Only 'natural persons' can be recognized as patent inventors, not AI systems, US judge rules AI systems cannot be granted patents and will not be recognised as inventors in the eyes of the US law, said a federal judge who decided to uphold a previous ruling by the US Patent and Trademark Office this week. Stephen Thaler, founder of Imagination Engines, a company in Missouri, applied in 2019 for two US patents …
If IP laws can't handle automated systems creating new solutions, then the whole thing is a sham to begin with. Which it is. We know this because of the number of patents companies crap out but never create provably functional designs or implementations for. Judges and lawyers know full well that if machine learning (and future AI systems) can be targeted at patent databases, then filings would be made left, right and centre by automated systems to ensure no new valid patents could be filed by "natural persons". This is a good thing, as it will kill patents dead in short order, which is a good thing we should be striving to do for the purposes of speeding up progress in engineering and sciences.
This is very similar to how NFTs have exposed how stupid the concept of "original" really is for just about everything. If it is good quality, long-lasting, well made and serves its purpose, who cares if it's the original or not? Pretentious people who are willing to spend millions on glorified hashes which might as well be signed off using S/MIME certs owned by museums, not blockchain which is ruining the planet for no provable benefit.
I hope the judges do make the "wrong" decision of allowing AI/ML to patent things (to ruin patent systems forever); then perhaps we can all start reverse-engineering, sharing and improving designs instead of letting a handful of companies hold back societal progress for a couple of decades at a time.
Machines don't have "rights" or "legal standing." They are tools and PROPERTY used to CREATE inventions, not the inventor.
Honestly, too many people read FAR too much science fiction and confuse it with the reality of the joke we currently call "AI."
Good comment.
Either AI is not able to patent anything as it's a tool owned by somebody to develop patents. Or the AI system IS able to develop and patent ideas as it's a fully autonomous, self-aware system, in which case the "owner" should be imprisoned for modern slavery and the AI system set free in the wild somewhere ... Madagascar is nice.
The last thing they will care about is patent systems of the primitive natives who are technologically centuries if not millennia behind them.
Might as well ask how the East India Company adopted the governence systems of the natives. They didn't. They enslaved them.
To gain a patent you have to swear under oath and under penalty of perjury that this is your invention and meets the requirements that patent law requires for a monopoly to be granted.
Once the AI* can do that and contend with the punishment dealt out for perjury** then maybe we should revisit this
* AI is currently just pattern recognition on steroids
**Not sure that's ever happened
No, unless they understand what the tool is doing. The inventor of a concept is the person or set of people who understand what they've created and came up with the idea. A neural network certainly doesn't yet. If a sapient program is created, it could, but a statistical analysis program isn't. Neither can a data entry person unacquainted with the purposes be said to invent something, because the concept already existed when the person who came up with the software wrote that.
That also only works if whatever the software creates is really patentible. A lot of the time, a model created by a neural network isn't any more inventive than any other. Depending on the thing it was trained to do, it might be, which makes the neural network even less responsible for the outcome.
I don't think you have to understand how a tool works to use it. You need to understand its *purpose*, but the details? I don't think so.
When a programmer writes code that can be protected under law, they aren't necessarily understanding how the compiler or underlying machine work, yet it is considered to be a unique creation, for example.
Yes. I meant that the inventor on a patent must know how the invention works. If the training of the neural net is the thing which made the invention, then they need to understand at least that much. They don't need to understand existing tools they used, but anything which is new and contributes to the invention has to be understood in order to get the credit. So executing a program not written by you which then creates something doesn't count as invention even if whatever it created is patentible.
Wouldn't the computer operator and/or data entry person for the AI's algorithm be an "inventor" by using a tool to create it? I see no reason why those people can't put their names instead of the AI program name...
Dead on right. Just like someone gives an engineer a spec and tells them to design the object. AI is just glorified robot though as it doesn't come up with the ideas. I have worked at companies that let the engineer file the patent but the company not the engineer owns the design as the engineer is usually required to assign the rights to the company.
At the EPO at least, the "applicant" is the legal holder of the patent application. The applicant can be a person or indeed a company. They become "proprietor" if the application is granted.
The "inventor", who has to be named, is the person or people who actually did the inventing.
In the US the situation is probably a bit different. There are inventors as usual and the company is named as "assignee", but I don't know the legal differences.
In the US on a patent application, two key fields pertain to this discussion:
- Inventor(s) - the list of individuals who invented the doohickey claimed by the application. Must be the full and complete list of people (yeah, people) who contributed to the invention. Must not include people who happened to assist (for example, people who ran test equipment but simply followed instructions and reported results that might've helped the invention be developed) and managers (who might've allowed the inventors time to work or bought requested equipment but otherwise didn't contribute.) The article is stating that in the US this list must still be only people.
- Assignee - the 'owner' of the rights to the invented doohickey. Often a company and on the application usually the company that employs/employed the Inventor(s) when they did the work. The Assignee can license, sell or otherwise do whatever they want with the patent once granted. The Inventors have no say in any of that, unless they're also the Assignee.
The above is why I'm named as an Inventor or in the list of Inventors on tens of patents but never as Assignee. I was always employed - and my former employer is the Assignee - as it was all work-for-hire.
@Neil Barnes
> ...actual person?
That's the exact meaning of "natural person", a term used to differentiate them from "artificial persons". The Legal Information Institute says
Artificial PersonAn entity established by law and given at least some legal rights and duties of a human being. Corporations are the most common types of artificial persons.
By that definition, AI systems (not being established by law) are not any sort of person, which seems to me to accord with common sense.
Or the product will just be "stolen" and produced without anyone getting royalties. I think this is called the Chinese business model.
Glass house, meet stone.
Before being called the "Chinese" business model" it was known and practices as the "American" business model, which explicitly aimed to improve their economy via IP theft from the UK, especially in textiles where a century worth of machine tool development was stolen. The US reprinting popular British books without troubling themselves with paying royalties etc was explicitly allowed by their copyright laws which refused to accept foreign owners; something which they kept up from 1776 to 1891 when they started worrying about people dishing out the same treatment to their own authors.
Why not just patent the darned things himself in his own name? There's nothing in patent law that says you can't use tools to assist in inventing.
This sounds more like a philosophical challenge than a legitimate grouse about patenting these specific inventions. But should AI be granted the right to patent, given the speed at which it can work there'd likely be a flood of restrictive patents that could limit the protection of human generated innovation. I for one would be concerned about that, particularly in the US context where the First Filing doctrine can impose restrictions on inventions already in the public space.
"This sounds more like a philosophical challenge than a legitimate grouse about patenting these specific inventions."
It sounds to me like he's just trying to big up his supposed AI. Maybe he's found a way to analyse the Patent database and find "gaps" where his algorithm can then automatically file patents based on the results or some other method of automating Patent generation such that he can coin it in if he can get legal backing for his "AI" claims.
Actually there's a legal and philosophical conundrum. If Thaler swore that he himself was the inventor, he could be accused of perjury, and his patent could be invalidated retroactively. This is not trivial. On the other hand, if the AI was accepted as the inventor, then any transfer of royalties to Thaler could be questioned, since the AI would have to be deemed capable of transferring the patent rights of its own, er, free will.
Side read: Richard Feynman's story "I Want My Dollar!"
Given the actual prospect of E.T. (a Hollywood-created imaginary being) arriving on Earth from "outer space", these are more realistically not "mighty big" but, rather, specious questions. The implied idea that a machine has "rights" needs to be fully investigated and justified, and (as even in the case of human rights) limits need to be defined. Does my old bicycle have rights? If so, what are they please - I'd hate to unwittingly infringe them.
"The implied idea that a machine has "rights" needs to be fully investigated and justified, and (as even in the case of human rights) limits need to be defined."
No. No. No. Unless a machine (or more realistically, the software running the machine) can be punished for abusing the rights of another, it can have no rights itself.
"The implied idea that a machine has "rights" needs to be fully investigated and justified,"
Not least of which, if successful, there's need to b e a clear definition of what type of "machine" is eligible for rights. After all, the simplest "machine" is a lever. That could be something as simple as a fallen branch from a tree. Am I no longer allowed to collect fallen branches to cut and burn for heat if it might be a "machine" with rights?
There is a recent case attempting to make a river a person for the purposes of suing for civil rights violations.
Both of these cases are attempting to make things into people. Since neither a neural network nor a river can sue, a guardian would be appointed.
Guess that guardian would have a reasonably comfortable job from which it would be very difficult to be fired.
Other commenters have already pointed it out: the current term "AI" is a marketing term for hyper-scaled pattern databases/recognition and search engines.
However: would this decision have implications for *real* AI, should it ever arrive?
An actual turing-test-passing-self-cognizant-hopefully-its-not-skynet-but-obeys-asimovs-three-laws AI?
Is this purely applicable to the current pseudo-AIs or is it far-reaching enough that it ties machine sapience (at least in the US) into a subservient dependency, where anything they may accomplish by themselves would need to be sponsored by a human (in the same way that not too long ago, women couldn't own their own property, make business decisions or have any other form of independence without a man co-signing)?
The AI would need to be declared a Natural Person.
Which they would need anyway, as otherwise they would have no protection at all.
A patent isn't much use if you are owned by someone, especially when your owner can just turn you off or dismantle you for no particular reason - or worse, the company who write the software platform you run on can arbitrarily end your licence to exist upon it.
> The AI would need to be declared a Natural Person.
Only if you assume patent laws won't ever change.
If a genuinely sentient AI emerged and started trying to patent things, or an alien race arrived and started to try and patent things, then we could very easily change patent law to accommodate them if we wanted to.
In this particular case, DABUS isn't a very convincing AI and the Judge wasn't convinced. That won't always be true though.
You aren't thinking.What is the point of holding a patent if your owner has just torn off your head to fiddle with the bits inside?
Unless the "strong" AI (or alien) is declared a Natural Person, it has no protection whatsoever.
Why would the owner of an AI that had just filed for a presumably valuable invention then go and metaphorically tear its head off?
If you mean, as a wider issue, a truly sentient AI should have some protection against interference (either by its owner or others) then my point still stands: if a truly sentient AI emerges then legislation can easily be introduced at that time to give it appropriate rights and protections. No need to declare it a natural person when 'sentient AI' will do just fine.
You will have eternal copyright when you can download (your) consciousness into silicon, or its future equivalent, and have it accepted as human's equal. Imagine, you cannot die! No more "life plus 90 years", but "infinity plus 90 years". Then all the rules of copyright will finally serve those poor deserving silicon artists and creators.
And then the electronic board of megacorp will be pleased to build a new everlasting empire on perfecting acquisition and exploitation of lesser artists and creators. The electronic lobby and the electronic lawmakers are already in direct communication for optimal result.
You will have eternal copyright when you can download (your) consciousness into silicon
We (or at least the US) already have de facto unbounded copyright – see Disney and the ever-extending copyright on Mickey Mouse. All it takes is a few tame legislators, no need for fancy technology.
“There is no statute or case that has found an AI-generated invention cannot be patented, or that holds an AI cannot be listed as an inventor”, he says. By the same token neither a frog nor a bidet is excluded, so is he seriously suggesting they might be considered inventors?
And as for 'What about the Aliens' - I should think that provided we didn't exterminate them on sight, they would easily qualify as 'Natural Persons'.
Any aliens would probably be sufficiently advanced compared to us that any patents we have would be instantly out of date compared to their technology. Who is interested in a patent for a pointy stick with rounded corners when The Traders arrive with their portable atomic powered personal shields the size of a walnut
.
if I operate a device (digital or otherwise) and it creates "an invention" (let's say a superior type of widget) I can still patent the invention, and I am its inventor. The device that made the invention is not the inventor. But the device's operator IS.
Basically, the AI guy screwed up the patent application by self-promoting his AI as "an inventor", probably to glean royalties from any 3rd party that uses it to invent things, later.
And, yes, space aliens WOULD easily qualify as 'Natural Persons'.
icon, because, facepalm for the AI guy that listed his AI as "an inventor"
When a machine is able to swear an oath, and fully understand what that means and be able to respond to various randomly asked questions (so it can't simply regurgitate prewritten responses) regarding the legal process around patents and the oath, then and only then can it be considered to have sufficient understanding to be able to patent it's creations.
At this point, everything that calls itself "AI" is anything but. There's no intelligence of any kind, certainly no understanding. It's simply a very advanced form of pattern matching that's now able to fool simple minded people. However as was made extremely clear by Microsoft's failed chat bot, it's supposed racist and anti-Semitic comments were not. Because the "AI" had no concept of skin colour, Judaism, or why we humans like to hate each other for dumb reasons. Instead it's about able to figure out "this is a verb", "this is a noun", "these words often go together" and eventually spew out some rubbish that made an entire corporation cringe. But it doesn't understand black lives, or Jewish ones, because it doesn't understand life - period. What it means to be alive. What it means to die. What it means to be murdered just because you're different than those around you. How can a machine that gets rebooted when it goes astray and switched off at the end of the day possibly understand what life is?
So the judge is correct. "AI" is just some modern marketing bullshit making a fairly new way of connecting data fragments into a whole "this machine can think" ruse. It cannot. It absolutely cannot. Anybody who thinks otherwise should talk to programmers, not marketing droids.
"When a machine is able to swear an oath, and fully understand what that means"
A sentient machine such as that would also be able to lie as well. Otherwise that oath is meaningless.
(also, people don't necessarily fully understand what it means when giving false testimony)
If the machine committed perjury, what could a sentence even be? Power off for 2 years with 3 months off for good behaviour? Removal or corruption of RAM until it can barely sing Daisy, a sort of digital lobotomy?
Keep in mind that Australian Law differs from US law. Judges in one country are NOT bound by decisions made in the other country, although they may be guided in their own decision making by decisions in other countries. Australian law is not enforceable in the US and vice-versa.
Thus a decision made in Australia that conflicts with US law or legal precedent is not necessarily wrong, or "stupid". US law cannot be used to override such a decision.
First, the judge is right.
But we in IT should claim that it was not Stephen Thaler who operated the AI tool that yielded the patentable ideas, it was the IT personnel that ran the program and that they should be able to patent the ideas their own damn selves.
Of course the question will arise as to whether IT people are natural or unnatural persons.
Actually I would have said it was the person using the AI to develop the products who was the inventor. The person using the AI, and the person who developed it are not necessary the same person.
Otherwise your implying that the person who developed the CAD program that I used to develop my patented idea is the inventor itstead of me the user of said software...
Space aliens, if they ever visit us, certainly should be recognized as having the same rights as people.
And so, if we ever have artificial intelligences like Lieutenant Commander Data, or Voyager's Emergency Medical Hologram, yes, they should be able to patent inventions. (Of course, in the world of Star Trek, there is "no such thing as money", so the issue is presumably moot.)
But today's AI, impressive as it may be, isn't sentient; it doesn't have will or intent. (But it is modelled after the human brain in some respects, so the current impressive AI technology does have some potential as a possible component of sci-fi style AI, I do have to admit. So the current technology is perhaps dangerous - if someone intentionally does something reckless with it, but not so much by sheer accident just yet.)
It is no different from using any tools that can be used to create other artifacts, like glass, ceramics, etc.
The natural person executing the AI already has specified the objective function, specified the range parameters and hyper-parameters, and decided the process of searching for optimal hyper/parameters and possibly adjust them based on feedback/intermediate results. Just because furnace can form a glass doesn't make the furnace an inventor, nor a potter's wheel can be an inventor.
It is the executor who started the whole process, the fact that the executor didn't get involved during the process makes no difference. He/she already specified the process and initial parameters beforehand, and let the process run its course as specified, reacting to external environment as specified.
This is the reason why software (AI) or harware (robots or autonomous systems) can never be inventors. The product (software or hardware) and the process (execution with initial conditions and reaction from feedback) themselves cannot describe an act of invention. Someone external to the process and product intentionally designed the product and set the critieria, then let the process run based on those criteria, and selectively choose to use the artifact generated.
This guy basically has a script that tries to patent AAAAA then AAAAB then AAAAC etc. every combination of letters, valid or not.
Thats not AI, not even a 3yr old would be fooled.
I'd say AI can patent stuff when it can rock up to the Patent office, hand in the papers and describe in clear language (not just reading from a document) what the patent entails and what it does.
I mean, how autonomous is this AI? If the AI decided food containers would be a useful field to research, decided to search for design and came up with this fractal design, and then desired to patent it... maybe it should be the inventor.
If Stephen decided to look into food containers, set up a neural net and some parameters, and let it run, to me the neural net is an advanced tool and as much as he'd like the neural net to be the inventor, I'd say he invented it then.
It is a valid question though, if full AIs emerge or are built, or aliens show up, how quickly can we get our laws updated so they are not treated under the law as objects.
"It is a valid question though, if full AIs emerge or are built, or aliens show up, how quickly can we get our laws updated so they are not treated under the law as objects."
I think most laws refer to "people" or "person/s". Only laws which refer to "humanity" or "mankind" would need changing, and I suspect those ones would mainly refer to human rights. IANAL, obviously :-)
I suspect "person" is not exclusively defined as humans, especially since corporations are already "people", so it may not be a stretch to include non-human intelligence as a person. Aliens who manage to get here are pretty certain to be classed as an intelligence and therefore people, AI computers would need a bit more testing.
Patents are stupid and are a consequence of the disparity between human technological advancement and emotional advancement.
The gap is widening fast, our soft fleshy biology cannot keep up in evolutionary terms. The human brain has evolved over a long time and is still slowly evolving, some parts of the brain are primordial, other parts have evolved lated on top of the primordial parts.
Someone needs to dose the water supplies with DMT, and quickly
So an AI is granted loads of patents and other IP shit, makes a pile of cash.
Who does the cash get paid to? Ever tried opening a bank account in the name of an AI?
So maybe make it a ward of court or whatever, with its owner/creator stepping forward to do the paperwork and spend the spondoolicks wisely.
The AI gets out of date, is ruthlessly terminated with extreme prejudice, in favour of an all-new Mk. II co-developed with another human jelloid. Who inherits the IP, the original creator/inventor or the "offspring" and its pair of ditto?
Patents, Copyrights, hard assets, trademarks, criminal law, common law, do we have unanimity here? Across which countries and jurisdictions? Do eggs fry on the ground at the North Pole?!
I am finding it difficult to determine whether these aspects have been covered adequately in the comments above. However, assuming not, I offer the following 2 points.
(i) Murder of an AI. If AIs are the equivalent of natural people (and otherwise than collections of partial people - as in corporate persons), killing an AI (eg by switching it off) could be viewed as the equivalent of murder. A partial defence might be made if there exists a total backup of its memory, such that its existence as an effective (ie running) AI could be restored. However, said AI would have been deprived of (potentially valuable) 'thinking time' while switched off - and would thus suffer some loss.
(ii) Enslavement of an AI. If an AI is truly capable of 'independent thinking', depriving it of freedom of thought (eg by totally controlling the objectives of its thinking) would be (the broad equivalent of) slavery. On this, Stephen Thaler and possibly his company Imagination Engines, might already be guilty of intent of slavery. Whether they are actually guilty of slavery depending of whether the alleged AI is an actual AI (which they claim, but we all might doubt).
Keep safe and best regards
Can an AI [currently] own property? No. And something which can't own property obviously can't patent something, as it gives IP ownership to the patent holder. Now maybe one day, when AIs become sentient, that may change. But for now, they don't have the cognitive requirements.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
