones that would be deemed most likely to work even without using a computer) and some barely work at all.

          By the way, this is an actual strategy used by companies whose names you would recognize but it gets even better: tons of applications (hundreds, at least) each focusing on different candidates (but without data!) are filed as separate provisionals with the hope that one of those applications discloses the magic. After a year of experimentation on the candidates, the “winning” provisional will be converted and the others will be abandoned so nobody recognizes the obvious 112 problem with the filing date.


            Genuine question for clarity:

            What do you mean by: “so nobody recognizes the obvious 112 problem with the filing date.”…?


              I understood what he meant as that each of the multiple provisional applications only claims a narrow species, not a generic claim covering large numbers of species and thus triggering an obvious 112 problem.


                And I would say that, however “obvious” it is under the jurisprudence of the USA, the “112 problem” is not a problem at the EPO. That’s because of the rise to dominance in Europe of the “plausibility” line of case law. Is there anything comparable in the USA, I wonder?


                Thanks Paul – I am aware that such is one possible interpretation (there are others), but I am looking to understand what Malcolm himself actually means with his invocation of ‘frailty” under 35 USC 112.


                As I figured, Malcolm is only interested in drive-by ad hominem rather than any actual discussion.


          Malcolm, not only the invention = conception mindset is a problem. A further aspect, peculiar to the USA, is the quacking from the sidelines about the “critical starting points” of the Locke-ian underpinnings to the full basket of rights under US patent law and the blather about certain inchoate Rights of Man, spelt MAN, from diverse legal theorists trying to attract attention. It’s a problem.


            As noted below at: link to patentlyo.com

            Leave it to Greg (and MaxDrei) to refuse to recognize the fundamental aspects of the US Sovereign origins

            I am reminded of the pile of MaxDrei horses decaying at the well of wisdom, for all of MaxDrei’s lament of “a mind willing to understand,” he would rather die of thirst than actually understand the US sovereign.

      3. 2.1.3

        David Stein: “ When an aggrieved DABUS itself makes such an assertion, then I will take seriously the issue whether to name an AI as an inventor.”


    2. 2.2

      Agree 100%, David. In Title 17, the entity that commissions a “work for hire” is the legal “author” of any copyrighted works that emerge from the commission, such that ownership vests originally in the commissioning entity (usually an LLC, Corp., etc.), not in the hireling artists. The easy solution here would be to amend to amend Title 35 along the same lines, so that the legal “inventor” is the entity that commissioned the R&D.

      Why should anyone care about tracing out exactly which employees of X Corp. contributed to the conception of an invention, so long as it is clear that all of them would assign to X Corp. in the end?

      1. 2.2.1

        “ The easy solution here would be to amend to amend Title 35 along the same lines, so that the legal “inventor” is the entity that commissioned the R&D.”

        Why is that easier than just saying “Of course computers can’t be inventors, you nutty freak?”

      2. 2.2.2

        Careful there Greg – Malcolm may place you in that same ‘one bucket’ that he has for everyone that he thinks is “wrong” (and does not FEEL how he feels).

      3. 2.2.3

        Why should anyone care about…

        Asked and answered — dating back to the discussion of the Stanford v. Roche case.

        Leave it to Greg (and MaxDrei) to refuse to recognize the fundamental aspects of the US Sovereign origins of patent law and the philosophical underpinnings of the Lockean nature of MAN’s (that is, real person and NOT juristic person) inchoate right that is the critical starting point of a patent (that inchoate right turned into a full basket of legal property rights).

    3. 2.3

      “What is the purpose of identifying the inventors”

      Besides preventing theft of intellectual property from the rightful inventors/owners?

    1. 1.1

      This really is the core catch-22 in Thaler’s argument: either DABUS is not a natural person capable of conceiving an invention, in which case it is not valid for it to be named as an inventor; or DABUS is a natural person capable of conceiving an invention, in which case Thaler has failed to demonstrate either (1) that he is in a position to vindicate DABUS’s rights to apply for a patent or (2) that DABUS has actually assigned its rights in the invention to Thaler.

      He wants DABUS to be a natural person without conferring the rights that a natural person would normally have. Going down that road would delve into some pretty science fictionesque concerns.

      1. 1.1.1

        There is no “catch-22” . Computers aren’t people and any argument to the contrary is stoopit.

        Given that there is no shortage of inventions or patents, it is beyond weird to want expand the system in this way, particularly when claims relating to computer instructions written by humans are already highly suspect or meritless.


          Feel free (me extending this invitation again) to remove ANY and ALL such technology from your life, Malcolm.


            You are a very silly person if you think your comment makes you look clever, informed, or anything other than an intellectually bankrupt hack.


                Malcolm really is a complete looney. Does anyone know him in real life, and is he such a ratmuncher in person?

Comments are closed.