Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Patents AI The Courts

US Patent Office Confirms AI Can't Hold Patents 44

The US Patent and Trademark Office (USPTO) asserts that only humans can be recognized as inventors on patent applications, not artificial intelligence systems, although the use of AI in the invention process is permitted and must be disclosed. The Verge reports: The agency published (PDF) its latest guidance following a series of "listening" tours to gather public feedback. It states that while AI systems and other "non-natural persons" can't be listed as inventors in patent applications, "the use of an AI system by a natural person does not preclude a natural person from qualifying as an inventor." People seeking patents must disclose if they used AI in the invention process, just as the USPTO asks all applicants to list all material information necessary to make a decision.

However, to be able to register a patent, the person using the AI must've contributed significantly to the invention's conception. A person simply asking an AI system to create something and overseeing it, the report says, does not make them an inventor. The office says that a person who simply presents the problem to an AI system or "recognizes and appreciates" its output as a good invention can't claim credit for that patent.

"However, a significant contribution could be shown by the way the person constructs the prompt in view of a specific problem to elicit a particular solution from the AI system," the USPTO says. The office also says that "maintaining 'intellectual domination' over an AI system does not, on its own, make a person an inventor" -- so simply overseeing or owning an AI that creates things doesn't mean you can file a patent for them.
This discussion has been archived. No new comments can be posted.

US Patent Office Confirms AI Can't Hold Patents

Comments Filter:
  • by dgatwood ( 11270 ) on Wednesday February 14, 2024 @05:44PM (#64240148) Homepage Journal

    Perfect. Now, can we have AI generate a few billion patent proposals and then publish them somewhere as prior art?

    • Any idea you think of patenting is already out there.
      Whether you make money of your patent only depends on how good your lawyer is.

    • You could but the USPTO nor the courts would take that seriously as a source of prior art.

    • A publication only invalidates a patent as prior art if it fully discloses the entirety of the claim and how to implement it. Your suggestion is basically the equivalent of the million monkeys with a million typewriters. And since the signal-to-noise ratio would be so low, the examiners at the USPTO most likely aren't going to spend the time going through all of them. It might not be worth the effort even for lawyers working to invalidate a patent as part of a lawsuit.
  • by MpVpRb ( 1423381 ) on Wednesday February 14, 2024 @05:53PM (#64240162)

    ...l can NOT invent. It's in their design.
    All they do is do remixes of their training data

    • by erice ( 13380 )

      ...l can NOT invent. It's in their design.
      All they do is do remixes of their training data

      Arguably, that is all humans do too. It is just a much larger and largely unknowable training set. Occasionally, someone does recognize what they think is the training data and the parties end up in court.

      • Not so. Humans can invent and come up with new ideas. Computers can not.

        We are not just wet computers. If we were then where did all the big inventions come from when there was literally no prior concept?
        Someone made the first rocket. Someone made the first computer. Someone made the first camera. Someone made the first shoe. Someone made the the first donut. Someone made the first LED screen. And so on.

        AI could not and has not done anything like that for us yet and with current type of technology

        • by gweihir ( 88907 )

          We are not just wet computers.

          Obviously. General intelligence and consciousness are characteristics machines do not have and, quite possibly, cannot have. At least with currently known Physics it is quite impossible to ever get there for consciousness (there is simply no mechanism) and likely impossible for general intelligence. A major extension would be needed, but while known Physics is far from complete, there is not even a hint where that extension would be possible.

          Obviously the deranged Physicalists regard that as heresy. And lik

        • We are not just wet computers. If we were then where did all the big inventions come from when there was literally no prior concept? Someone made the first rocket. Someone made the first computer.
          Someone made the first camera. Someone made the first shoe. Someone made the the first donut. Someone made the first LED screen. And so on.

          AI could not and has not done anything like that for us yet and with current type of technology can not and never will.

          Just so I understand humans were able to evolve from nothing without the benefit of design, computers, minds or Ahura Mazda using nothing more than trial and error yet somehow you see fit to make claims mind is now somehow required to do things that are "truly original".

          These AI are very good at mixing and matching existing things to come up with new variants no one thought of yet. Nothing truly original.

          All I want is for someone to provide a concrete example of "remixes of their training data" vs. a concrete example of "truly original". I doubt it'll ever happen.

          • > Just so I understand humans were able to evolve from nothing without the benefit of design, computers, minds or Ahura Mazda using nothing more than trial and error yet somehow you see fit to make claims mind is now somehow required to do things that are "truly original".

            No. I have no idea what you're even trying to say which leads me to say "no".

            > All I want is for someone to provide a concrete example of "remixes of their training data" vs. a concrete example of "truly original". I doubt it'll eve

            • No. I have no idea what you're even trying to say which leads me to say "no".

              It is not clear what about the statement you did not understand given its trivial nature.

              I was making a rather obvious point if a mindless process of trial and error resulted in the "truly original" outcome of life/human beings then asserting mind is required to create original things is incorrect.

              Remixing: computer mixes a zillion combinations of random chemicals to make a zillion new drug candidates. Each new combination is run through a simulator to see if it actually does anything.

              Original: the telescope, rocket, LED screen, camera, Roman concrete, farming, domesticated animals, the twin turbo v6, run flat tires, ceramics, transistors, X-rays for medical use, satellites, thx sound and video, poetry, art, rock gardens, Hitch Hikers Guide to the Galaxy.

              Obviously modality (how something is achieved) is distinct from outcome (what is achieved).

              What outcomes are necessarily the result of "true originality" and what outcomes are the result of "remixing"? Can you te

      • by gweihir ( 88907 )

        Bullshit. Maybe that is all _you_ can do, but general intelligence is a thing. Again, maybe a thing you do not have. People that can have original ideas are rather rare.

    • by jonadab ( 583620 )
      Right, the issue is that we're calling something "AI" that is basically just a stastical engine. ChatGPT is essentially a markov chain generator on all the steroids ever. Calling it "intelligence" is misleading.

      In the unlikely event that anybody ever figures out how to make an AI that is _actually_ intelligent, in the sense of actually understanding what things mean and being self-aware to a meaningful extent (and I mean a meaningful extent, as in sapient/sentient, a higher standard than the level of "sel
      • Right, the issue is that we're calling something "AI" that is basically just a stastical engine. ChatGPT is essentially a markov chain generator on all the steroids ever. Calling it "intelligence" is misleading.

        LLMs are neural rather than statistical language models (SLMs). The intelligence is in the ability of these systems to generalize learned concepts and apply them generally. For example the ability to add fractions is applied when asking the LLM to double my chocolate chip cookie recipe.

        In the unlikely event that anybody ever figures out how to make an AI that is _actually_ intelligent, in the sense of actually understanding what things mean and being self-aware to a meaningful extent

        Does intelligence require self-awareness or can it persist independently?

        and being self-aware to a meaningful extent (and I mean a meaningful extent, as in sapient/sentient, a higher standard than the level of "self-aware" that we see in animals), then we would need to grant
        them civil rights, which would presumably include, among other things, the right to own property, including intellectual property such as copyright and patent.

        I'm the only conscious mind in the universe. The rest of you are fake organic automatons.

  • by r_a_trip ( 612314 ) on Wednesday February 14, 2024 @05:59PM (#64240174)
    Just come up with a half plausible story and just claim the invention as your own? Who is going to check if it was AI generated?
    • Anyone you sue for practicing the patent will check if it was AI generated, so it has to be at least 51% plausible. Regardless, I think this will change the scope of a "person of ordinary skill in the art", since many more practitioners can craft AI prompts than break new ground by inventing something. I like to think a thousand prompts to a dozen AI's would reveal the fraud on the patent office (and maybe scare some bonafide inventors out of further litigation).

      But, your are correct that if you are able to

    • by gweihir ( 88907 )

      Patent fraud happens to be a crime and it happens to make the patent invalid. So yes, people are going to check.

  • by Local ID10T ( 790134 ) <ID10T.L.USER@gmail.com> on Wednesday February 14, 2024 @06:19PM (#64240212) Homepage

    I would like to disclose that I used a pencil while doodling my idea. I would also like to confirm that the pencil I used does not hold a patent on my idea.

    Its a tool. It has a user.

    We do not have independent AIs. Maybe we will -someday. But for now, it is just a tool.

  • Imagine a future AI that is 100% as creative, intelligent, and self-motivated as a human being. It files a patent. This brings about a bjillion questions that are completely unsolvable today:

    * What is the legal given name of the AI?
    * What is the AI's age and can the AI legally sign the paperwork?
    * What country is it a citizen of?
    * What is the mailing address for sending the paperwork?
    * What bank account will the AI use to remit payment? If it receives a royalty check in the mail, can it deposit that chec

    • A few of those are answered by requiring that an AI be "registered/born" in some country upon creation. That would provide a birthdate, country, etc, etc.

      The rest, hell if I know. I'm glad I won't be there to have to figure it out if true AI ever comes to pass.

      • by gweihir ( 88907 )

        A few of those are answered by requiring that an AI be "registered/born" in some country upon creation. That would provide a birthdate, country, etc, etc.

        Actually, it would not. It would just provide a convenient fiction. About as valuable as arbitrarily assigning these "characteristics". Oh, and what do you do when a copy is being made. Same identity? Different one? And when you merge them back together? The fact of the matter is that AI is software and data and these do not have identity.

    • by gweihir ( 88907 )

      Indeed. AI is a machine, and not even really an entity, much less a person. No idea why so many fall to a complete irrational animism when AI is concerned.

    • by jvkjvk ( 102057 )

      * What is the legal given name of the AI?
      Whatever the AI want's it to be?

      * What is the AI's age and can the AI legally sign the paperwork?
      Age of consent

      * What country is it a citizen of?
      Wherever it's system reside

      * What is the mailing address for sending the paperwork?
      Wherever it establishes a PO box for.

      * What bank account will the AI use to remit payment? If it receives a royalty check in the mail, can it deposit that check? With what picture ID?
      The bank account the AI sets up. Why couldn't it deposit a

      • by MobyDisk ( 75490 )

        uhh.... not sure if you are kidding, but just in case you aren't, none of those answers actually work in the real world.

        * What country is it a citizen of?
        Wherever it's system reside

        The system resides in cloud servers spanning multiple countries. Even if it were in a single country, residence != citizenship. Today, to get citizenship, one must usually be born from a woman who resides in the country or is a citizen of that country. Most countries don't just want random citizens "assigned" to them.

        The bank account the AI sets up.

        Riiight, by sending it's photo ID to the bank, and providing it's Tax

        • by jvkjvk ( 102057 )

          >The system resides in cloud servers spanning multiple countries. Even if it were in a single country, residence != citizenship. Today, to get citizenship, one must usually be born from a woman who resides in the country or is a citizen of that country. Most countries don't just want random citizens "assigned" to them.

          Well, let's say AI has to be incorporated then, so it hash a country of residence.

          >>The bank account the AI sets up.
          >Riiight, by sending it's photo ID to the bank, and providing it

  • I typed in a comment in some code I was about to write today describing what I was going to do and the AI assistant wrote it all for me as a suggestion in about 2 seconds. I hit the tab key to accept, thank you very much.

    But I have all kinds of ideas, and I think that pretty soon I will be able to just blue-sky them to my assistant. After some review I will tell it to file a patent application.

  • An AI is a machine, not a person at all.

    • The USPTO is using the legal definition of "person". A "natural person" is a human. A "non-natural person" is, for example, a corporation, which are already prohibited from being listed as the inventor on a patent.
      • by gweihir ( 88907 )

        Hmm. So a non-natural non-person is excluded by default then? What about a natural non-person?

        This US legal idea of corporations being "persons" is beyond bizarre.
           

  • Part of the problem is software patents, that require no physical representation whatsoever. Far too many of them are fraudulent in too many ways, and used to smother innovation related even by fraudulent lawsuits and SLAPP suits to prevent participation in the industries. Simply ending the preactice and discarding software patents would free up a great deal of competition and innovation. Copyrights would still be available to hinder cut&paste abuse of corporate intellectual property, but the outrageous

  • If I ask an AI to solve a problem and it comes up with a solution someone else patented does that invalidate the patent as obvious in the eastern district of Texas?

    Let's stipulate for the sake of argument it can be proven the AI does not know about the patent because the AI was trained prior to the filing of the patent or some similar arrangement.

To stay youthful, stay useful.

Working...