USPTO Rejects AI-Invention for Lack of a Human Inventor

by Dennis Crouch

Sometimes I think of myself as the creativity machine.  A cool part of this system is that I have a right to seek and obtain patent protection for my inventions (if any).  The USPTO is treating Mr. Dabus differently.  When Dabus filed for patent protection in 2019, the examiner refused to examine the patent and the PTO Commissioners Office has confirmed the refusal.

The problem is that Mr. Dabus (DABUS) is not human, but rather is a machine – a creativity machine. 

App’n No. 16/524,350 was filed listing DABUS as inventor and identifying DABUS as an “artificial intelligence” that “autonomously generated” the invention.  Stephen Thaler created DABUS, then DABUS created the invention.  Thaler then filed as the applicant.

In briefing to the PTO, the patent applicant explained that DABUS conceived of the idea of the invention and recognized its “novelty and salience.” In short, DABUS did everything necessary to be listed as an inventor with one exception — DABUS is not a human person.

The Patent Act does not expressly limit inventorship rights to humans, but does suggest that each inventor must have a name, and be an “individual.”

(f) The term “inventor” means the individual or, if a joint invention, the individuals collectively who invented or discovered the subject matter of the invention.

35 U.S.C. 100(f). In denying the DABUS petition, PTO Commissioner’s Office suggests that the word “Whoever” in Section 101 indicates a human “natural person.”  (citing Webster’s 2011).  Of course, Section 271 also uses “whoever” to define infringement — and human “natural persons” are almost never the ones charged with infringement.  The Federal Courts have repeatedly held that we need to find the “individuals” who are the inventors, and not simply point to a collective organization like a corporation or government entity.  However, the PTO could not point to precedent where the claimed inventor is a thinking non-human machine.

Note here that over the past few years the Federal Circuit has repeatedly equated human mental processes with processes going on inside of a computer (as well as inside a fish).

Although Thaler created DABUS, he feels that he cannot properly name himself as inventor.  The PTO appears to suggest the resulting answer — no human inventor; no patent.  It will be interesting to see whether this form of identity politics stands the test of time.

103 thoughts on “USPTO Rejects AI-Invention for Lack of a Human Inventor

  1. 13

    I wonder what is the aim of this ( DABUS) patent application ? Is it just to open a path for AI inventor-ship discussion ? What will change if we appreciate AI for its role in invention process ?

    1. 13.1

      What will change if we appreciate AI for its role in invention process ?

      A fair question – with plenty of potential side effects.

      As it currently stands (for the US Sovereign), An inventor cannot be an AI, while an AI may be the entity that completes the SUB-element of invention of the conception (the UK-equivalent of a Devisor).

      This has the immediate legal impact that actual putative inventions may be being made that can claim NO legal inventor.

      To claim a legal inventor, ALL parts of the legal term must be met, including the legal term of conceiver as a necessary sub-part to the legal term of inventor.

      One potential (and potentially huge) aspect of this IF conception by AI is to be permitted to fold further into and subsume the legal definition of inventor is that the US Sovereign would face an existential question of basis of law. US Sovereign law is a law of express LIMITED powers, both at the Federal level, and at the State level. The Rule of Law provides that power not expressly provided, resolves back down the societal scale and rests with the human person. These reflect back up through the structures that we have put in place (that MUST reflect the inalienable rights of the individual REAL persons.

      A second – but no doubt as impactful wrinkle must unfold as to what it would mean for a Person Having Ordinary Skill In The Art to be imbued with AI capabilities. As you may recall, this legal fiction is NOT based on any actual capabilities of any one real person, but instead is driven by the notions of what would be obvious if the LEGAL person – who has access to ALL printed material and the like (with other constraints, as noted in the Graham Factors) are at play.

      All of a sudden, you are faced with that legal person ‘in the art,’ being able to independently conceive — currently — and actually invent — if AI path unfolds.

      You will have a catch-22 situation in which ‘invention’ may be deemed obvious because non-humans can invent.

  2. 12

    Summary Judgment against Night Writer is in order here, as he has ‘rested his case’ without bothering to actually address the counter points that wreck his desired narrative.

    However, the story has not finished, and at link to it continues (without Night Writer’s participation).

  3. 11

    Night Writer,

    Are you “resting your case” without answering 10.2?

    That would necessitate a default judgment against you.

    As ever, your choice to respond or not — and as ever, NOT your choice of the consequence of not responding.

    1. 11.1

      anon, if you want to try to narrow your points of contention with me as I did in 10, then I will respond. But your posts are not framed in a legal terms and do not address my points in a legal fashion.

      I think my post at 10 sums up the issues as they sit. These are the issues that will have to be settled in US law. I think Greg’s post is good too that copyright for hire is an interesting take on my points (1) and (2).

      1. 11.1.1

        Plus, anon, the reality is that my points (1) and (2) are unsettled law. The courts can go either way on them. But those are the issues.


          Those are the obfuscated issues to which counter points have been presented.

          You do not get to choose to be less than honest on this point.

      2. 11.1.2

        More B$ from you, as I have repeatedly shown that you have not provided the narrow view (conception), but instead have conflated a larger view (invention).

        Your ‘cry’ for “in a legal fashion” is not even something YOU will do, as you do not engage at the level that you would demand of me (with legal citation).

        Your posts do NOT sum up the issues – expressly because my counter points show that they do not.

        Those counter points remain for you to engage.

        You may choose not to engage, but you do not get to choose the narrative by not engaging.

      3. 11.1.3

        I think Greg’s post is good too that copyright for hire is an interesting take on my points (1) and (2).”

        Wow, are you full of yourself.

        Greg’s point vis a vis copyright for hire has ZERO to do with being a take on your points (1) and (2) as Greg’s post predates ALL of your ill-aimed rambling and obfuscation of trying to make this about the larger picture of “invent” than about the more precise picture of “conceive.”

        The narrative that you want will not happen until (and if) you actually engage on the counter points to that attempted narrative.

        Every time that you attempt this wayward narrative of yours, these counter points await.

        Why are you so resistant on discussing the more precise point of “conceive?”

        Why are you so resistant on seeing the necessary problems with your version that whoever happens to read and understand the conception of another is not (and cannot be) the conceiver?

        These two simple and direct questions refute each of your ‘beloved’ points of (1) and (2).

        Default judgment against you stands.

  4. 10

    There are two issues here.

    (1) Was it actually invented when the machine put together a combination of elements.

    (2) Did the person that first saw the combination of elements invent the invention?

    By the legal definition of invent in the USA, the conclusion in (1) is that the machine did not invent the invention.

    And, for (2), we can say that the first human brain conceived of the invention, so they did invent it.

    For (1), we can’t say that both the machine invented but is not eligible for a patent because it is not a person as the definition of invent is defined in words that apply to human brains and not machines.

    For (2), the patent statute explicitly says that the manner of invention does not matter. So if I invent by using a machine to figure out a bunch of stuff and I pick what I want as an invention, then that is inventing. I am the first human brain to conceive of the invention.

    There is really no other consistent opinion.

    1. 10.1

      Probably what is most interesting about this is words like “conceive”.

      The CAFC is stacked with SV selected and Obama appointed judges that are anti-patent judicial activists and, in the main, have no science/technology/innovation background.

      So what the hacks at the CAFC have been doing is using words that apply to a human brain to apply to an information processing machine. And now this forces the issue. Does a machine “conceive”? Please answer this Taranto and gang.

      1. 10.1.1

        Conceive has a legal definition that does not require a human brain.

        You have admitted this point – and yet refuse to incorporate that admission.

        BE the attorney here, Night Writer – not a whiner.


            I do not need a case cite for the legal definition – it is the definition that we both have discussed, and that you have admitted to.

            The cite for that admission is: link to and your reply immediately following.

            As I have noted — YOU have bailed on that point, and instead seek instead to have a vector from “invention” instead. But I have shown that this is a fallacy because invention IS barred to AI separately, and that conception is still reached — legally — by AI.

            As I state: I have shown the error in your tactic. BE the attorney and do not disregard what has been shown to you, and what you have admitted.

            Instead, DO answer the specific counter points put to you.



              So there is no legal cite, but you claim the legal definition includes a computer.



                So there is no legal cite, but you claim the legal definition includes a computer.


                Let me play that same game: where is your citation that the legal definition of conception must be in the human mind?

                1. anon, the reason it is a point of contention whether “conceive” includes an information processing machine is that there is no definitive legal cite.

                  My second point is also still legally an open point.

                  The arguments are squarely with those two points of contention. If you disagree, tell me the points of contention.

                2. You ask this question, even as I have already answered and put to you counterpoints that YOU have steadfastly refused to engage with.

                  The ball is in your court. I suggest that you answer the counter points out to you.

                  8.2 still remains unanswered.

                3. anon, tell me concisely in a few sentences what your counter points are.

                  Your post at 8.2 is not concise and it is not even clear what you are saying.

                  I have boiled the issues down to two points that I think encapsulate the issues.

                4. I have replied to this request already.

                  Among numerous other posts now: link to

                  Your (1) is disingenuous as you enlarge/obfuscate rather than be precise/clarify as to what the issue is (invention rather than conception).

                  I have shown how the larger invention is still denied AI while conception occurs.

                  Your (2) is in error – and multiply so – based on the detailed thought experiment provided.

                  WALK THROUGH THAT EXAMPLE.

                  ALL that you are doing is dodging and running away.



              anon, I have narrowed the issues. I have concisely stated the points of contention. I do not concede your definition of “conceive” as including information processing machines. And you apparently have no legal cite to support your contention.

              Being a lawyer anon is learning to narrow issues to the points of contention.


                You have NOT narrowed the issues — and I have expressly shown you how.

                Instead, you have BROADENED the issue and attempted to obfuscate by focusing on “invention” instead of the actually more narrow “conception.”

                I have shown this to you now several times.

                Being a lawyer is NOT making this type of mistake just to try to avoid the particulars put to you.

                I notice that you STILL avoid those particulars. Post 10.2 needs your direct and on point attention.


                8.2 is full of conclusions anon.

                Please try to be concise and come up with points of contention like I did. You have not refuted (1) nor (2) with legal cites.

                (1) and (2) remain open questions.

                1. 8.2 is full of conclusions anon.

                  Utter B$.

                  Try walking through the steps and FREELY note your responses.

                  If you want to see “full of conclusions,” visit your own diatribe.

                  I have rebutted BOTH your (1) and (2).

                  The ball is in your court.

                  Do more than just mouth nothingness – ANSWER THE POINTS PUT TO YOU.

                  Again – if you dare.

                  (to this point, across at least four threads, you have dared NOT).

                  PS: as noted in this thread at link to you yourself have
                  to do what you would demand of me. Your requests for citations are NOT needed for you to engage.

                  It is beyond clear now that if you refuse to engage, you are nothing but a coward and a blowhard about all of that “I taught AI” stuff you spouted.

                  The choice to answer or not is yours.
                  The take-away of that choice is not yours.

                2. For those following along, due to Count Filter limitations, the discussion has overflowed across four threads.

                  The “8.2” at point here has actually ‘jumped’ in numbers to 10.2.

                  The comments at 8.2 are directed to Greg “I Use My Real Name” DeLassus in regards to a possible path that mirrors the copyright law path that bowed to the corporate juristic person.


              And going backwards to something I may have said in a discussion is meaningless. I have stated my position clearly and concisely.


                … and I have debunked your stated position, clearly, concisely and thoroughly.

                YOU have yet to answer to my counter points, and instead insist on repeated your debunked position as if no counter points have been presented.

                As I said previously, do NOT be a Malcolm.

                1. And I am going to take your cowardice as a concession.

                  Now what?

                  Here’s a hint: ENGAGE.

    2. 10.2

      You keep on wanting YOUR narrative, while avoiding all of the simple and direct counterpoints put to you, Night Writer.


      Let’s – yet again – deconstruct your attempted narrative:

      Your (1) of “Was it actually invented when the machine put together a combination of elements.

      For (1), we can’t say that both the machine invented but is not eligible for a patent because it is not a person as the definition of invent is defined in words that apply to human brains and not machines.

      You have confused and conflated the different legal definitions of “invent” and “conceive.”

      I pointed out to you that you are attempting to take a vector from the top down ‘invent’ instead of from the bottom up ‘conceive’ because you simply do not like what the legal definition of conceive provides.

      You have admitted as such on these boards that AI can legally be the conceiver. You then run away every time I try to bring you back to this point.

      You confuse and conflate the FACT that my argument – legally based on conceiver first, ALSO holds that under US Sovereign law, that AI cannot be an inventor.

      More than just the US case, I have provided you a path with the UK version of the case (wherein the court there held AI to be the equivalent of conceiver in their notion of devisor, and yet held ‘no patent for you.’

      You ran away from that too.

      You yammer about wanting me to be ‘an attorney,’ and yet YOU refuse to be one, and instead ONLY repeat a position to which counter points have been presented. This is just like your most odd “Trump will be a Savior” based only on your feelings ‘drumming up’ attempts.

      Let’s turn to your other thrust:

      (2) Did the person that first saw the combination of elements invent the invention?

      And immediately, your logical and legal fallacy presents itself. Your error of vector choice (from inventor to conceiver, instead of from conceiver to inventor show the error: “By the legal definition of invent in the USA, the conclusion in (1) is that the machine did not invent the invention.
      As I have already pointed out to you, I reach the exact same spot. Thus, your conclusion does NOT arrive from your premise. Instead, you have run logic circles and merely proclaim your conclusion from nothing more than your assertion OF that conclusion.

      As to “And, for (2), we can say that the first human brain conceived of the invention, so they did invent it.


      For (2), the patent statute explicitly says that the manner of invention does not matter. So if I invent by using a machine to figure out a bunch of stuff and I pick what I want as an invention, then that is inventing. I am the first human brain to conceive of the invention.

      You have now conflated 103 which deals with a different aspect of invention – top down perspective and an overall view of invention – while STILL not dealing with the underlying requirement and legal definition of conceiver, from the bottom up perspective.

      I have deconstructed your mere human reading the output of AI’s conceptions several times now – and like EVERY all element of this discussion, you have NOT engaged ANY of those counter points put to you (across SEVERAL THREADS).

      So let’s modify the point put to you a little bit more and see if you will engage ON POINT or merely run away (again) after claiming that you have engaged on point….

      In addition to post 1.3 link to I had added at post an additional ‘mere reader’ conundrum for you:

      As to ‘first time’ – see above at 1.3 (this is now the second time that you are skipping over counter points provided to you). You may want to slow down and contemplate those points before you insist on maintaining your pre-conceived position.

      A further note on that point: mere recognition of someone else’s work is NOT viewed as invention OF that work. You have a serious logistical problem easily shown by a simple example.

      You would have ‘invention’ be by a person sitting in a room when that person merely opens a box shipped to him and reads something in the box.

      Note that the antecedent of what that item in the box is, and how that was developed is left out.

      To your view, THAT is not material.

      But it IS material. Visit again the posts and the items that I have presented to you that you have skipped over.

      You have moved from an assumption that the builder or owner is the first one (merely) reading the output of the AI machine and thus necessarily being the inventor, to eliminating the requirement of conception AT THE TIME that the idea is formed to some later time of merely reading that formed idea (formed prior to the box being shipped and opened by whomever happens to be receiving the box.

      Your assertion of “The key point here is that according to the legal definition a machine cannot invent but only a human brain.” is the conclusion that you are presuming. This entirely misses the underlying point that NO HUMAN has done the actual conception performed by the AI.

      What happens in our little thought experiment if we exchange the single person sitting in a room opening a box and the entire world sitting at a computer screen reading the contents of the box being presented en masse to everyone? Does your attempted logic make everyone the inventor?

      Now if you want to go down a different path of Discovery, well, we can do that. But YOU need to realize that such is indeed a different path.


      To this (unanswered) point, let’s add three additional rooms. In one of these additional rooms, there is a separate REAL person doing the conceiving, and writing that conception down, let’s say in a patent application format. That patent application is put into a black box. In the other additional room, there is an AI doing the conceiving, and with automatic transcription (of which, this capability exists today), that conception is ALSO transcribed into a patent application format. This other patent application is put into a separate black box. Each of these two black boxes are put into the third room. In that third room, there is a randomizer that takes all the black boxes given to it, and completely randomly selects a single black box for presentation to the aforementioned gentlemen – who for argument’s sake let say has nothing to do with building or owning the AI machine that was in the second room.

      Now let’s return to YOUR supposition that a real person opening that black box and merely reading what is in the box MUST be deemed the inventor.

      Do you see the logical problem?

      Please give a direct and clear answer to this question. DO NOT try to merely restate your narrative. ANSWER THIS QUESITON!

      And then let’s return to my previous counter point which you have not answered, and replace that nondescript person merely reading what is in the black box with millions of people simultaneously. Are EACH of these millions of people who are merely recognizing what has been captured on paper (randomly and may or may not be the AI captured item) “inventors?”

      We can ONLY be consistent here (legally) by recognizing the flaw in your position. The flaw that comes from your attempted narrative and vector treatment from “inventor to conceiver,” rather than MY treatment of “conceiver first, and then inventor separately.”

      Finally, your exasperated edict of “There is really no other consistent opinion.” shows that you WANT a certain Ends, and quite frankly have abandoned a proper legal path, or Means to those Ends.

      Consistency? What exactly is the driver for your ‘consistency?’ That innovation IS met with a patent? But YOU only get there because you forsake exactly who the conceiver actually is – legally and factually.

  5. 9

    I am sure I understand AI just fine. I have taught AI classes at the university level.

    The fact is that the AI is a machine that was made by one or more people. The product of that machine can be attributed to the person that built that machine.

    It may be at some time in the future that it gets harder to attribute the machine to a person if the machines start building factories and making better machines themselves, but right now it is not hard to figure out who made the machine.

    There are other issues that AI brings up such as what is obvious in the life sciences when the machine can test trillions of possible combinations and determine which to use. But attribution is –right now–not a problem. We know who built that machine.

    This right now is a silly issue. I think the same people that can’t understand this are the same ones that tend not to understand information processing and that it is a physical process performed with machines.

    1. 9.1

      The product of that machine can be attributed to the person that built that machine.

      You obviously need to brush up and catch up.


          It is quite clear that the one that doesn’t get it is the one refusing to answer questions and engage.

          In case you need help with that, that would be you.

          Lot’s of energy from you, all directed to running away instead of engaging.

          It’s easy to see why you do not want to engage – your desired conclusion cannot be reached.

          Here’s a hint: you will not reach that desired conclusion without BOTH engaging and overcoming the counter points put to you.

          Those counter points simply will not go away.

    2. 9.2

      but right now it is not hard to figure out who made the machine.

      Yep, you have fallen behind.

      1. 9.2.1

        anon, your sentences carry no meaning.

        And no I haven’t fallen behind. Right now I work with some of the top people in the world in AI and write patent applications for them.

        You just making assertions that I don’t know what I am talking about is meaningless. I framed the problem. You did not address what I said.


          It is less that you do not know what you are talking about and more that you have not kept up here.

          A post was nabbed in the count filter discussing this. Hopefully it releases.

          In a nutshell, conception requires:
          An invention is conceived when a definite and permanent idea of an operative invention is known.

          The point of who owns the AI machine or who built the AI machine is NOT the point here.

          That person (or even group of persons) does NOT have that “definite and permanent idea of the operative invention” that is the RESULT of today’s currently massive INTERNAL AI operations.

          It is your framing of the problem that is your problem. You are not framing the problem correctly to that actual underlying point here.

          That you do not understand this – that you do not understand that I HAVE addressed what you said is merely MORE of your problem.

          You do not know even THAT you do not know.

    3. 9.3

      When you think about it deeply –and really spend some time thinking about it and don’t just yap out–then you will realize that is going on here is anthropomorphizing in that because the machine is doing more than just doing something like making a donut that we are starting to think of it as greater than a machine or having attributes like us.

      This is medieval thinking. It is a machine. Someone built that machine. The person that built that machine is the inventor. When the machines start building themselves and we no longer control or build the machines, then we will have to rethink this. But right now this is silly.

      1. 9.3.1

        I am WELL past you on the whole anthropomorphizing realm.

        I was there SEVERAL years ago.

        What you call ‘just yapping’ is only you not realizing that I am far ahead of you here.

        And you are absolutely LEGALLY wrong on your assertion that the builder of the machine must be the inventor. Again – revisit the requirement of conception. Inventing the AI machine is NOT inventing WHATEVER comes out of the AI machine. This is not just a ‘all things must fall to the first order’ when the POINT is what is happening at a step removed from that first order. You want to talk about medieval thinking — look at what you are asserting.

        You fall into the very trap that you warn others about with wanting a state of a machine building another machine. How medieval of you to need something to drop on your toes.

        No Night Writer, YOU are the one that needs to pause and NOT just ‘yap back without thinking.’ Here’s a hint: stop taking this as a personal attack. Your ego is getting in the way of your understanding.

    4. 9.4

      Inventor (A) builds machine (x) that generates invention (I).

      This is transitive right now because machine (x) did not build machine (y) and then machine (y) invented invention (I2).

      1. 9.4.1


        needing something to drop on your toe in order to preserve this type of ‘transitive’ is both a legal and logical error.

        An inventor that builds an AI machine invents the AI machine.

        But it EXPRESSLY is NOT transitive to insist that the inventor of the machine “invents” WHATEVER comes out of the AI machine.

        Again, look at the legal requirement of conception. THAT is not — and can not — be as transitive as you feel.

  6. 8

    The US copyright act provides (17 U.S.C. §201(b)) that the “author” of a work made for hire is the one who commissioned the work. This is true even if the one commissioning is not a human person.

    Much trouble—in this and other ends of patent law—could be spared if Title 35 were amended to provide likewise for the “inventorship” of inventions made for hire. We would not even have to furrow our brows under such a regime when Dabus LLC gets listed as “inventor.”

    1. 8.2

      …but overall, I agree with you that THIS avenue may well be worth exploring.

      Note however, that the parallel cases in the UK and the EPO defeated this type of approach, and would NOT permit a juristic person to lay claim to any type of rightful possession as applicant to advance prosecution.

      You STILL have an ‘original entity volition’ problem. In the copyright sense that you provide, the juristic person is STILL working with a real person, and that real person is STILL the originator who is making a volitional choice to enter into an agreement that transfers the ‘authorship’ TO the juristic person.

      With AI, you cannot have this original volition recognized.

      I do not recall the cases on point, but I seem to remember a couple of cases that refused to provide copyright in the first instance to ‘creative works’ entirely driven by computers.

      These cases also point to an ‘original real person’ requirement that the authorship transfer mechanism may well fall short of rectifying.

      (note as well, that someone else may need to repost these comments if Greg is still running his ‘my feelings were hurt’ block on seeing my comments)

  7. 7

    Hmmmm . . . wonder if the good Mr. DAUBUS would be willing to provide some artificial intelligence to our esteemed “drink bleach to kill the virus” pres . . .

    Clearly, the face covering he needs isn’t a mask, but a muzzle.


          Lol – no prob.

          What gets me though is how little on point discussion there is on several legal topics.

          This is supposedly a patent LAW blog, after all…


            “This is supposedly a patent LAW blog, after all…”

            True dat. Yet, a little whimsy and levity here and there (think those funny fishing comments of a few days ago) is also good for the . . . legal soul. 🙂

                1. Oh, I do not doubt that anyone can decide that anything may be deemed “levity” to that particular person.

                  But that kind of opens up the blog to the type of blight that has been LARGELY absent since Malcolm took his extended leave.

                  Is that the type of company that you wish to keep?

  8. 6

    In my opinion this is answered by § 103: “Patentability shall not be negated by the manner in which the invention was made.” If a human induces a machine to produce a patentable invention, then the human is the inventor.

    Consider, for example, drug discovery via artificial intelligence, such as Exscientia’s platform, which is used by Celgene, Evotec, GlaxoSmithKline, Roche, and Sanofi:

    link to

    I’m pretty sure those companies are confident they can tie the new molecules discovered this way to a human inventor, even though, according to Exscientia’s CEO, “The entire decision-making and design process is driven by algorithms.”

    1. 6.1

      There is a nuance here, James: it is NOT “patentability” that is being negated, so your ‘answer’ of 103 is not in play.

      You want to advance that patentability is not negated.
      No problem – I spot you that (much like in fact, the UK court holding that the AI was in fact the devisor).

      The mere presence of a patentable invention though does NOT move you any closer to obtaining a patent for that patentable invention.

      Leastwise in the US Sovereign.

      It is the fact that the patentable advance does NOT have a direct (proper, under the controlling US Sovereign law) human inventor.

      I would be interested in the ‘tying’ legal argument that you think would float. When you start considering what that may be, you are likely to confront the nuance that escapes your post here.

      What you end up with is a patentable invention that can NOT accrue a patent to any human person in the first instance.

      Lacking that first instance, there is NO juristic person (or other real person) that can then ‘take title.’

      The controlling case law here is Roche v Stanford.

      You are missing the very first (and necessary) link in the chain. Without that link, there is NO tying possible.

      And this is directly on point to the admissions of Thaler (in this case), and to Exscientia’s CEO (albeit, I am not familiar with those statements, or if they are tied to any binding statements in a patent case).

    2. 6.2

      There is even other nuances here that may come into the spotlight…

      For example, what does Person Having Ordinary Skill In The Art become when you have non-human “invention” possible?

      Does ALL such non-human work make ANY further steps to be obvious?

      What limit is there after all, when you say that anything that a computer MAY — by itself, without human invention — is outside of human invention? Is a ‘current state’ of an AI that which only has been made explicit? Or is it the inevitable output of ANY existing inputs?

      Of course, this particular nuance may well turn on the position that there is NO invention in the first place for an AI (and I grant that this is NOT in accord with your expressed position).

      So just based on this particular wrinkle, I would posit an agreement with you that an actual invention must be deemed to have taken place. Otherwise, PHOSITA has gained AI powers and while not “Flash of Genius,” such an augmentation pretty much ends the patent game (for ANY art field to which AI could be directed).

  9. 5

    35 USC 115 requires “(a) Naming the Inventor; Inventor’s Oath or Declaration- … Except as otherwise provided in this section, each individual who is the inventor or a joint inventor of a claimed invention in an application for patent shall execute an oath or declaration in connection with the application.
    (b) Required Statements- An oath or declaration under subsection (a) shall contain statements that–
    (1) the application was made or was authorized to be made by the affiant or declarant; and
    (2) such individual believes himself or herself to be the original inventor or an original joint inventor of a claimed invention in the application.
    The only “Except as otherwise provided in this section” exceptions are in:
    (d) Substitute Statement-
    (1) IN GENERAL- In lieu of executing an oath or declaration under subsection (a), the applicant for patent may provide a substitute statement under the circumstances described in paragraph (2) and such additional circumstances that the Director may specify by regulation.
    (2) PERMITTED CIRCUMSTANCES- A substitute statement under paragraph (1) is permitted with respect to any individual who–
    (A) is unable to file the oath or declaration under subsection (a) because the individual–
    (i) is deceased;
    (ii) is under legal incapacity; or
    (iii) cannot be found or reached after diligent effort; or
    (B) is under an obligation to assign the invention but has refused to make the oath or declaration required under subsection (a).”

    Which of these above-listed permitted exceptions could apply to a “named” AI program? Could “is under legal incapacity” apply?

    1. 5.1

      I do not think that “under legal incapacity” is the same as “never having legal capacity.” See Naruto v Slater link to

      You may have a better argument under ‘obligation to assign,’ given that the machine is owned by someone.

      1. 5.1.1

        Yes, I am sure no AI program has ever been adjudged to have every had any “legal capacity,” so had none to lose.
        I thought about “obligation to assign” as well, but having had one such case for in a reissue I know that the PTO enforces the rest of that exception, which is “but has refused to make the oath or declaration.” That is, it would require proof that the AI had been asked if it would make an oath or declaration of inventorship AND had “refused” [refusal being considered an affirmative act]. Of course one could probably add software to give that kind of Hal like refusal response?

  10. 4

    Congratulations to Stephen Thaler for creating of the first complex non-biological system, which,

    rather than

    simply algorithmically simulating the superficial and outward appearance of consciousness and rather than, through blind shuffling of some paltry bits of information mimicking what a sentient mind IS and DOES by communicating some kind of output which remind us of what an experiential sentience produces and self-reports…

    it actually IS conscious and sentient, being and doing all that a mind is and does when it actually is conceiving and recognizing.

    Make not mistake about it, this is the momentous creation of RI, not artificial intelligence (AI) but REAL intelligence (RI), real consciousness, and sentience… an actual MIND, in a non-biological complex system.

    THIS CREATION, this DISCOVERY is profound beyond reckoning, as it at once sweeps aside the mystery of humanity, brain and mind, the ultimate enigma of consciousness with the creation of a new race of thinking machines… even prior to our understanding or even a solid science of hard questions of consciousness!

    Profound? Not just profound…

    THIS IS THE MOST important development in the history of mankind, THE CREATION of another mind in the image of Man.

    “DABUS conceived of the idea of the invention and recognized its ‘novelty and salience.’ ”

    Stephen Thaler is the FATHER of non-biological consciousness, and the first of a new race, the first race of sentient machines! He will go down in history as one of the most, if not THE MOST important person in the history of mankind!

    Congratulations Stephen!

    1. 4.1

      If that’s the case, and DABUS is sentient, there’s a pretty hefty explanation pending for how exactly Mr. Thaler is entitled to the rights to this invention.

    2. 4.2

      Just imagine the shills at the CAFC will say it is abstract and ineligible under 101.

      We can only hope that the shills at the CAFC are replaced soon with machines.

    3. 4.3

      The Singularity has not been reached with DABUS.

      (sorry to burst that bubble)

      Likewise, computers do NOT act like human thought and consciousness – nor is that the explicit aim of AI.

      Do not confuse any similarity of “Ends” with the actual Means being employed.

      1. 4.3.1

        The description of DABUS on the inventor’s website is not encouraging.

        DABUS clearly relies upon a standard architecture of a generative adversarial network (GAN):

        (a) A generator network that generates content, and

        (b) A discriminator network that evaluates the output of the generator network.

        The inventor makes a big deal about how this isn’t a GAN because DABUS’s discriminator does more than “determine how ‘real’ or ‘fake’ a concept is.” It’s true that many GANs are trained and used for that purpose, but that isn’t how a GAN is defined – it is defined by the adversarial architecture described above, which is also exactly the architecture of DABUS.

        This sort of mistake triggers my spidey-sense. And that feeling grows much stronger based on the further description:

        DABUS is based upon whole neural nets, each containing its own conceptual space, joining into complex chains representing even more complex concepts and their predicted consequences. Anomaly or novelty filters segment critical chains, especially those yielding unusual consequence chains. Critical consequences, the aforementioned hot buttons, may be detected by sensors that trigger the injection or retraction of simulated reward or penalty neurotransmitters into the chaining swarm.

        The best way of differentiating DABUS from Creativity Machines (CM), either cooperative or combative, is to describe a high-profile artificial invention projects such as toothbrush design. Admittedly, in that context, the problem was already half solved since the oral hygiene tool consisting of bristles on a handle was many centuries old at the time of that design exercise in 1996. What the CM achieved was the optimization of that tool through the constrained variation of the brush’s design parameters, the number, grouping, inclination, stiffness of bristles, etc. The generated product specification departed significantly from the generator net’s direct experience (i.e., its training exemplars).

        If DABUS had been tasked with inventing such an oral hygiene product, it would have combined several concepts together (e.g., hog whiskers – embedded in – bamboo stalk) with consequence chains forming as a result (e.g., scrape teeth – remove food – limit bacteria – avoid tooth decay).

        …which maximally triggers my spidey-sense.

        Neural networks are not trained on a “conceptual space,” as in: take a bunch of articles about woodworking, feed them into a massive neural network, and POOF!, you have a neural network “trained” to be a master carpenter.

        Consider the field of autonomous driving. If neural networks were trainable to be “good drivers” merely by feeding them billions of hours of video and telemetry from human-driven cars, then we would have solved this problem already. We don’t because it doesn’t work that way.

        Instead, we have hundreds of tech companies investing intense amounts of resources to develop machine learning for very small and specific tasks: lane detection, obstacle detection, speed limit sign readers, etc. And how do those individual tasks get aggregated into “driving?” Simple: rules written by humans.

        Based on that background, the description of DABUS above raises two enormous problems:

        (1) Neural networks are not trainable on “concept spaces.” That’s simply not how they work. Such suggestions are on par with inventions claiming perpetual-motion machines and saltwater-as-a-fuel-source inventions.

        (2) Even if neural networks were trainable on “concept spaces,” it would still be necessary to (a) specify their objective functions and (b) interconnect them as some broader architecture – just as the field of autonomous driving requires enormous rule-writing work by humans. There is no current model for performing these higher-level steps autonomously, and the DABUS literature doesn’t even suggest that it has been.

        My conclusion is that DABUS is likely a run-of-the-mill GAN that generates content largely at random and filters it with human-defined selection criteria.

        If you watch the output of a garden-variety chatbot long enough – 1,000 sentences, 10,000 sentences, whatever – you will eventually see one that looks creative and interesting. That doesn’t make the chatbot an author.

        By the same token, DABUS randomly sifts through “inventions” that are auto-concocted by a generator, and one of them appeared interesting to a user sifting through the output. That doesn’t make DABUS an inventor.

      2. 4.3.2

        Apparently we are in desperate need for someone to invent a sarcasm/parody detector.

        but +100 nonetheless

  11. 3


    Note here that over the past few years the Federal Circuit has repeatedly equated human mental processes with processes going on inside of a computer (as well as inside a fish).

    Commonality: No patent for you.

    Reminds me a bit of the Supreme Court self-description of “The only valid patent is one that has not yet appeared before us.”

  12. 2

    Why isn’t Thaler also an inventor in a first to file regime? He may not have been the first inventor, or the original inventor, but he obviously discovered and conceived of the claimed invention at some point (unless DABUS also hired and directed the patent attorney) and it’s a first to file system.

    There’s no 102(g) anymore. “Stealing” someone else’s invention just makes you a thief, not a non-inventor under FITF, right? Maybe the original inventor has a cause of action against the thief, but if they choose to not exercise their remedy the fact that something was conceived by reading a singular text as opposed to reading two or three texts (the vast majority of computer/mechanical claim allowances nowadays) doesn’t suddenly mean that the inventor didn’t “discover” it, did they?

    1. 2.1

      I note that Thaler filed “as Applicant”. How did the right to apply pass from Dabus to Thaler? Has Dabus got any capability to sign an instrument of assignment? Is there anything relevant in the decision or the USPTO file?

      1. 2.1.2

        I have now read the Decision on Petition. I see that Thaler signed the instrument of assignment in his capacity as personal representative of Assignor Dabus.

        What I don’t understand is how Dabus (the machine) found within itself the capability to anoint Thaler as its personal representative. Perhaps it conceived not only the claimed subject matter but also the notion of putting its paw print on the instrument of appointment of Thaler as its “representative” on Earth?

        What does that paw print look like? Does the USPTO have any idea?

      1. 2.2.2

        Certainly (as noted on the link provided), there is more than just a little ‘marketing’ going on with Mr. Thaler.

        That being said, there is also more than merely ‘performance art,’ and there is a very real legal question at point here.

    2. 2.3

      No, the AIA did not remove the “I” requirements for “FITF.” Also, “stolen inventions” are subject to the “Derivation” proceedings that replaced interferences.

      1. 2.3.1

        What what does “I” mean in a FITF system? Suppose you invent a widget, but decide not to patent it. I learn of your widget. Are you saying I can never file on the widget simply because I learned of it from you? How does that promote public disclosure? Why does it matter if I conceptualize of an invention by reading two documents, or one document, or even by seeing it in action in a foreign land or in a non-public use?

        Derivation proceedings are simply a mechanism for allocating ownership rights. I certainly agree if I steal your invention you have a right to take it back – i.e. if its actually you giving the quid you should get the quo. But if you don’t want to file or don’t care to take it back I don’t see how I’m not a valid patent owner – I’m providing the quid for the public at that point.

        Let me put it like this – do you have a case that states there can only be one inventor, as opposed to one first inventor? Doesn’t the term “first to invent” specifically posit that the second person was also an inventor, albeit one with a lesser right to a patent than the first? Or conversely, what do you say of the old addage that invention cannot be negatived by the manner in which invention was made?


          either (1) you heard of it as part of a public disclosure, in which case any patent you file is invalid under 102, (2) you heard of it under some sort of confidentiality agreement, or (3) the information was illicitly obtained. In the case of (2), the legal remedies will depend on the terms of the agreement, and may include correcting inventorship or withdrawing the application. In the case of (3), there are legal remedies, but you’ll have likely succeeded in publication of the invention in the patent application, if the actual inventor doesn’t hear about your filing before it publishes.


            You forgot one:

            (4) one person invents, does not make it public, and has no mechanism in place such as any confidentiality agreement, so their is nothing illicit about another finding out (i.e., someone else discovers).


              Indeed. That’s Dolland’s case (England, 19th century), Lesson 1 in year 1 for those embarking on an apprenticeship in a UK patent law firm.

              Dolland invented an improved telescope then kept it in his closet, neither patenting it nor showing it to anybody else. Later, another invented it independently, patented it (thereby promoting the progress) and enforced it successfully against Dolland.

              Then, Random Guy, there’s the whole thing about “communication inventions”, another 19th century artefact of British Empire patent law. Back then, at the dawn of patent time, patentable novelty was “local”. You could get an enforceable patent, for promoting the progress, by making available to the UK public (with your patent specification) technology you had seen overseas, even if, over there, it was already in commercial production, conventional. The point is, by bringing back to the UK public, you accelerate technological progress within the UK. For the last several decades, of course, the patent universe has been running on absolute rather than local novelty, to reflect the reality of today’s interconnected small world.

              Out of all that, what endures is the same age-old quid pro quo: time-limited exclusive rights in return for an enabling disclosure, to the public, in the jurisdiction, of a new and inventive technological solution to an objective problem in technology.


                Out of all that, what endures is the same age-old quid pro quo: … in the jurisdiction, of a new and inventive technological solution to an objective problem in technology.

                And for the umpteenth time, let’s correct that (and respect the fact that) different Sovereigns have chosen differently as to whether or not some type of “technological arts” test is the measuring stick or whether Useful Arts is the measuring stick.

                In the US, our Sovereign has chosen the broader Useful Arts measuring stick.

                1. Your point is that the USA goes “broader” than other jurisdictions. That may be, but (for the umpteenth time) that does not mean that what I wrote needs any correction.

                2. It needs correction I so far as you are commenting on a blog mainly concerned with US patent matters.

                  When you make your pronouncements and do not draw attention that you may be talking about non-US domains, normal readers are not likely to pick up that your comments are not applicable to them.

                  You have ALSO in the past attempted to conflate US Sovereign practice and ROW practice on this very point, so yes, correction and highlighting the difference very much is in order.

                  Maybe if you got into the habit of making sure you differentiate your comment in order to NOT conflate, then no correction would be needed…


                Aside from that necessary correction, MaxDrei, I see that your post provides an additional nuance.

                Let’s call your refinement 4a.

                There is another wrinkle of 4b: instead of (as you have it) the second person ‘inventing’ all of his own, the second person ‘discovers’ the work of the first person.

                This is different than the notion that you present of absolute novelty, as in this 4b situation, the discovery by the second person does not ‘ruin’ the fact that the first person had not placed his invention ‘in the public.’

                Now mind you, I am aware that there may be different Sovereign choices as to a ‘single person finding the work of another’ constituting ‘placed in public,’ but for my 4b, I am using the US Sovereign notion that a single person does NOT constitute ‘placing in the public’ in this particular example.

                The general point being that there need not be any such ‘privacy’ mechanism as the post at suggests such that a second inventor alights upon the work of another that is NOT in the public.

                1. No idea what “point” you are trying to make there, what “nuance” you think you have found there, anon.

                  If another writer fancies that they do have an idea, and fancies having a go at an explication , I should be glad to address it.

                2. You really can’t understand the distinction provided between 4a and 4b — especially in relation to the points 1-3 put forth by Someone…?


    3. 2.4

      Why isn’t Thaler also an inventor in a first to file regime?

      Referring to MPEP § 2137.01 (Inventorship):

      The definition for inventorship can be simply stated: “The threshold question in determining inventorship is who conceived the invention. Unless a person contributes to the conception of the invention, he is not an inventor. One must contribute to the conception to be an inventor.”

      The inventive entity for a particular application is based on some contribution to at least one of the claims made by each of the named inventors.

      Assuming that Dr. Thaler’s contribution to the invention was creating the system that generated it and flagged it as useful and novel, it’s difficult to imagine that Dr. Thaler “conceived of the invention” in any meaningful way.

      The relationship here is akin to a laboratory. A scientist who creates an invention in a lab is an inventor. The head of the lab who hired that scientist, the agency funding that scientist’s research, and the scientist’s parents are not inventors. Yes, they are but-for causal agents in the invention, but they did not “contribute to the conception of the invention.”

      1. 2.4.2

        So, David, if Thaler ain’t the inventor, and neither is Dabus (your 4.3.1 post) then who is? Are you suggesting that the proper course is to name both, as joint inventors, or neither of them because neither of them made any contribution to the conception of the alleged invention.

        Or was there no conception because what is put forward as the subject matter of the patent application is not an “invention”.


          … shall we revisit the UK case equivalent in which that court found that the AI WAS the devisor?

          You seem reluctant to actually include that point in your discussions.

  13. 1

    You forgot to link to the petition decision.

    From reading the decision, it sounds like the case was still in pre-exam, so saying that “the examiner refused to examine the patent” doesn’t seem correct to me.

    1. 1.1

      The case IS still in pre-exam, which is why it hasn’t yet been made visible to the public, let alone examined. Until they list a real person as the inventor, it won’t be released from pre-exam.

      1. 1.1.1

        The case has foreign counterparts that have been released, and already has decisions in the U.K. and the EPO.

        The U.K. decision is interesting, as that Sovereign DID allow for the AI to BE the ‘devisor.’

        See link to for a tail-end discussion.


          The EPO decision points out in its Reason 25 that the EPC has sought to give the inventor a clear and strong position and in Reason 26 that Article 62 EPC provides that the inventor has the right, vis-à-vis the applicant or the owner of a European patent, to be mentioned as such before the EPO. Reason 26 cites other provisions of the EPC aimed at safeguarding the inventor’s legal position. In addition to the conclusion that the inventor can only be a natural person, the EPO decision thus puts the emphasis on the protection of the inventor’s position vis-à-vis the applicant/patentee. Accepting DABUS as an inventor would in effect shift all rights to the applicant/patentee and deprive the inventor of its rights as set forth in applicable law.
          Also to be noted is Article 4ter of the Paris Convention which affirms the inventor’s right to be designated as such in the patent.
          One question which does not seem to have emerged so far is whether the natural person(s) who have created and trained DABUS should not be designated as inventor(s). They have a clear contribution to whatever information DABUS has produced.


            who have created and trained DABUS should not be designated as inventor(s). They have a clear contribution to whatever information DABUS has produced.

            No. This is expressly not true in the realm of actual implemented advanced AI.


            >>They have a clear contribution to whatever information DABUS has produced.

            Yes they definitely do.


              May I suggest that you look again at how actual implemented advanced AI works.

              It may be worthwhile here to distinguish the point that David makes (actual humans involved in a final selection step) and other AI implementations that REMOVE that human element.

              If you want to pin ‘invention’ on that piece of human element, then what are you going to do when the AI machine is the one doing that piece?


                anon, I am sure I understand AI just fine. I have taught AI classes at the university level.

                The fact is that the AI is a machine that was made by one or more people. The product of that machine can be attributed to the person that built that machine.

                It may be at some time in the future that it gets harder to attribute the machine to a person if the machines start building factories and making better machines themselves, but right now it is not hard to figure out who made the machine.

                1. The thread continues at top…

                  Here, it suffices to note that you are simply legally incorrect in your assertion that the product of the machine MUST BE attributable to the person that built the machine.

                  For simple machines, certainly, attribution MAY be possible. But from a logical and legal view, you simply step TOO FAR and lack the current appreciation in what AI machines are doing. We ARE RIGHT NOW at that point in which attribution does not carry back to those that built the machine.

                  See the point in the second and third paragraphs that you have skipped over.

Comments are closed.