The book “against intellectual monopoly” has shaped a lot of my thinking on this topic - economists have looked at the various occasions in which patents were introduced into an industry (or extended in scope), and there is no evidence they actually improve innovation/efficiency/outcomes (including the pharma industry!). I was quite surprised as my whole life, it was sold to me as an incentive-boosting measure which in turn would lead to said outcomes.
With that lens, I welcome gradually phasing this stuff out, especially as we navigate into the unknown game-theory landscape AI-as-inventors brings.
You spend billions to get a drug from concept to approval - and then once you've invested all that money, someone else can just sell it too, free loading on all the studies you ran? Why would anyone invest in drug studies?
I need a bit more depth and detail to believe that this doesn't destroy the pharma industry.
What would the empirical evidence even look like? It's not like the modern pharma industry existed before patents.
Patents are an incentive to encourage an inventor to lay out an invention or process in exchange for the state protecting that process, we did this because there have been in the past inventions that have been lost, that were valuable largely because the inventor died without documenting what they did (keep in mind the first patent was issued in 1331[0], this is old law)
The complicating factor is that as time passes our ability to reverse engineer has grown, however I'm not sure that invalidates the need for patents, the question is whether the new patents are being assessed well from a novelty / inventiveness perspective
I'm having a hard time even grappling with how that could be true?
I always assumed that intellectual property was invented in order to protect against a specific use case:
If researching a new product is extremely cost intensive. But once a product is invented, it is easy to reverse engineer how the product works. Then the first firm will need intellectual property to put in the initial cost, otherwise they will not do so, as they know they will not have enough time to recoup their costs in the market before a competitor moves in with a copy-cat product without having to paid the initial costs.
There are two things to tackle here which I’m keen not to mix up as I think their epistemological properties are quite different:
1) [the stronger one] while the scenario/narrative is a compelling one (or maybe it just feels compelling as I’ve heard it so many times), if it doesn’t have experimental/data backing I have to abandon it.
2) [the weaker one, as it replaces a narrative with another narrative within a complex system] I’ll only give the highlights as the arguments are a lot more eloquently laid out in the book; part of it is comparing the force of “many inventor nodes building on top of many invention nodes” vs “inventor nodes (with more investment individually?) building on top of fewer invention nodes”, part of it is the game theory effect of companies collectively investing less (proportionally) in R&D as the ROI from lawyers under this regime has more power, part of it was that actually, the reverse-engineering-simplicity story was too overblown and that actually the friction + domain knowledge has a stronger effect than people think (they published a paper on this). There were others, but it’s been a while now!
In many research-intensive products go-to-market costs are bigger than the cost of actual invention. You buy a pharma startup for a few million for their patents, then spend tens of millions on certification, trials and manufacturing pipelines. Your competitors would spend most of that too on the same markets.
Also, true inventions are rare. A lot of stuff that is being patented is just effort spent, that a lot of people could reproduce (and routinely reproduce, then hit the patent and spend more time to find a workaround).
Indeed. Patents incentivize investment in R&D. There is an argument to be made that the scope of patentable inventions should be more limited, in particular preventing trivial patents that didn’t require substantial R&D, and maybe also that patents shouldn’t last as long.
But doing away completely with patents would certainly stifle companies’ willingness to invest in R&D. They’d rather wait for someone else to invent something they can copy.
Reverse engineering may be easy for really simple inventions, but quickly becomes so hard you may as well invent the thing from scratch. Look at USSR's domestic chip production. At one point they succeeded in reverse-engineering chips like Intel's 8086, VAX etc., but chip design very quickly became so complex reverse-engineering the entire chip became impossible. I would say if an invention can be easily reverse-engineered, then it is a simple foundational idea that should not be patented, and if it is truly innovative, then it cannot be reverse-engineered easily and doesn't need patent protections.
Then there is the fact that when something is patented, that has a chilling effect on competition, making the market less efficient.
There are also a lot of really silly patents that end up benefitting no-one, not even their inventor, but only result in needless litigation. The recent lawsuit between Nintendo and PocketPair comes to mind.
While there are cases in which patent law can help individual people profit from their invention, once all consequences are tallied, the overall effect of patent law on society appears to be negative.
Sanity! No AI has accountability so also should not own any benefits (not only patent but anything beneficial). Violate that and you created a blackhole of value creation.
This appears to be confusing patent inventors with patent owners. It’s the latter who benefit and presumably are accountable for the use of the patent and potential plagiarism.
If an invention was trivial enough to be invented by AI then why should we allow that action to be patented? The expenditure of labor to research that invention was minimal and definitionally not novel.
Broadly speaking, IP law generally exists to protect the rights of humans. The law doesn't generally recognize that inanimate objects have rights.
The idea that an AI could have some sort of property rights is a nonstarter, legally speaking. It's just as invalid of a legal idea as claiming that a tree could have a patent on the shape of its leaf.
So when people go to the patent office and say "I didn't make this! an AI invented this", the obvious response from the patent office is "cool, well only humans get rights, and if you didn't make it, you can't get a patent on it, so too bad". This isn't a judgement of AI.
Now, a lot of people come to presume that this means that anything that AI touches is not subject to any IP rights -- but that's not what this means at all. Humans are allowed to use tools to create things that they have IP rights to. Your typewriter itself can't hold a copyright to a book, but if you use a typewriter, you can still hold the copyright to the book.
Ultimately, whether or not the use of AI is disqualifying to a human inventor doesn't really have anything to do with AI -- it all hinges on whether or not the human meets the requirements of holding the patent.
>"The Patent Office ordered the plaintiff to provide the name of a person as the inventor. The plaintiff refused to do so, and the application was rejected."
implies that if he provided his name as the inventor, the application may not have been rejected.
Oh, please let it be the second option. Let AI be the thing that kills the "intellectual property" because humans will never manage to shake off that terribly wrong decision by themselves.
I don't personally feel the inevitable UBI/subsistance will make intellectual property much of a patentable/profitable field (...for too much longer), thanks to generative AIs' massive transformations (entrylevel &+).
The US ruled similarly to Japan, but years ago, from copyrights through patents... from my limited POV, the LLMs (specifically) and art models are just regurgitating stolen concepts... echo'ing Picasso's great artists steal! mantra. The US has already ruled this is legal (e.g. newspaper content isn't "stolen" when a genAI summarizes it for a 3rd-party user).
Having sat with published authors, discussing their work/book with LLMs... it is really an interesting perspective on "readers' perspective(s)" [human ¬].
>The US ruled similarly to Japan, but years ago, from copyrights through patents... from my limited POV, the LLMs (specifically) and art models are just regurgitating stolen concepts... echo'ing Picasso's great artists steal! mantra.
No, in the US AI output is ineligible for copyright not because "art models are just regurgitating stolen concepts", but because only human created works are eligible for protection.
>only works created by a human can be copyrighted under United States law, which excludes photographs and artwork created by animals or by machines without human intervention
> from my limited POV, the LLMs (specifically) and art models are just regurgitating stolen concepts
Formally speaking, it's not the case, though this is commonly misunderstood. Statistical models are definitionally ampliative, otherwise they wouldn't be statistical. One can argue about it until they're blue in the face, but it almost always comes down to a misunderstanding of what the models are, what the mathematics behind them is a description of, and what the underlying logical structures represent.
The thing is that the position and objection to these models isn't actually a substantial, reasoned position where the words have a direct meaning. Though it's dressed up like reason, it's not the point. It's a kind of metaphor. This actually does reflect the nature of intellectual property law. The legal framework is knowingly illogical at an object-level, because the end its seeking is completely divorced from the means. It has to be, because the idea of intellectual property is absolutely unjustifiable in-and-of-itself. It's just a useful legal fiction to make sure people are getting paid by commoditizing ideation. That's not a bad thing, it just means you have to be mindful that bottom-up reason will lead you astray when dealing with it.
I know this is about the Japan and not the US, but software patent law has been incompatible with traditional IP protection in the US for a long time and it really doesn't make sense in the current age.
> "... the plaintiff submitted an application in 2020 for food containers and other items invented by DABUS, an artificial intelligence the plaintiff had created."
If you were seriously trying to patent some AI-created invention, why would you claim it was created by AI? You would simply put your own name on it. This was obviously a case of pushing the envelope to see how far he could go.
This is just as hair-brained as going down to the police station and claiming that your gun just murdered someone, then being surprised when they don't put the gun in jail.
The law does not recognize the anthropomorphization of inanimate objects.
One thing i've got to wonder. Would this always remain the case, at what point should society seriously consider the "personhood" of an AI (as a noun).
I agree with the other top-level comment next to yours (at the time of writing): when we're willing to enforce consequences for them in the same way we would for people. If I violate laws, I can get put in jail, and then I (most likely) can't use any computers until I get out. To consider an AI a person, it needs to have legal liability in the same way a fleshy person does.
Eh. No one has been able to prove to my satisfaction that they're conscious, or even simply define what it is that they claim to possess. Pick something else.
In my opinion, no jurisdiction in the world would be able to approve AI as an inventor on patent applications.
And for a very simple reason: you could easily overwhelm any intellectual property bureau just by having your AI drown them in AI slop. Even if most of these patents get refused, just refusing a patent is a lot of work, I imagine.
The plaintiff is Stephen Thaler who has made a career of this litigation all over the world.
To my knowledge, he has notched only one win (i.e., granted patent) in South Africa, where patents are only cursorily examined [1].
The last word in the US is from the Federal Circuit a couple of years ago [2]. Same basic outcome: only a human being can be an inventor.
That said, the new Director of the USPTO has indicated that inventors should feel free to use AI however much they want as long as a human name is on the patent. However, it should be stressed that the Director's guidelines have not been litigated yet.
This is a news article about a dumb publicity stunt where a crank put his "AI" on a patent application, and the court said "you have to put your own name on it". It has no bearing whatsoever on debates about whether AI is good or bad, or whether it's ok that OpenAI looked at your Github, whether your coworker Gary is committing too much slop with Claude Code, or whatever else people want to make it about.
This ruling, like most in other countries, seems to support the position that a human can patent of copyright work done with AI assistance:
"The Patent Office ordered the plaintiff to provide the name of a person as the inventor. The plaintiff refused to do so, and the application was rejected."
> human can patent of copyright work done with AI assistance
Not sure about patents in the US but irt copyright, only the parts that are not LLM output are copyrightable. All LLM output is automatically public domain.
So if you have a work that was done with AI assistance, only the pieces of that work that are human authored can be subject to copyright. The AI parts cannot, if there are any.
I think it's long past time we get rid of the silly idea of intellectual property all together. If AI has the potential to do any good in the world in its current form, its that.
That is not exactly true under US law. You're simplifying what the copyright office has said to the point where you're missing the key points of what they were trying to convey.
The copyright office has affirmed multiple times that whether or not you use an LLM is irrelevant. Copyright eligibility requires "sufficient human-authored expressive elements". It doesn't matter what tools you use -- an LLM, a troop of trained monkeys, etc.
Ultimately all that matters is whether or not the human creativity involved qualifies. Because copyright is ultimately a right that protects human creativity.
So yes, if you put "write me a book" into ChatGPT -- that clearly does not quality for copyright. "Write me a book" itself is not creative enough for copyright.
Now on the other hand, if you spend 1000 hours writing a book, and you run it through ChatGPT for suggestions and/or edits -- there is no reason why that LLM output would not qualify.
Yes, provided you can separate the two (e.g. a book and illustrations in one case). AFAIK the courts have still not ruled on what happens when AI and human contributions cannot be separated etc.
It varies a lot in other countries, but in most (if not all) an AI cannot hold a copyright.
Thats why *SOME* humans will still be needed. They'll be accountability sinks when (NOT IF) the AI in charge goes off the rails. The human will then be summarily be blamed.
This is how the reverse centaur operation works. LLMs suck and not work in increasingly bad ways, and the companies who sell them treat them as one would buy psychic services (read: entertainment). So they need a token human to person-wash this slop.
I really can't understand the moral compass of people who would pirate other peoples' works under "fair use" to train AI, only to turn around and try to claim ownership of them when AI regurgitates it.
note that this was in 2020 (pre-chatgpt), with the author's own "ai", "DABUS", and it appears that the author wanted solely DABUS to be listed as the patent holder, which does not seem to indicate any insane greed or whatever.
the likelihood of one single guy having the same data scraping & storage capabilities as the big players, years before them (i see info about DABUS back to 2018), is slim.
Because AI doesn't just regurgitate it. Make up a new word and ask ChatGPT use it in a sentence - you've now got a brand new sentence that was not in its training data. If it only regurgitated data then it wouldn't be able to use that word in a sentence.
The same applies to image generation - they can generate images that almost certainly were not in the training data.
The truth is as long as there is competition, having morals does not exist in the tech/crypto/ai industries given the goal is to make money. That’s it.
Only after the participant has completed their grift or extraction operation then they begin virtue signalling their ‘morals’. It is fake.
If you are here for asserting morals, this is the wrong industry.
This is consistent with rulings in other courts globally around IP rights. IP protects content created by humans. Your AI slop is effectively public domain.
>>That doesn't mean anything produced by those tools is public domain
They can't produce anything on their own. They have to be prompted which is initiated by humans at this point, so the patents can be owned by the initiator(human) not the tool.
That isn’t what the courts have decided. They just decided it has to be a human on the patent application name. You can use whatever tool you want to get there, but if you patent a thing, it has to be a human in the name.
I think we’re saying the same thing. If you’re using AI as a tool to support human creative content that’s one thing. But what courts are pushing back on is trying to patent/protect content where the core creator was AI. That’s what most people mean when they say “AI slop.” There courts are consistently saying you can’t protect this.
I haven't been able to square this belief (This is what i believe too.) with what I perceive as so, so many people making projects, putting them on github and slapping an MIT/GPL license on them.
If IP rights can't be applied to generated code then how are they able to apply a such a license to them?
I've asked this before and the response was along the lines of people thinking their multiple prompting amounted to human creative process and therefore it was covered but ... how? Any lawyers around that can ELI5 it for us? Maybe links to a lawyer somewhere who did?
Because the “AI slop is uncopyrightable” people are misunderstanding court rulings like this. It’s not that AI output can’t by protected by IP, it’s that AI is not a person and so you can’t assign IP rights to it. You CAN assign IP rights to the human who did it (if they can show it’s non-trivial, like a haiku or photographer).
AI use is slowly creeping into pure mathematics and proving theorems or providing legging to mathematical breakthroughs. Just go watch some Terrance Tao videos to see some recent work. In addition, theorem provers and the likes have been around for awhile. Some of these systems create novel ideas or bridge novel ideas in ways that are arguably not “obvious” in any sense of the term.
While as a species our key strength has been our intelligence and it’s been core to our identity, and computing has slowly over decades infringed on this forcing us to rewrite what it is to be human, I understand the defensive view.
I also see LLMs and other AI systems spit out complete nonsense that’s truly obvious to most people. But that doesn’t make any of these systems, in my opinion, incapable of creating or bridging novel new ideas that I would call far from obvious had we substituted a human in place of it. I didn’t look at the patents in question, plenty of obvious patents make it through anymore, so that could be the case here, but I believe AI isn’t far away if not already there of creating truly patentable inventions if someone were to push it.
No, I mean legally they should be considered obvious, as the difficulty to create them becomes small. It makes no sense to give someone a monopoly on an idea that anyone could get just by prompting an AI.
Now if the invention also includes some real world work, or if the AI took a huge amount of tokens/money to reach the conclusion, ok. But otherwise an AI coming up with the idea at low cost should invalidate a patent of that idea (the AI not being trained on the patent of course.)
The book “against intellectual monopoly” has shaped a lot of my thinking on this topic - economists have looked at the various occasions in which patents were introduced into an industry (or extended in scope), and there is no evidence they actually improve innovation/efficiency/outcomes (including the pharma industry!). I was quite surprised as my whole life, it was sold to me as an incentive-boosting measure which in turn would lead to said outcomes.
With that lens, I welcome gradually phasing this stuff out, especially as we navigate into the unknown game-theory landscape AI-as-inventors brings.
You spend billions to get a drug from concept to approval - and then once you've invested all that money, someone else can just sell it too, free loading on all the studies you ran? Why would anyone invest in drug studies?
I need a bit more depth and detail to believe that this doesn't destroy the pharma industry.
What would the empirical evidence even look like? It's not like the modern pharma industry existed before patents.
Patents are an incentive to encourage an inventor to lay out an invention or process in exchange for the state protecting that process, we did this because there have been in the past inventions that have been lost, that were valuable largely because the inventor died without documenting what they did (keep in mind the first patent was issued in 1331[0], this is old law)
The complicating factor is that as time passes our ability to reverse engineer has grown, however I'm not sure that invalidates the need for patents, the question is whether the new patents are being assessed well from a novelty / inventiveness perspective
-[0]: https://en.wikipedia.org/wiki/History_of_patent_law
I'm having a hard time even grappling with how that could be true?
I always assumed that intellectual property was invented in order to protect against a specific use case:
If researching a new product is extremely cost intensive. But once a product is invented, it is easy to reverse engineer how the product works. Then the first firm will need intellectual property to put in the initial cost, otherwise they will not do so, as they know they will not have enough time to recoup their costs in the market before a competitor moves in with a copy-cat product without having to paid the initial costs.
There are two things to tackle here which I’m keen not to mix up as I think their epistemological properties are quite different:
1) [the stronger one] while the scenario/narrative is a compelling one (or maybe it just feels compelling as I’ve heard it so many times), if it doesn’t have experimental/data backing I have to abandon it.
2) [the weaker one, as it replaces a narrative with another narrative within a complex system] I’ll only give the highlights as the arguments are a lot more eloquently laid out in the book; part of it is comparing the force of “many inventor nodes building on top of many invention nodes” vs “inventor nodes (with more investment individually?) building on top of fewer invention nodes”, part of it is the game theory effect of companies collectively investing less (proportionally) in R&D as the ROI from lawyers under this regime has more power, part of it was that actually, the reverse-engineering-simplicity story was too overblown and that actually the friction + domain knowledge has a stronger effect than people think (they published a paper on this). There were others, but it’s been a while now!
In many research-intensive products go-to-market costs are bigger than the cost of actual invention. You buy a pharma startup for a few million for their patents, then spend tens of millions on certification, trials and manufacturing pipelines. Your competitors would spend most of that too on the same markets. Also, true inventions are rare. A lot of stuff that is being patented is just effort spent, that a lot of people could reproduce (and routinely reproduce, then hit the patent and spend more time to find a workaround).
1 reply →
Indeed. Patents incentivize investment in R&D. There is an argument to be made that the scope of patentable inventions should be more limited, in particular preventing trivial patents that didn’t require substantial R&D, and maybe also that patents shouldn’t last as long.
But doing away completely with patents would certainly stifle companies’ willingness to invest in R&D. They’d rather wait for someone else to invent something they can copy.
5 replies →
Reverse engineering may be easy for really simple inventions, but quickly becomes so hard you may as well invent the thing from scratch. Look at USSR's domestic chip production. At one point they succeeded in reverse-engineering chips like Intel's 8086, VAX etc., but chip design very quickly became so complex reverse-engineering the entire chip became impossible. I would say if an invention can be easily reverse-engineered, then it is a simple foundational idea that should not be patented, and if it is truly innovative, then it cannot be reverse-engineered easily and doesn't need patent protections.
Then there is the fact that when something is patented, that has a chilling effect on competition, making the market less efficient.
There are also a lot of really silly patents that end up benefitting no-one, not even their inventor, but only result in needless litigation. The recent lawsuit between Nintendo and PocketPair comes to mind.
While there are cases in which patent law can help individual people profit from their invention, once all consequences are tallied, the overall effect of patent law on society appears to be negative.
Sanity! No AI has accountability so also should not own any benefits (not only patent but anything beneficial). Violate that and you created a blackhole of value creation.
This appears to be confusing patent inventors with patent owners. It’s the latter who benefit and presumably are accountable for the use of the patent and potential plagiarism.
If an invention was trivial enough to be invented by AI then why should we allow that action to be patented? The expenditure of labor to research that invention was minimal and definitionally not novel.
1 reply →
Isn't patenting an action to protect a benefit (whichever patent or benefit would be)?
1 reply →
Can the petitioner re-file with his own name as the inventor, or does this mean that all AI-generated inventions are unable to be patented?
Broadly speaking, IP law generally exists to protect the rights of humans. The law doesn't generally recognize that inanimate objects have rights.
The idea that an AI could have some sort of property rights is a nonstarter, legally speaking. It's just as invalid of a legal idea as claiming that a tree could have a patent on the shape of its leaf.
So when people go to the patent office and say "I didn't make this! an AI invented this", the obvious response from the patent office is "cool, well only humans get rights, and if you didn't make it, you can't get a patent on it, so too bad". This isn't a judgement of AI.
Now, a lot of people come to presume that this means that anything that AI touches is not subject to any IP rights -- but that's not what this means at all. Humans are allowed to use tools to create things that they have IP rights to. Your typewriter itself can't hold a copyright to a book, but if you use a typewriter, you can still hold the copyright to the book.
Ultimately, whether or not the use of AI is disqualifying to a human inventor doesn't really have anything to do with AI -- it all hinges on whether or not the human meets the requirements of holding the patent.
[delayed]
>"The Patent Office ordered the plaintiff to provide the name of a person as the inventor. The plaintiff refused to do so, and the application was rejected."
implies that if he provided his name as the inventor, the application may not have been rejected.
Oh, please let it be the second option. Let AI be the thing that kills the "intellectual property" because humans will never manage to shake off that terribly wrong decision by themselves.
You can't prove something is/isn't created with AI.
Also, if AI generated content cannot be copyrighted, they can't infringe copyright as well
5 replies →
I don't personally feel the inevitable UBI/subsistance will make intellectual property much of a patentable/profitable field (...for too much longer), thanks to generative AIs' massive transformations (entrylevel &+).
The US ruled similarly to Japan, but years ago, from copyrights through patents... from my limited POV, the LLMs (specifically) and art models are just regurgitating stolen concepts... echo'ing Picasso's great artists steal! mantra. The US has already ruled this is legal (e.g. newspaper content isn't "stolen" when a genAI summarizes it for a 3rd-party user).
Having sat with published authors, discussing their work/book with LLMs... it is really an interesting perspective on "readers' perspective(s)" [human ¬].
>The US ruled similarly to Japan, but years ago, from copyrights through patents... from my limited POV, the LLMs (specifically) and art models are just regurgitating stolen concepts... echo'ing Picasso's great artists steal! mantra.
No, in the US AI output is ineligible for copyright not because "art models are just regurgitating stolen concepts", but because only human created works are eligible for protection.
>only works created by a human can be copyrighted under United States law, which excludes photographs and artwork created by animals or by machines without human intervention
https://en.wikipedia.org/wiki/Monkey_selfie_copyright_disput...
This is true, but I feel like it accomplishes the same spirit/thing.
3 replies →
> from my limited POV, the LLMs (specifically) and art models are just regurgitating stolen concepts
Formally speaking, it's not the case, though this is commonly misunderstood. Statistical models are definitionally ampliative, otherwise they wouldn't be statistical. One can argue about it until they're blue in the face, but it almost always comes down to a misunderstanding of what the models are, what the mathematics behind them is a description of, and what the underlying logical structures represent.
The thing is that the position and objection to these models isn't actually a substantial, reasoned position where the words have a direct meaning. Though it's dressed up like reason, it's not the point. It's a kind of metaphor. This actually does reflect the nature of intellectual property law. The legal framework is knowingly illogical at an object-level, because the end its seeking is completely divorced from the means. It has to be, because the idea of intellectual property is absolutely unjustifiable in-and-of-itself. It's just a useful legal fiction to make sure people are getting paid by commoditizing ideation. That's not a bad thing, it just means you have to be mindful that bottom-up reason will lead you astray when dealing with it.
You're absolutely right!
I know this is about the Japan and not the US, but software patent law has been incompatible with traditional IP protection in the US for a long time and it really doesn't make sense in the current age.
> "... the plaintiff submitted an application in 2020 for food containers and other items invented by DABUS, an artificial intelligence the plaintiff had created."
The plantiff is Stephen Thaler: https://imagination-engines.com/founder.html
If you were seriously trying to patent some AI-created invention, why would you claim it was created by AI? You would simply put your own name on it. This was obviously a case of pushing the envelope to see how far he could go.
This is just as hair-brained as going down to the police station and claiming that your gun just murdered someone, then being surprised when they don't put the gun in jail.
The law does not recognize the anthropomorphization of inanimate objects.
What I want to see is patent officers using AI to label patent applications as "not novel" if the AI can invent it.
But, since the income of a patent office is determined by how many patents they approve, one can dream ...
One thing i've got to wonder. Would this always remain the case, at what point should society seriously consider the "personhood" of an AI (as a noun).
I agree with the other top-level comment next to yours (at the time of writing): when we're willing to enforce consequences for them in the same way we would for people. If I violate laws, I can get put in jail, and then I (most likely) can't use any computers until I get out. To consider an AI a person, it needs to have legal liability in the same way a fleshy person does.
If there’s a consensus that AI is sentient and conscious and there are ways it can act autonomously, probably.
Corporate personhood has already been disastrous enough. We don't need to compound it with AI personhood on top.
Consciousness?
Eh. No one has been able to prove to my satisfaction that they're conscious, or even simply define what it is that they claim to possess. Pick something else.
[dead]
In my opinion, no jurisdiction in the world would be able to approve AI as an inventor on patent applications.
And for a very simple reason: you could easily overwhelm any intellectual property bureau just by having your AI drown them in AI slop. Even if most of these patents get refused, just refusing a patent is a lot of work, I imagine.
How would you even recognize the assignment of property rights to a big box of numbers?
And even if you did, it's entirely inanimate, how would it even exercise them?
Those applications cost money and would create thousands of jobs for displaced AI workers.
The plaintiff is Stephen Thaler who has made a career of this litigation all over the world.
To my knowledge, he has notched only one win (i.e., granted patent) in South Africa, where patents are only cursorily examined [1].
The last word in the US is from the Federal Circuit a couple of years ago [2]. Same basic outcome: only a human being can be an inventor.
That said, the new Director of the USPTO has indicated that inventors should feel free to use AI however much they want as long as a human name is on the patent. However, it should be stressed that the Director's guidelines have not been litigated yet.
[1] https://artificialinventor.com/patent/
[2] https://www.cafc.uscourts.gov/opinions-orders/21-2347.OPINIO...
This is a news article about a dumb publicity stunt where a crank put his "AI" on a patent application, and the court said "you have to put your own name on it". It has no bearing whatsoever on debates about whether AI is good or bad, or whether it's ok that OpenAI looked at your Github, whether your coworker Gary is committing too much slop with Claude Code, or whatever else people want to make it about.
This is fixable by simply replacing the AI with a human I dont think the laws catch or can determine the differences
Well who invented it then? User of said AI or owner of said AI or no one?
If you’re in the EEA or UK and reject the tracking, you can still use your browser’s Reader Mode to read the text. Or on the console:
> Your AI slop is effectively public domain.
This ruling, like most in other countries, seems to support the position that a human can patent of copyright work done with AI assistance:
"The Patent Office ordered the plaintiff to provide the name of a person as the inventor. The plaintiff refused to do so, and the application was rejected."
> human can patent of copyright work done with AI assistance
Not sure about patents in the US but irt copyright, only the parts that are not LLM output are copyrightable. All LLM output is automatically public domain.
So if you have a work that was done with AI assistance, only the pieces of that work that are human authored can be subject to copyright. The AI parts cannot, if there are any.
I think it's long past time we get rid of the silly idea of intellectual property all together. If AI has the potential to do any good in the world in its current form, its that.
> All LLM output is automatically public domain.
That is not exactly true under US law. You're simplifying what the copyright office has said to the point where you're missing the key points of what they were trying to convey.
The copyright office has affirmed multiple times that whether or not you use an LLM is irrelevant. Copyright eligibility requires "sufficient human-authored expressive elements". It doesn't matter what tools you use -- an LLM, a troop of trained monkeys, etc.
Ultimately all that matters is whether or not the human creativity involved qualifies. Because copyright is ultimately a right that protects human creativity.
So yes, if you put "write me a book" into ChatGPT -- that clearly does not quality for copyright. "Write me a book" itself is not creative enough for copyright.
Now on the other hand, if you spend 1000 hours writing a book, and you run it through ChatGPT for suggestions and/or edits -- there is no reason why that LLM output would not qualify.
https://www.copyright.gov/ai/Copyright-and-Artificial-Intell...
Yes, provided you can separate the two (e.g. a book and illustrations in one case). AFAIK the courts have still not ruled on what happens when AI and human contributions cannot be separated etc.
It varies a lot in other countries, but in most (if not all) an AI cannot hold a copyright.
1 reply →
Thats why *SOME* humans will still be needed. They'll be accountability sinks when (NOT IF) the AI in charge goes off the rails. The human will then be summarily be blamed.
This is how the reverse centaur operation works. LLMs suck and not work in increasingly bad ways, and the companies who sell them treat them as one would buy psychic services (read: entertainment). So they need a token human to person-wash this slop.
I really can't understand the moral compass of people who would pirate other peoples' works under "fair use" to train AI, only to turn around and try to claim ownership of them when AI regurgitates it.
note that this was in 2020 (pre-chatgpt), with the author's own "ai", "DABUS", and it appears that the author wanted solely DABUS to be listed as the patent holder, which does not seem to indicate any insane greed or whatever.
the likelihood of one single guy having the same data scraping & storage capabilities as the big players, years before them (i see info about DABUS back to 2018), is slim.
Because AI doesn't just regurgitate it. Make up a new word and ask ChatGPT use it in a sentence - you've now got a brand new sentence that was not in its training data. If it only regurgitated data then it wouldn't be able to use that word in a sentence.
The same applies to image generation - they can generate images that almost certainly were not in the training data.
You cant make a man understand the moral compass when his salary bla bla bla...
Don't forget exceptionalism: this is so disgustingly wrong... except when I do it. In my case it is moral and perfectly justified.
The truth is as long as there is competition, having morals does not exist in the tech/crypto/ai industries given the goal is to make money. That’s it.
Only after the participant has completed their grift or extraction operation then they begin virtue signalling their ‘morals’. It is fake.
If you are here for asserting morals, this is the wrong industry.
This is consistent with rulings in other courts globally around IP rights. IP protects content created by humans. Your AI slop is effectively public domain.
That's not how I understand it.
AI is a tool, like your keyboard or your code editor.
Those can't own patents. That doesn't mean anything produced by those tools is public domain, it just means the attribution has to belong to a human.
>>That doesn't mean anything produced by those tools is public domain
They can't produce anything on their own. They have to be prompted which is initiated by humans at this point, so the patents can be owned by the initiator(human) not the tool.
4 replies →
That isn’t what the courts have decided. They just decided it has to be a human on the patent application name. You can use whatever tool you want to get there, but if you patent a thing, it has to be a human in the name.
I think we’re saying the same thing. If you’re using AI as a tool to support human creative content that’s one thing. But what courts are pushing back on is trying to patent/protect content where the core creator was AI. That’s what most people mean when they say “AI slop.” There courts are consistently saying you can’t protect this.
5 replies →
> Your AI slop is effectively public domain.
I haven't been able to square this belief (This is what i believe too.) with what I perceive as so, so many people making projects, putting them on github and slapping an MIT/GPL license on them.
If IP rights can't be applied to generated code then how are they able to apply a such a license to them?
I've asked this before and the response was along the lines of people thinking their multiple prompting amounted to human creative process and therefore it was covered but ... how? Any lawyers around that can ELI5 it for us? Maybe links to a lawyer somewhere who did?
a person publishing as if a AI is the creator is publishing under a pseudonym.
AI has all the IP rights of a pen, pencil, chalk, or crayon.
Because the “AI slop is uncopyrightable” people are misunderstanding court rulings like this. It’s not that AI output can’t by protected by IP, it’s that AI is not a person and so you can’t assign IP rights to it. You CAN assign IP rights to the human who did it (if they can show it’s non-trivial, like a haiku or photographer).
2 replies →
It heavily depends on human involvement. AI is merely a tool.
Your ai slop is effectively something you own, because you wrote the prompt.
If an AI can invent something then it should be considered obvious.
The problem is that you have creations that aren't really obvious. Have you seen that rocket engine with a crazy laborious design made with an AI?
AI use is slowly creeping into pure mathematics and proving theorems or providing legging to mathematical breakthroughs. Just go watch some Terrance Tao videos to see some recent work. In addition, theorem provers and the likes have been around for awhile. Some of these systems create novel ideas or bridge novel ideas in ways that are arguably not “obvious” in any sense of the term.
While as a species our key strength has been our intelligence and it’s been core to our identity, and computing has slowly over decades infringed on this forcing us to rewrite what it is to be human, I understand the defensive view.
I also see LLMs and other AI systems spit out complete nonsense that’s truly obvious to most people. But that doesn’t make any of these systems, in my opinion, incapable of creating or bridging novel new ideas that I would call far from obvious had we substituted a human in place of it. I didn’t look at the patents in question, plenty of obvious patents make it through anymore, so that could be the case here, but I believe AI isn’t far away if not already there of creating truly patentable inventions if someone were to push it.
No, I mean legally they should be considered obvious, as the difficulty to create them becomes small. It makes no sense to give someone a monopoly on an idea that anyone could get just by prompting an AI.
Now if the invention also includes some real world work, or if the AI took a huge amount of tokens/money to reach the conclusion, ok. But otherwise an AI coming up with the idea at low cost should invalidate a patent of that idea (the AI not being trained on the patent of course.)