Where We Are on AI Inventorship and Where We Should be Heading

“It is likely a matter of time until an AI will be able to simulate human thought, think creatively, and independently identify and solve problems…. If current laws remain unchanged…the owner of the AI-generated IP can and likely will attempt to protect AI-based inventions as trade secrets to the extent possible.”

AI - https://depositphotos.com/236313962/stock-photo-cyber-law-internet-law-concept.htmlThe past few years saw a meteoric rise of artificial intelligence (AI) products, services, and applications. AI has evolved from merely a buzzword or a cool new idea to a substantively used tool in a variety of applications, including autonomous driving, natural language processing, drug development, finance and cybersecurity among others. Companies, universities, and inventors world-wide noted the importance of AI and began seeking to patent various aspects of AI technology. Until 2018, these patent applications identified a human inventor who invented a particular aspect of the AI technology. Then, Dr. Stephen Thaler filed a patent application for a food container and a light emitting device that identified an AI, known as DABUS, as an inventor.

Dr. Thaler’s DABUS application became a poster child for what is now a worldwide question: can AI be an inventor, at least with respect to a patent application? The DABUS application was initially filed under the patent cooperation treaty (PCT) and is now being examined in more than a dozen countries, with divergent results. To date, only South Africa has granted a patent on the DABUS application. Australia presents some promise. There, the Australian Federal Court rejected the patent office’s initial denial and ordered the office to examine the DABUS application on its merits. On the other hand, in the United States, United Kingdom, Germany and European Union, the courts have rejected the DABUS application, holding that AI cannot be an inventor. Most recently, the England and Wales Court of Appeal upheld lower rulings that the DABUS applications were deemed to be withdrawn, but the three judges were split, with the two patent specialists on the panel taking different views.

Defining an Inventor

A basic reason for these disparate decisions is how the term “inventor” is defined by statute. South Africa and Australia reached a result favorable to the DABUS application because their laws do not exclude a non-human, e.g., AI, from being an inventor. The laws of the United States, United Kingdom, and Europe, on the other hand, require the inventor to be human. This is because the inventor engages in the mental step of conception which, according to respective patent offices or the courts, can only be performed by a human.

Proponents of IP rights in the United States believe that international consensus of strong IP protections is vital to technological growth. After all, the United States Constitution grants inventors a limited monopoly to make, use, and/or sell their inventions in exchange for disclosing these inventions, in full, to others. This limited monopoly is an impetus driving innovation.

One might ask: are there immediate ramifications to technological growth if AI cannot be named as an inventor? Unlikely.

AI is Still a Tool

Although AI has become more and more advanced over the years, AI is still far away from inventing meaningful technological improvements and itself driving innovation. AI today is incredibly good at performing tasks, computing and parsing data, finding patterns, determining trends, and identifying solutions. All these tasks, however, are based on training the AI on previous examples, datasets, and patterns to solve for like outcomes. This is true across various AI industries. Cars are capable of driving autonomously because the AI component of these cars has been trained on data recorded over millions of driven miles. By the same token, AI can use natural language processing to summarize text, translate documents, and even chit-chat because it has been trained on millions of documents, question-answering datasets and previous conversations. AI can identify protein sequences and drug formulations, identify financial fraud, or diagnose cybersecurity attacks because AI has been trained to identify those patterns. There is even research in AI writing executable programing code from a proposed business case. Notably, all these tasks require human involvement, oftentimes significant. A human being designs the AI, trains the AI, and verifies when the AI produces a result. Due to advances in computing power and available training data, the AI can perform these tasks faster, more accurately, and better than humans.

AI today excels in performing tasks, yet it is still deficient in creative thinking, abstract and original thought, and identifying problems and undefined corresponding solutions. These deficiencies are, according to some jurisdictions, the hallmarks of inventorship. Until AI can transcend these limitations, it will remain a tool, rather than an inventor.

The Law Must Think Long-Term

This begs the question: are there long-term impacts to technological growth if AI cannot be named as an inventor? Definitely.

It is likely a matter of time until an AI will be able to simulate human thought, think creatively, and independently identify and solve problems. At this point, AI is likely to satisfy the current definition of what it takes to be an inventor. If current laws remain unchanged, and the AI cannot be an inventor, the owner of the AI inventions would not qualify for IP protections. The owner would need to decide whether AI inventions are even worth the cost of investment, regardless of the potentially immense benefit they could offer. The owner of the AI-generated IP can and likely will attempt to protect AI-based inventions as trade secrets to the extent possible. Alternatively, pursuing and concentrating on innovations where AI is not an inventor is another option. This will certainly impede technological growth.

One Thing Certain: We Need Certainty

AI is still in its infancy. DABUS has brought the AI inventorship issue to the forefront worldwide, marking this a good time to ponder, debate, and reach international consensus on whether and under what circumstances an AI can be an inventor. One potential solution is to pass statutes stating that AI cannot be an inventor. Another is to treat AI inventions as legal persons, similar to how corporations are treated now. Regardless of which solution is reached, it is clear that IP owners need more certainty on the issue from the courts or governments—or both.

Image Source: Deposit Photos
Image ID:236313962
Copyright:phonlamai 

The Author

David McCombs

David McCombs is primary counsel for many leading corporations in inter partes review (IPR) and is regularly identified as one of the most active attorneys appearing before the Patent Trial and Appeal Board (PTAB). His clients benefit from his 35 years of practice, which include appellate argument, patent litigation, and portfolio development.

David McCombs

Eugene Goryunov is a partner in the Intellectual Property Practice Group in the Chicago office of Haynes and Boone and an experienced trial lawyer that represents clients in complex patent matters involving diverse technologies. He has extensive experience and regularly serves as first-chair trial counsel in post-grant review trials (IPR, CBMR, PGR) on behalf of both Petitioners and Patent Owners at the USPTO.

David McCombs

Dina Blikshteyn is a counsel in the Intellectual Property Practice Group in the New York office of Haynes and Boone. Dina’s practice focuses on post grant proceedings before the U.S. Patent and Trademark Office, preparing and prosecuting domestic and international patent applications, as well as handling trademark and other IP disciplines. Dina is also a co-chair of the artificial intelligence practice at Haynes and Boone.

Warning & Disclaimer: The pages, articles and comments on IPWatchdog.com do not constitute legal advice, nor do they create any attorney-client relationship. The articles published express the personal opinion and views of the author as of the time of publication and should not be attributed to the author’s employer, clients or the sponsors of IPWatchdog.com. Read more.

Discuss this

There are currently 9 Comments comments. Join the discussion.

  1. Anon October 12, 2021 9:47 am

    I could not disagree more.

    There ARE immediate effects: think for example of joint inventorship (and challenging the notion that AI is merely a ‘tool used’).

    AI today is incredibly good at performing tasks, computing and parsing data, finding patterns, determining trends, and identifying solutions. All these tasks, however, are based on training

    The moment you have both structured and unstructured training (the moment the human ‘mind’ takes its hands off the wheel — to turn a phrase), is the moment that you no longer can claim human as the LEGAL inventor.

    Once you have that — and we have that NOW — then you have to realize that (in several Sovereigns, including the US), the prohibition against machine as legal inventor casts the shadow of doubt of ANYONE being able to legitimately be deemed inventor enough to have patent protection. Naming only the human inventors opens the door to legal challenge of enforceability.

  2. Moshe October 12, 2021 10:22 am

    First let’s restore meaningful protection to inventions by humans (after it’s been gutted over the past 15 years – eBay, Alice, AIA, etc etc), and THEN we can discuss –> The owner would need to decide whether AI inventions are even worth the cost of investment, regardless of the potentially immense benefit they could offer.

  3. Anon October 12, 2021 11:30 am

    One other (small) nit – patents provide what is deemed to be a negative right – not a positive one.

    Statements such as “After all, the United States Constitution grants inventors a limited monopoly to make, use, and/or sell their inventions in exchange for disclosing these inventions, in full, to others” thus is less than helpful, and invites confusion on the legal term of monopoly. One does NOT earn a positive right to make, use or sell. This is also important to keep in mind as to the larger majority of patents are improvement patents and a baseline item (an item being improved upon) may well still be under patent protection to another, such that one could NOT be making using or selling (without some agreement with that other).

  4. MD Hoyle October 12, 2021 3:51 pm

    A horrible idea of the first order. By using AI, large corporations will create an interconnected blanket of patents that will forever block any attempts of single inventors from ever getting into the field.

    This will absolutely be the end of the small inventor.

  5. Seth A Milman October 13, 2021 8:44 am

    This topic is broader than just inventorship in a patent. It has implications in property law in general. If an AI is listed as an inventor, that AI would have ownership of the invention under the patent law. But AI’s are currently seen as tools, not persons, under other areas of law like copyright. I don’t think our laws would change to allow an AI to own property for a very long time. So what does that if we acknowledge an AI as an inventor in a patent, where inventors are automatically owners of the invention? Should the AI be viewed as a tool under the patent law, like they are under the copyright law? If so, can a tool actually *conceive* an invention?

  6. Curious October 14, 2021 5:20 pm

    Once you have that — and we have that NOW — then you have to realize that (in several Sovereigns, including the US), the prohibition against machine as legal inventor casts the shadow of doubt of ANYONE being able to legitimately be deemed inventor enough to have patent protection.
    That’s your OPINION. My opinion is that AI is a human-created tool that identifies potential solutions to problems already identified by the first mover (i.e., the human).

    To me, an inventor should be able to do the following:

    1) identify a problem, devise a way to solve (or at least partially solve) the problem, and/or
    2) create/discover something novel and then recognize a utility for that something.

    A sophisticated AI system that generates (randomly or not) a number of organic compounds without recognizing a utility for these compounds is not an inventor — even if one of those organic compounds turns out to be a cure for cancer.

    As I have written before, I suspect there will be a time when an AI would be capable of being an inventor in the manner I discussed above. However, we aren’t even remotely close to that point. An AI that is able to identify a problem and spontaneously devise a solution to that problem gets us far closer to Vernor Vinge’s singularity that I believe exists today. One could argue that such an AI would be a sign that the singularity is upon us.

    I read the DABUS application involving fractal surfaces. It is arguably both non-enabled and ill-conceived. Moreover, unless I, as a patent attorney, could question the so-called “inventor” as to certain aspects of the so-called “invention,” I would not be comfortable putting my name on any application naming an AI as an inventor. I would have the same hesitancy if I was approached by a parent whose 10-year old child was the alleged inventor of something. Unless I could question the child about the invention, I would be concerned that I was being used, by the parent, as a tool to further the reputation of the child.

    If one wants to defend a so-called “AI-created” invention, I would wait for better ones to defend. The two “inventions” I am aware of raise to just above the level of being cringe-worthy — and not being cringe-worthy is a very, very low bar.

    This point aside, where are the other inventions? Its been 2 years since the food container patent application was filed. I would expect that an AI truly capable of being an inventor would be pumping out hundreds, thousands (if not a lot more) of inventions by now.

    If Thaler wants positive publicity for the capabilities of DABUS, I would suggest an approach involving publicly disclosing vast numbers of valuable inventions created by DABUS — and then publicly declaring that there are even more/better inventions to be disclosed if DABUS could be declared an inventor. That would truly put the capabilities of AI on the forefront of the public consciousness. Right now, we’ve got two low-grade disclosures — both of which I would describe as being on non-commercially valuable technology. If Thaler wants more recognition for his creation, he needs it to produce better — a whole lot better.

  7. Anon October 15, 2021 9:54 am

    That’s your OPINION. My opinion is that AI is a human-created tool that identifies potential solutions to problems already identified by the first mover (i.e., the human).

    No.

    That is NOT my opinion – that is an objective view taking into consideration the legal meanings.

    It is also the assertion of the owner of DABUS, as he has disclaimed being the one that provided the point of invention.

    It is also the point that I have explored in my scenario of a person in a second room being presented a box WITH an output from a first room and wanting that person in the second room being deemed legal inventor merely by opening the box and reading the contents therein.

    I have already debunked your attempted de minimis treatment of “just a tool.”

    That ship has sailed.

    As to your other musings, meh, perhaps people are waiting for the dust to settle before jumping into the fray – I mean, would you want to follow based on Thaler’s lack of success to date?

  8. Curious October 16, 2021 9:54 am

    That is NOT my opinion – that is an objective view taking into consideration the legal meanings
    It isn’t a fact. Hence, it is an opinion — your opinion.

    It is also the assertion of the owner of DABUS, as he has disclaimed being the one that provided the point of invention.
    Like he has no agenda he is trying to push. Come on. You never struck me as being capable of ascending to the level of naivety you are exhibiting here. There is a reason why I brought up the example of a parent coming to me — saying that their 10-year old child was an inventor. Given the circumstances, I would be highly skeptical. Wasn’t it Carl Sagan who stated “extraordinary claims require extraordinary evidence”? Since computers have yet to be inventors, I need detailed evidence that establishes that the inventor was, in fact, the computer.

    It is also the point that I have explored in my scenario of a person in a second room being presented a box WITH an output from a first room and wanting that person in the second room being deemed legal inventor merely by opening the box and reading the contents therein.
    It is a poor analogy. A half-baked idea is not an invention. I need to know far more about the process than what your analogy provides.

    I have already debunked your attempted de minimis treatment of “just a tool.”
    You’ve debunked nothing. It is but a tool. It is created by humans. It receives input selected by humans. It is directed by humans. The output is interpreted by humans. If you want to argue that AI is more than that, then I need to see the evidence.

    As to your other musings, meh, perhaps people are waiting for the dust to settle before jumping into the fray – I mean, would you want to follow based on Thaler’s lack of success to date?
    People are waiting? Are you saying that “people” are directing the inventive process?

    Regardless, the point remains is that the two so-called Ai-generated inventions presented for our review are of exceptionally low quality. They have very little practical utility and their disclosures are also of particularly low quality. As for Thaler’s lack of success to date, one could hardly ask for poorer vehicles by which to change patent law than the applications he presented.

  9. Anon October 26, 2021 11:41 am

    Apologies Curious, as this thread had escaped my notice (and replying so late may escape your notice).

    Let’s pick up the discussion at the next opportunity.

Post a Comment

Respectfully add to the discussion.

Name *
Email *
Website