by Dennis Crouch
Oral arguments in the AI-inventorship case of Thaler v. Vidal are set for June 6, 2022. Prof. Ryan Abbott is set to argue on behalf of the patent owner (and AI creator) Stephen Thaler. Assistant US Attorney Dennis Barghaan will argue on the USPTO’s behalf.
Thaler developed an AI that he calls DABUS. DABUS apparently created two inventions–a “neural flame” and a “fractal container.” But, Thaler refused to name himself as inventor. Rather, he says that it was DABUS who did the inventing. But, the PTO refused to issue the patent without a human listed inventor. After being rejected by the PTO, Thaler filed a civil action in the E.D. Va. That court dismissed the case–holding that a non-human device cannot quality as an “inventor.” Abbott and Barghaan litigated the case below as well.
The question on appeal:
As stated by Thaler: “whether an AI-Generated Invention is patentable.”
As stated by the Gov’t: Whether “an artificial intelligence device comprised solely of source code can[] qualify as an ‘inventor’ under the Patent Act.”
Lots of folks might resist Thaler’s suggestion that the AI is the one who invented, but at this dismissal stage of the litigation that fact is procedurally assumed to be true. Both sides in this case have good arguments, but we should probably be troubled if the result is that “two otherwise patentable inventions cannot receive patent protection” despite the statutory statement that “[p]atentability shall not be negated by the manner in which the invention was made.” 35 U.S.C. 103.
Federal Circuit asked to Decide whether US Patent Law Excludes Non-Human Inventors
Professor Crouch, I think you fell into a linguistic trap by Thaler. That is, the question isn’t whether the invention is patentable (which might bear on 35 U.S.C. § 103) but whether DABUS can be named as an inventor. These are separate, orthogonal questions. The government is perfectly within its rights to accept that invention is patentable but refuse to grant a patent on that invention to a thing.
DEMON SEED (1977), a film adaptation of Dean Koontz’s 1973 novel of the same name, features an AI computer named Proteus IV. Proteus IV creates a cure for leukemia is just a few days. Some scientists ask Proteus IV’s designer, Dr. Alex Harris, “Did you apply for a patent?” The computer also functions very much like an AMAZON ECHO.
You are stretching the last sentence of 103 way too far. Patentability is negated by the status of the inventor (which is only in a very broad sense a part of “the manner in which it was made”) all the time. Pre-AIA 104–part of the same statute that added the last sentence of 103–explicitly excluded activity outside the United States, except by inventors who were serving in the US military. Pre-AIA 115 also requires the applicant to swear an oath that he regards himself as the original and first inventor. An AI can’t swear any oaths. It is true that both of those provisions have been amended, but they show the original meaning of 103 did not extend to barring exclusions of non-human inventors from patentability.
if the result is that “two otherwise patentable inventions cannot receive patent protection” despite the statutory statement that “[p]atentability shall not be negated by the manner in which the invention was made.” 35 U.S.C. 103.
That’s an interesting way of looking at it. I suppose that even those Office employees are specifically excluded from applying for patents, the employee could simply tell someone close to them and that person would be a valid “inventor” because the patentability is not negated by the manner in which in the invention was made? Is that the conclusion – I just need a trusted confidant?
I am not sure what you mean regarding effect on POSITA.
As for programming, I am no expert regarding AI, but I assumed some issue or problem was presented to DABUS, and that someone had to ensure that DABUS had the input from which to draw a solution. But, I may be wrong.
Until AI reaches the ability to solve general problems like humans do, as opposed to tasks it is assigned and programmed to accomplish, I think the answer is that inventorship should be shared with whoever did the programming. That is, the AI does not enjoy sole inventorship. Now, when AI becomes like VIKI in iRobot, the answer may change.
Even the “non-sole-inventorship” though raises issues NOW.
For example: what does such do to that OTHER non-human juristic person known as the Person Having Ordinary Skill In The Art?
(by the by, the notion that AI “is programmed to” misses the point as the actual inventive aspect is on the table as exceeding that “programmed to,” which is one reason why recognizing that humans involved cannot themselves satisfy the legal definition of “inventor” is part and parcel of the current discussion.)
I am not sure what you mean regarding effect on POSITA.
As for programming, I am no expert regarding AI, but I assumed some issue or problem was presented to DABUS, and that someone had to ensure that DABUS had the input from which to draw a solution. But, I may be wrong.
The reference to PHOSITA has to do with handling of “patentability” of those inventions not perpetrated by a human inventor.
This touches on your second uncertainty – presenting a problem and ensuring inputs do not rise to the level of a human meeting the legal requirement of being an inventor.
Pardon rePeat posts (count filter)
The reference to PHOSITA has to do with handling of “patentability” of those inventions not perpetrated by a human inventor.
This touches on your second uncertainty – presenting a problem and ensuring inputs do not rise to the level of a human meeting the legal requirement of being an inventor.
From the decision below:
If one asked “how many individuals were onboard the Millenium Falcon when it left Mos Eisley?”, what answer would you give? It does not seem crazy to my mind that one might include C3PO and R2-D2 in the count.
But if I asked you how many individuals were aboard Luke’s X-wing during the assault on the Death Star, would you include the targeting computer in your count?
I don’t believe anyone here is saying that no AI could ever be an inventor, but rather that it is implausible that this “AI” is an inventor.
I don’t believe anyone here is saying that no AI could ever be an inventor, but rather that it is implausible that this “AI” is an inventor.
The court below said that—as a matter of U.S. law—no AI could ever be an inventor. That is the point in discussion. Does the word “individual” in Title 35 categorically exclude AI from inventorship, absent a statutory amendment?
As much as you like to try to ‘shoo away’ Stanford v. Roche, that case very much informs the issue here.
Hint: statutory amendment may well NOT be enough for the US Sovereign.
Does the word “individual” in Title 35 categorically exclude AI from inventorship, absent a statutory amendment?
Yes. Yes it does.
If you want to change the law, talk to Congress or the Supreme Court. Personally, I don’t think they’ll be receptive.
FWIW, I agree. I do not think that current U.S. law permits an AI inventor (although I am less convinced that the plain meaning of “individual” is what precludes AI).
I also doubt that Congress would be much interested in changing that if one asked them specifically to amend the statute clearly to allow an AI inventor to be listed. I could see Congress, however, allowing a corporation to be an “inventor” in a work-for-hire situation, much as Title 17 currently allows for “authors” in work-for-hire arrangements.
I could see Congress, however, allowing a corporation to be an “inventor” in a work-for-hire situation
Congress already made the split between inventor and applicant.
Fair enough. I’m clearly out of my depth because I don’t even know if the meaning of the term in the statute is anchored to any particular timepoint.
But I’m already looking forward to the next debate about whether a Borg hive-mind can be an inventor under current law.
>It does not seem crazy to my mind that one might include C3PO and R2-D2 in the count.
That’s…not…possible… mostly because it would make Luke an “enslaver” (in woke parlance)
That’s…not…possible…
Impossible and impolitic are not the same thing.
I’ll forward your edits Mr. Lucas.
That said, George himself was “impolitic” on this point. He shows robots being tortured, which implies they aren’t mere machines. But he also depicts robots as being owned, which makes their owners evil per se.
So, by process of elimination, is Darth the real hero of the story?
George himself was a bit impolitic on this point. He showed robots being tortured, strongly implying they were sentient. He also shows most of the galaxy “owning” one at some point, thus making those characters evil per se.
Amusingly, by process of elimination, this reasoning makes Vader the “least evil” being in the galaxy. Heck, you could justify the destruction of Alderaan as just killing a bunch of enslavers (see also punching Nazis).
Circling back to the main subject, if/when AIs are recognized as “persons,” then their owners automatically become “enslavers” i.e., horrible people. Moreover, anyone who uses products made from slave labor is complicit in said slavery. I’m not sure Thaler, much less your typical iPhone owner, is going to welcome the logical implications of this lawsuit.
You don’t need to go to the future or even call out AI for iPhone slave labor.
If one asked “how many individuals were onboard the Millenium Falcon when it left Mos Eisley?”, what answer would you give?
Well there were two force sensitive people there and the force binds and unites all things, so I’m going with one less than the general viewpoint.
The mention of the inability for AI to execute an oath reminds me that the same argument was used to deny slaves patents. See link to historynet.com. At the time, slaves were thought of as much less than free persons. Now, AIs are also thought of as less than individuals.
As people (not you in particular) seem to like to forget things, let me remind all that I was the first to provide the historical analog to the patenting (by others) of the inventions of slaves.
Out of curiosity, what did Thaler submit for an oath/declaration from Dabus?
Presumably, a substitute statement + the “is under legal incapacity” option. If so, did he get some other court/agency to appoint him legal guardian?
FWIW, one notes that Stanford v. Roche was not cited even once in the decision below. Presumably that is because Stanford v. Roche does not actually touch on any legal points relevant to the dispute in this case.
Lol — sure, keep flailing there Greg.
It’s not as if you have any credibility left to lose.
Any of the Thaler supporters here want to dig into 2015/0379394 and really explain to us how DABUS works?
How would that be pertinent?
How would that be pertinent?
To perhaps explain how DABUS has (allegedly) made such a huge jump in AI capabilities.
Not pertinent — let’s stay on point, eh?
Not pertinent — let’s stay on point, eh?
To a summary judgement motion — no.
However, in the very unlikely event that Thaler defeats the summary judgement motion, then the facts will be at play. I know you have tried really hard to avoid talking about the real facts (as opposed to the asserted facts), but any ruling that DABUS is an inventor is going to be relying upon real facts, which likely includes an explanation as to how DABUS really works.
Now you are being disingenuous with “ I know you have tried really hard to avoid talking about the real facts ”
Real facts are as currently presented – this may differ from your version (you really just want different facts).
God knows how long this will stay in the filter …
Real facts are as currently presented
They aren’t real facts. They are alleged facts. You either don’t understand the difference (this I doubt) or you are intentionally avoiding the issue.
BTW, what I want (in terms of facts) is irrelevant to my position(s).
I have a certain understanding of AI based upon my own personal experience. And based upon that personal experience, I doubt that the conclusion alleged by Thaler (i.e., DABUS is the inventor) is not supported by the real facts. The problem is that Thaler doesn’t explain why he believes DABUS is the inventor. As far as I know, Thaler isn’t a patent attorney. From my experience, lay-people usually have a very poor understanding of patent law. As such, while Thaler may think DABUS is an inventor based upon certain facts — those facts, if shared with a patent attorney, might cause the patent attorney to think that Thaler was the inventor. However, we cannot know any of this until underlying facts of what really happened is shared by Thaler.
These facts are important because what Thaler is alleging is beyond the capabilities of AI, as I know it. I’ll repeat the quote attributed to Carl Sagan — “extraordinary claims require extraordinary evidence.”
Wt,
Looks like items in the filter were (selectively) cleared – not just on this thread, but on more than a half dozen of them.
As to “real” versus “alleged” facts, I use the accepted meaning of the word in which facts are first alleged, and then when they are accepted by all parties they are taken as true.
It is quite clear that you simply want a different set of facts — one that has NOT EVEN BEEN alleged. YOU want to try to make a point on facts NOT present. I fully “get” why you FEEL this way, but your feelings simply are not in the case.
Trust me – I simply have the better position here.
Not pertinent — let’s stay on point, eh?
A valid way of avoiding the negation question would be to conclude that no AI could ever make a nonobvious invention. You don’t need to decide negation if they categorically only produce noninventive results. The court isn’t likely to say that, but they could.
Oh wait, you’ve been more than clear that you do not WANT to stay on point (8.2 below and prior).
“Step 1: Assert Facts.
Step 2: Deny relevance of discussing basis of Facts.
Step 3: Win Argument.”
Is this the sort of argumentative skill I could hope to learn in law school, WT?
Did you even come close to beginning to establish any relevance of the “facts” that you want to offer?
Here’s a hint Ben – under Evidence, facts must be pertinent. You are not even close to being there, and as such, you have not even established “an argument” that could be won.
<i.“[p]atentability shall not be negated by the manner in which the invention was made.”
This statement is inapplicable. The patents are not be rejected based on the manner in which they are made. Rather, they are being rejected because the inventor refuses to list himself as the inventor.
Expressly wrong: “because the inventor refuses to list himself as the inventor.”
Have you been following along? At all?
Genuine question: in the US is inventorship a question of fact, or of law?
In reversing the initial judgment that DABUS could be named as inventor, the appeals court in Australia made an interesting point, observing that inventorship is a question of law, and that given the agreed ‘facts’ the question of whether there might be a human inventor (i.e. Thaler) had not been considered.
In truth, we do not know how these inventions were devised. We only know how Thaler asserts that they were devised. It is a strange thing when a court is asked to assume a ‘fact’ that is, conventionally, not a fact, but a question of law underpinned by a factual analysis that cannot be carried out because the relevant facts are not in evidence.
Patentability is not being negated, in this case, by the manner in which the inventions were made. Automated generation is a perfectly valid manner of invention, just as are discovery or serendipity. Rather, progress of the application is being negated by the stubborn refusal of the applicant to submit the true facts of the invention’s devising for proper evaluation by the court, such that the question of inventorship can be properly resolved.
I am not sure this is correct in all contexts, but the Federal Circuit has deemed that “Inventorship is a question of law entitled to de novo review by the Federal Circuit.” Vapor Point LLC v. Moorhead, 832 F.3d 1343 (Fed. Cir. 2016)
Patentability is not being negated, in this case, by the manner in which the inventions were made.
There is patentability and then there is inventorship. These are different concepts and are reflected in to two different questions on appeal presented by the parties.
BTW — under summary judgment, all facts are to be construed in the manner most favorable to the non-movant (i.e., Thaler).
Thaler has presented the question of “whether an AI-Generated Invention is patentable” because patentability has nothing to do with inventorship. Conditions of patentability are 102 and 103. Patentability doesn’t care about who invented but rather what was invented as compared to the prior art. It is a clever way to present the issue since Thaler almost automatically wins if that is the issue to be decided as the presumed facts are that the invention itself is patentable (regardless of the devisor).
While clever, Thaler’s arguments are an attempt to rewrite what the District Court held. The issue before Court was not patentability. Rather, the issue before the Court was inventorship, which is reflected in the USPTO’s proposed question on appeal: “Whether ‘an artificial intelligence device comprised solely of source code can[] qualify as an ‘inventor’ under the Patent Act.'”
Scanning through the District Court memorandum decision, they determined the the term “individual” refers to a “natural person.” They also determined that the “policy considerations” presented by Thaler aren’t enough to overcome the statute’s plain language. In essence, the Court said that if want to argue policy considerations and to expand the scope of the meaning of inventorship, then take it up with Congress.
As a matter of law, I think the District Court nailed it down pretty tightly, and I very much doubt the Federal Circuit is going to want to breathe any life into this. I’m wondering if the Federal Circuit will choose to put the final nail into the coffin by Rule 36’ing this.
As a matter of fact, and the facts are not being debated here, I highly doubt that what DABUS did raised to the level of inventorship. This is an opinion I’ve made known in the past. This isn’t to say that AI cannot at one time be so capable. I just don’t believe that time is now or any time in our near future. I see current AI as a tool used by humans. It isn’t sophisticated enough to be anything more.
As for the “policy considerations,” my opinions should also be known. I see no pressing need to muddy the patent law waters by introducing an issue (AI as inventor) that has the potential to do more harm to patent law than good.
I would be interested in discussing:
i) the current law
ii) the facts of AI as I understand them
iii) the policy considerations Congress may contemplate in allowing AI to be an inventor
iv) the underlying motivations of Thaler
However, I’m not interested in discussing the implications of what happens after AI is allowed to be an inventor.
[T]he District Court… determined the the term “individual” refers to a “natural person.”… As a matter of law, I think the District Court nailed it down pretty tightly…
What lines of statutory text would you point to as establishing that “individual” conveys “natural person”? How do these lines of statute rule out broader readings than “natural person”?
Also expressly wrong: “stubborn refusal of the applicant to submit the true facts of the invention’s devising for proper evaluation by the court,”
Just because what has been provided is not liked is just not the same thing as “stubbornly [refuse] to submit true facts.”
The hot issue is whether under the present law, and given the assumed fact that DABUS is indeed the inventor, that invention by DABUS can be accepted as patentable under the existing law. At the court of appeal in England, the three judges split 2:1. Will there be a split at the Federal Circuit, I wonder.
Apart from all that though, does anybody here seriously deny that, looking into the future, the patent statute has to be able to recognise as patentable inventions made by AI’s?
To your last question, are you willing to recognize the Lockeian nature of patent law?
If not, then I suggest that you bow out of the conversation — leastwise as it pertains to US Sovereign law.
“has to be able”? Why would it have to be able? Assuming that AIs are not recognized as inventors, people won’t use AIs for getting patents. They would still be used for problem solving. The only possible loss would be the value, if any, of fewer patent publications.
One of the policy points (put forth by others), NS II, is that if patent protection is indeed denied, then the protection available would be Trade Secrets, and AI already has enough issues with lack of traceability (technical, as well as poli-philo-equitable).
“we should probably be troubled if the result is that “two otherwise patentable inventions cannot receive patent protection””
Only if we’re literally begging the question.
Actually no — this is NOT an instance of begging the question.
It’s simply that answering THAT question is not dispositive. Pro Say at 2 below is actually much closer than you at point.
“we should probably be troubled if the result is that “two otherwise patentable inventions cannot receive patent protection”
One would only be troubled had they not looked at the inventions attempted to be claimed. Of all the patents/applications I have ever been aware of in my professional capacity (that number is probably close to 10,000 based upon both what I am prosecuting and the art cited therein), I would say that the two Thaler patent applications are bottom 1% in terms of disclosure and utility. Actually, I would be comfortable saying bottom .1% and even then I would be hard pressed to recall 10 patents/applications that were worse. Bottom 0.1-0.2% would be about accurate in my opinion.
Moreover, if we were to actually investigate the real facts, I suspect that these “two otherwise patentable inventions” could easily have Thaler as the named inventor.
“I suspect that these “two otherwise patentable inventions” could easily have Thaler as the named inventor.”
Exactly my point. There’s nothing to be troubled if we don’t prematurely assume he is not an inventor.
Wow – the illiterate are demanding facts to be “convenient.”
What could go wrong with that?
the illiterate are demanding facts to be “convenient.”
Don’t confuse facts with alleged facts.
Procedurally, the District Court presumes that the alleged facts to be true. However, there are a great many of us who doubt that the alleged facts are true. While you are comfortable ignoring that issue, many of us are not.
How is it “an issue?”
Unless you are challenging existing court procedures, you have no case.
But please — you have already made it clear that you are not interested in discussing the legal points AT point – why are you still around?
you have already made it clear that you are not interested in discussing the legal points AT point
You have already made it clear that you are not interested in discussing the facts/evidence that underlie the legal issues.
It is “convenient” for you to sweep those facts/evidence (or lack thereof) under the table.
Again – it is you that simply do not like the present facts and want different ones.
I “get” that you don’t like Thaler and that he is merely gaming the system with false assertions that he is not the inventor. But those are not facts at hand. Those are only facts as you would want them to be.
Again – it is you that simply do not like the present facts and want different ones.
Too bad this comment will get filtered for some indeterminate amount of time.
Let us remember what the good professor stated elsewhere: “Inventorship is a question of law.” Thaler has stated that DABUS is the inventor. This is not a statement of fact but a conclusion of law by a person that is unqualified to make such a conclusion.
You keep talking about the “present facts” as if they exist. They don’t. Thaler has made allegation as to a conclusion of law. He hasn’t presented any facts as far as I know.
Regardless, for purposes of the Federal Circuit’s review, the “facts” are not relevant for reasons I stated elsewhere (i.e., the facts are going to be viewed in a light most favorable to non-movant). The facts (whatever they happen to be) pertaining to whether DABUS actually did the inventing would become relevant only if Thaler was to prevail on the legal issue as to whether the patent statutes allow an AI to be an inventor.
This is the order by which DABUS could become an inventor.
1) Federal Circuit permits an AI to be an inventor.
2) Thaler presents facts at the District Court showing that DABUS met the normal requirements for inventorship.
3) Federal Circuit will have to agree that DABUS met those requirements based upon the facts found by the District Court.
Only until those three things happens will DABUS become an inventor and your desired conversation of the impact of such an event be relevant. However, none of those things will come to pass — and my prediction has nothing to with what I want or don’t want. For example, I have long predicted that Congress and the Court will do nothing of import with regard to 35 USC 101 despite my fondest wishes that they do so. I can separate my wishes from my predictions. In this instance, my prediction is that Thaler’s goal of AI inventorship is dead as a doorknob.
The present facts DO exist.
YOU keep on wanting a different set.
And sure, the question of law that stems from the facts remains a question of law (a bit of a strawman from you as I do not suggest otherwise).
Procedurally, the District Court presumes that the alleged facts to be true. However, there are a great many of us who doubt that the alleged facts are true. While you are comfortable ignoring that issue, many of us are not.
Ok, but pause and consider the implications of assuming the alleged facts. If the CAFC decides this issues on the assumptions implied by the procedural posture of this case, that does not mean that DABUS gets the patent. Rather, if the CAFC vacates the summary judgment on this record, then that just means that they remand it back to the district court, who can conduct additional fact-finding during the trial that never happened by the summary judgment stage.
In other words, there is ample opportunity for the case to end with the district court finding the facts that you believe that they will find. We can only get to a fact-finding as you contend, however, if first we get past the present summary judgment disposition of this case.
In other words, there is ample opportunity for the case to end with the district court finding the facts that you believe that they will find. We can only get to a fact-finding as you contend, however, if first we get past the present summary judgment disposition of this case.
I agree with just about everything your wrote except the “ample opportunity for the case to end with the district court finding the facts that you believe that they will find.” To get to this step, the USPTO needs to lose at both the Federal Circuit and the Supreme Court. One, the USPTO isn’t going to lose at the Federal Circuit. Two, if by chance the USPTO did, they would certainly appeal it to the Supreme Court, and this is something too juicy for the Supreme Court to ignore. Three, there is no way that the Supreme Court would endorse a position that an AI could be an inventor. Hence, there is not “ample opportunity” for the District Court to find anything. This case ends at the Federal Circuit. Thaler may try to go en banc or to the Supreme Court but they’ll shoot him down immediately.
We can only get to a fact-finding as you contend, however, if first we get past the present summary judgment disposition of this case.
Absolutely true, but it isn’t going to make it that far — despite the fondest wishes of certain people here.
To get to this step, the USPTO needs to lose at both the Federal Circuit and the Supreme Court… Hence, there is not “ample opportunity” for the District Court to find anything. This case ends at the Federal Circuit.
Good point. I share your expectation that the CAFC will rule for the PTO.
Thaler will surely petition for cert. at that point. I have no firm guess as to how that will end. The probabilities, however, surely favor the idea—as you suggest—that the summary judgment below stands.
If Thaler wins, I look forward to follow on inventions by optimization functions and calculators.
“Dr. Thaler’s DABUS
I did chuckle – but Thaler will not win here.
Thaler will not win here.
On that, you and I can agree.
For a victory, the following is likely needed:
i) a plaintiff who isn’t leveraging the issue as part of an underlying goal to elevate AI into personhood
ii) important and valuable inventions at issue
iii) evidence that additional important and valuable inventions can be had if AI is granted inventorship
iv) evidence/arguments that these important and valuable inventions won’t be disclosed if AI is not granted inventorship
None of these factors are present here. Thaler seems to be a few cards short of a full deck, the inventions are lousy, and there is no evidence that good ones are forthcoming. He hardly makes for a sympathetic plaintiff.