IP and Human Creativity in the AI Age: (Panel 1) AI’s Challenge to Inventorship

ROBERT MERGES: We're hearing from all quarters today.

The idea, of course, is that there's a lot of developments around IP law impacted by artificial intelligence technology.

And they're not just happening in the district court in San Francisco and New York, they're happening all over the world.

And some jurisdictions are actually deciding for the first time, some of the first substantive cases in the IP field.

So we're starting to see opinions that dig down below the initial level.

We're starting to see some more doctrinal opinions about how do you enable a patent.

That's an invention, that's AI assisted.

And we're starting to see some decisions on how will we determine patentability in the AI era, what kind of prior art is relevant, et cetera.

We're going to talk about all that.

But the microcosm is that this conference, put together by three different pieces of the Berkeley Law world.

Berkeley Center for Law and Technology, of course.

And the course that we teach Dr. Hao and I on IP and human creativity is under the BCLT umbrella.

And we're in our third year on that course.

And it's been really eye opening for me.

The other sponsor-- the second of 3 is the Robbins Collection here at UC Berkeley, which if you were in the world of civil law and canon law scholarship.

If you were somebody who needed to see a medieval commentary on Roman law, you would know the Robbins Collection because they have one of the world's great collections of fundamental texts in civil law, in canon law.

And we've been looking to partner with them for a long time.

My long-term colleague, Laurent Maile, who's here.

Laurent, you want to just give us a wave?

[APPLAUSE] Yeah.

[APPLAUSE] I've been trying to figure out a way-- I respect him a lot.

And we've been colleagues for a long time.

For a long time, I've been trying to figure out a way for us to do something together.

And it turns out that this space is perfect.

Because a lot of developments are happening in many different kinds of jurisdictions.

And the civil law focus of the Robbins Collection makes a lot of sense.

And then just in talking to him today, he was reminding me that because we take of a human creativity angle, certainly in our course.

And we really think about the impact of AI on human creation and the human creative professions.

Laurent was reminding me that the birth of humanism was really one of the generative moments in the Renaissance and in late medieval times.

Particularly, he was talking about the 14th century.

He said that was the first time people really began to think about human creativity and move it out from a purely theological umbrella into a kind of humanist focus.

So I think the overlap is even stronger.

The last component is we wanted to hear from as many Asian countries as we could, because AI technology is taking off very quickly.

And so we brought in also the Korea Law Center at the law school to help us fund some people to come in from Korea.

And today we have Judge Kim from Seoul, and we have a number of Korean practitioners who are here.

And we're very happy about that because as I say, AI technology is really popping up everywhere, and we don't want to limit our focus to the handful full of US cases, which so far as are clustering around copyright, in particular fair use.

Not that it's not important.

Sometimes I say that copyright the field should be re-titled fair use studies.

But I think there are other forms of IP that are going to be relevant.

And our last panel-- I'm especially interested in because if we think about moral rights, the right of publicity, the softer IP rights connected with reputation and artistic style, I think AI technology is going to be a very serious challenge in those areas.

And so we're trying to get out ahead as well as cover our bases.

I don't want to take too much time because we have so much of substance to talk about a couple of housekeeping details.

We will have a break at 10:25.

This is an old building.

There are bathrooms scattered around.

You can find them.

They do work, if you are patient.

And the other thing that I wanted to mention was that the-- so our first panel is going to go right up until that break.

And then we have several in a row.

So if everybody could come back from the break promptly, we can get started on that second part.

So to summarize the intro, I'm really happy everybody's here.

Many of you have come a long way, and even those of you who only came over the Bay Bridge fought a noble battle to get here.

Maybe it's not the same as a 12-hour flight, but sometimes it feels like it.

So I'm really grateful that people made the effort, and I hope that we have a good and productive day.

So I'm actually going to allow Panel 1 to get started.

Dr. Hao will introduce everybody on the panel.

But I should introduce her.

She's our Senior Research Fellow at BCLT, has taught Patent Law and IP Law at Tsinghua University in Beijing.

Sometimes co-teaching with the Federal Circuit Former Judge Randy Rader.

Many of you know.

And she's been with us now several years running the Berkeley Antitrust and IP Competition Center, and she's been co-teaching the Human Creativity and AI course with me, not just co-teaching.

She conceived it and reduced it to practice.

So she should be the sole inventor on the patent for that course.

I'm just a tag along.

And she will moderate this panel and hopefully during the day, introduce some of her own research and thinking on AI, which has been revelatory to me.

I think it's a very high-end contribution.

And I've been very pleased to participate.

So, Dr. Hao, let me turn over Panel 1 to your hands.

[APPLAUSE] Thank you so much, Rob.

That's very kind introduction.

Again, my name is Yuan and it's wonderful for me to be here today with you guys.

I'm going to be the moderator of the First panel, AI challenges to inventorship.

I'm thinking because we have an audience with diverse background, and some of whom may not have been very familiar with patent law and AI challenges to inventorship.

So maybe I'll just spend a few minutes, doing a very brief introduction of what we are talking about here today on this panel.

So if you are patent experts, please just bear with me for a few minutes.

So at the very outset, what is so special about state-of-the-art AI today?

So the AI that marries our lengthy discussion of these challenges to various patent doctrines.

It's special because in the first time in human history, we now have a tool that's capable of making significant contributions to the technological solution.

And sometimes this contribution would be so significant that assuming the AI were a human being, and no doubt she would have been found as an inventor.

Thus, should we assign the honored position of inventorship or co-inventorship to the machine then?

And in other situations, it may even become a question that even though there's human beings involved in this process, but the human being's contribution may not be sufficient to merit a inventorship under the traditional inventorship doctrine.

So what do we do about those?

So the way I see it, I call the latter situation inventory list inventions.

Shorthanded as II, which is a common challenge faced by multiple jurisdictions in the near future.

So the way I see it, there are two basic scenarios of the II situation.

The first scenario is illustrated with the very famous starburst case.

I'm sure several panelists will talk about that case.

If you're not familiar with that case, it's this owner of an AI system called DABUS, and the owner is Dr. Steven Thaler.

And he claimed that all the inventing work has been done by the AI, and he filed many patent applications around the world claiming the AI is the sole inventor.

And this generated a series of cases, both judicial cases and regulatory, in some jurisdictions around the world.

Despite it's fame, I still think, personally, DABUS might belong to the easier wave of inventorship challenges.

Because if we take the facts on its face value, there's really no human being involved in the process.

Because Dr. Thaler claimed that AI DABUS did every bit of the job, which we might touch on this later.

But I don't think it's entirely consistent of with my own understanding of the state of the art AI.

But leaving that aside, this is still an easy case.

Also because the regulatory branches and courts around the world, they were actually asked to just do a statutory interpretation of what constitutes inventorship based on their respective laws.

But the harder case would be next wave.

The legislature people might face the more daunting challenge that whether naming AI inventors would be consistent with deeper IP principles.

So all of these are fantastic issues and also very challenging issues.

But we have several experts from multiple jurisdictions here, Chris, from US and-- I'm sorry.

Dr. Mammen-- Chris Mammen from US.

And Mr. Chris De Mauny from UK, Mr. Yoon Woo Shin from Korea.

And also, we are very privileged to have Ms. Nalini-- Justin, can we maybe play.

SPEAKER: [INAUDIBLE] YUAN HAO: OK.

Hi Nalini.

Good morning?

--from USPTO.

I'm going to do a very brief introduction of our panelists.

Then we will start the discussion.

The last housekeeping point is that we're going to divide the panel into two sessions.

In the first half an hour or so, we're going to stay on the doctrinal level.

So our panelists will inform each other and you guys about the status quo of their jurisdictions.

What's going on there.

Then depending on the flow, I might push back with just one or two questions to facilitate the discussion.

Then hopefully after this round, we still have 20 minutes or so to go a little bit beyond the doctrinal level, diving into maybe some of the deeper normative principles, whether through this lens, how this would inform people's understanding of the AI challenge to inventorship.

So I'm going to do just a very brief introduction.

Nalini is a Senior Legal Advisor in the Office of Patent Legal Administration at the USPTO.

In this position, she drafts rule packages, memoranda, and associated guidance on topics such as subject matter eligibility under 35 USC 101, and disclosure requirements under 102.

She also coordinates the development and implementation of patent examination guidance related to artificial intelligence and other emerging technologies at the PTO.

And sitting on my left-hand side is Dr. Chris Mammen.

Between law school and entering law practice, Chris Mammen earned a doctorate in legal philosophy from Oxford.

And for the past 30 years, he has practiced IP litigation in the Bay Area, and on a recent sabbatical from practice returned to Oxford as a visiting academic while he convened a multidisciplinary working group on AI, IP, and creativity.

And now next speaker there is Mr. Chris De Mauny.

He is a patent litigation specialist based in San Francisco.

He advises clients from a range of industries, including tech, life sciences, and sustainable technologies.

Chris is licensed to practice law in England and Wales, and registered as a foreign legal consultant with the State Bar of California.

And last but not least, Mr. Yong Woon Shin is an attorney at JIPYONG Law firm in Korea.

He handles a wide variety of ICT and IP cases, and also serves on the special committee for responding to artificial intelligence and legal tech at Seoul Bar Association.

All right.

So I can't wait to start our US perspective.

So each panelist firstly has six minutes or so to introduce the status quo in their respective jurisdictions.

And after that, we might have a little discussion.

Nalini, are you ready?

NALINI MUMMALANENI: Yes.

Can you hear me?

YUAN HAO: Yes.

NALINI MUMMALANENI: Awesome.

Thank you, Dr. Hao.

I will talk about the AI and mentorship from the US standpoint.

So a quick overview of the inventorship laws in the US.

If you look at the definitions 35 U.S.C 100(f) and g, inventor or co-inventor or joint inventor are defined as an individual or individuals who invented the subject matter of the invention.

35 U.S.C 101 says, whoever invents or discovers suggesting a natural person.

And 35 U.S.C 115 116 require the patent application to include the name of the inventor or joint inventors.

And also requires inventor or joint inventor to execute an oath or declaration.

So language of 35 U.S.C 115 and 116 also refers to individual himself or herself, person, et cetera.

Indicating that the inventor or joint inventors are natural persons.

So we can conclude that the statute requires the naming of the natural persons who invented the claimed invention, irrespective of the contributions provided by advanced tools like AI.

And there is also no section in the Patent Act that supports the position that inventions that are created by natural persons using specific tools like AI systems, result in improper inventorship or are otherwise unpatentable.

Now, the Federal Circuit has consistently made it clear that conception is the touchstone of inventorship.

This is true in the case of joint inventorship context.

Also, where a person who shares in the conception of a claimed invention is a joint inventor of the invention.

In joint inventorship scenarios, the inventive entity for a particular application is based on some contribution to at least one of the claims made by each of the named joint inventors coming to USPTO Practice, The USPTO generally presumes that the named inventor or joint inventors in the application are the actual inventor or joint inventor to be named on the patent.

Now, Dr. Hao had been talking about this DABUS case.

We encountered the issue of AI inventorship in 2020.

In April 2020, the Office received a couple of petitions in two patent applications where an AI system was named as the sole inventor.

As you may be aware, the AI system was named DABUS.

In the petition [INAUDIBLE], the office explained that under current US patent law, inventorship is limited to natural persons.

And this case went all the way up to the Federal Circuit.

And there the Federal Circuit said based on Supreme Court precedent, the court referred to the definition in 100(f) et cetera, found that individual is a natural person unless Congress provided some indication otherwise.

And in this case, Congress provided no such indication.

The court concluded that the Patent Act supports the customary meaning that an individual is a human, and the court explained that it was not confronted with the question of whether inventions made by human beings, with the assistance of AI are eligible for patent protection.

The Supreme Court denied cert in this case.

And following this case, we revised the manual of patent examining procedure.

For those of you who are practicing before the USPTO, you know about this in accordance with this Thaler additions and clarified that an inventor must be an actual person.

And by extension, any joint inventor must be an actual person.

And patent applications naming a machine as an inventor or joint inventor will be considered by the USPTO as having improper inventorship.

And in Thaler, we dealt with AI as the sole inventor.

Generally, inventors use AI systems in different capacities and different points during the invention process to arrive at their inventions.

So we had to address the situation where AI is used by natural persons in the invention creation process.

While the Thaler case was going through the court system, USPTO conducted a many stakeholder engagement activities like requesting RFCs, requesting public comments.

We also held public listening sessions on this issue.

And following this, we issue the mentorship guidance for AI-assisted inventions February last year.

The key takeaways from the guidance were that AI-assisted inventions are not categorically unpatentable.

We provided a framework for determining inventorship based on the Federal Circuit decision in Pannu Lab 1997 case to determine whether a human's contribution is sufficient for proper inventorship.

Practically, as AI is used in different aspects or stages of the invention creation process by the inventors, we provided some guiding principles to help the practitioners, et cetera, understand how to use that.

Based on the statutory language of 35 U.S.C 161 and 171, the guidance applies to plant and design patents and patent applications also.

The guidance further discussed potential impact on areas of patent prosecution like foreign priority claims, how you deal with that, what you can expect from old declaration practice, duty to disclose rule 56, et cetera.

And please note that this guidance is currently under review for alignment with administration priorities and policies.

So keep an eye out if there is any changes.

And for your reference, the MPP sections are given here and the link to the Federal Register Notice that I just talked about is also provided.

Back to you.

Dr. Hao.

YUAN HAO: Thank you so much, Nalini.

Really fantastic.

But I know the guidance is currently under review.

Are you allowed to briefly just introduce the five guiding principles or not?

If not, I'm going to skip that.

NALINI MUMMALANENI: Yeah.

So the five guiding principles generally talk about various scenarios, there were more permissive in that they allow for humans to be named as inventors in different scenarios.

Like building the underlying AI systems or finding solutions to problems and fine tuning the output so that they arrive at the invention.

All these things.

But in essence, they provide more guidance as the Federal Register notice has more details about those guiding principles.

YUAN HAO: Oh, sure.

Yeah.

Thank you.

If I can just ask one more question is that-- I think it's very important in this guidance you said that AI-assisted inventions are not categorically unpatentable for improper proper inventionship.

Because the way I observe it-- not to steal our later panelists thunder-- is that we are actually witnessing a lot of case studies of AI-assisted inventions already during the past three or four years.

And two typical industry sectors would be materials science-- material research and drug discovery.

So categorically, if excluding all these AI-assisted inventions outside the patent law protection, to me sounds not very practical at least.

But the thing is that-- again, not to steal thunder from later panelists-- some of our speakers like Professor Arti Rai, did some amazing research regarding case studies of AI-assisted drug discovery, for example.

But found that maybe many of those applications even though there was substantial AI involvement, but because of the unpredictability of the law right now.

So not too many people want to review AI's role in a patent applications.

So from the perspective of a senior patent examiner in the United States, do you think the patent applicants in the future they should disclose that or how do you evaluate the inventorship requirement?

Can you give us some basic idea there.

NALINI MUMMALANENI: Yeah.

So we do not impose any additional duties to disclose the use of AI tools.

We already have existing rules and policies for all types of patent applications.

And AI-assisted inventions are no different.

So to give you an example.

Consider some of the scenarios.

If you look at the inventorship scenario, we consider AI as just another tool, like a microprocessor or a sequencer that's used by natural persons in the conception of the claimed invention.

YUAN HAO: OK.

NALINI MUMMALANENI: A patent application needs to identify only human inventors in the application data sheet.

There is generally no additional duty to disclose information about how these inventors used AI in the invention process, unless it's requested by the USPTO during examination, which I'll go over in a bit.

YUAN HAO: Right.

NALINI MUMMALANENI: If the AI tool or its use is material to patentability, you may need to disclose the required information pursuant to existing rules.

For example, duty to disclose in Rule 56.

Some tips to determine inventorship are given in MPP 2004, where the practitioners can decide who the actual inventors of the application are.

Also, if a claimed invention is itself related to a particular AI system, you may have to comply with the disclosure requirements under 35 U.S.C 112 and that one you are already aware.

Now, coming to the second later part, once you file the patent application, coming to examiners evaluating inventorship.

Typically examiners do not make inquiries regarding inventorship.

The USPTO presumes that the named inventors in the patent application or the patent are actual inventors.

That is the information the practitioners provide to us in the application data sheet or declarations, we presume that they are the actual inventors.

Because of this presumption, the rejections under 35 U.S.C 101 or 115 for improper inventership are usually very rare.

And we are seeing the same trend.

We don't see any heightened trend with regards to AI-related inventions.

But when I said I'll discuss it later, if there is evidence to the contrary, like examiner finds some evidence and the examiner makes a rejection under 101 or 115 for improper inventorship.

Now the applicant can come with appropriate evidence to traverse that rejection.

Just like you traverse other statutory rejections, you will have to provide evidence to traverse rejections.

Here also, there is no change in practice.

USPTO does not prescribe a specific format in which evidence is presented, as long as the objective evidence is supported by actual proof.

So we do not have any changes in practice or anything to deal specifically to disclose use of AI in arriving at your invention.

YUAN HAO: Thank you for the nuanced explanation, Nalini.

That was very helpful.

I think we will just maybe keep it to the-- we will have 10 minutes or so at the end of Panel 1 for our audience to have Q&A, if that's OK with you.

So for now, we'll just proceed with Chris Mauny, who will kindly introduce the UK and EU practice on AI in challenging to inventorship for us.

Chris.

CHRIS DE MAUNY: Thanks very much.

Can I get my slides up on the screen or?

YUAN HAO: It's coming.

CHRIS DE MAUNY: Great.

Thanks.

Well, as indicated, I'm going to try to give some highlights of the position in Europe.

Everybody talks about Europe as if it's a homogeneous bloc, which for patent law is partly true in some respects and very much untrue in others.

So let me start by getting into-- OK, this is a slightly different deck.

But let me start by getting into the position from the European Patent Convention.

So this is the international treaty that effectively serves as the legislative basis for the operation of the European Patent Office.

And for much of substantive patent law among the European Patent convention states.

And what you'll see is Article 61 gives the right to a European Patent, to the inventor, or to a successor in Title.

So that's the foundational principle of the origin of the right to a patent.

Now, in the DABUS case, the issue was-- as already indicated, whether the scope of the term inventor could include an AI system that was said to have done all of the inventing activity.

And when this case was heard by the legal board of the EPO, they stated very clearly that a machine is not an inventor within the meaning of the EPC.

And we'll unpack that slightly on the next slide.

But it's a very clear statement.

That quotation is from the head note.

So that's basically summary.

No, machines are inventors.

And to analyze that position, they turned to Article 61 and identified three functions that are embodied within that short text.

And I think that the second and the third are really where the focus of the case is.

But fundamentally, it all flows from the first function, which is that unless you fulfill the conditions of Article 61, there is no right to a patent.

So you've got to fulfill those conditions for there to be anything to successfully file.

And the analysis-- I've obviously pulled out some very short passages from what is quite a long decision, but the analysis is really very simple.

The interpretation of the term inventor means a person with legal capacity.

That's its ordinary meaning, and there's no reason to think that the legislation should bear any other meaning than the ordinary meaning.

So a person with legal capacity has to be named as an inventor.

If you don't have that, there's no right to the invention.

And there's no way for that right if it existed to flow to another person, such as Dr. Thaler in this case.

So the immediate, more practical question is, well, if Dr. Thaler had approached the matter in another way or in another, less extreme example, could a patent that is derived from an eye inventing activity be filed and successfully designated inventor?

The EPO did address this point.

And the argument was made that if they reached the conclusion that they did, that would treat certain classes of potential inventors or people deriving their rights from inventors unfairly.

And I guess this probably harked back to things like the Paris Convention Equal Treatment, et cetera.

And the board simply said, they're not aware of any case law that would prevent the user or owner of a device involved in inventive activity to designate themselves.

So they don't get into detail as to the level of involvement, what kind of right that-- I mean, they use the word owner.

But in detail what kind of right that person would have as regards the inventing machine, let's call it.

But they acknowledge that in principle, a patent that has derived from activity that involved an AI system could be patentable by reason of having an inventor who is a legal person.

In the UK, this case went-- the corresponding case went all the way to the Supreme Court.

And I'm not going to go through the whole of this slide, but the person thing is at the bottom.

An inventor within the meaning of the 1977 Act, which implements the European Patent convention, must be a natural person.

And DABUS is not a person at all, never mind a natural one.

So same conclusion after three layers of litigation.

What about that more practical, more real situation?

The court dealt with this slightly differently to the EPO.

They said that this is more than a formal objection.

It goes to the heart of the system.

So they took the view that it's not merely about who you name on the form.

In principle, it is actually about having a human person.

And that's a slight, subtly different from how the EPO expressed themselves.

Because the EPO effectively said, if you have an owner or something involved, then you may be OK.

I think one needs to bear in mind, however, is the practicalities of how things work in Europe, which will turn to in a moment, mean that this is unlikely to be challenged in most cases.

If a human inventor is named and said to be the inventor, that is likely to get through.

But what's particularly interesting is that in September, another of Dr. Thaler's applications was analyzed by a first instance appellate judge, having come from the UK Patent Office.

And it was rejected because although Dr. Thaler attempted to name himself as the inventor, the history, both of that application and more generally, indicated that he himself did not believe that he was the inventor.

So I'm not going to go through all of that.

But effectively, he had changed his position both on the suite of inventions that came from DABUS that had all been filed at the same time.

And on this specific application, he had changed his position from, I am not the inventor because I did not contribute, to I am the inventor because I'm a human, as it were.

And when this was challenged in these proceedings, the explanation was that he had changed his position because his understanding of the law had changed following the Supreme Court decision.

But the Patent Office and then the court held that this wasn't persuasive.

Because what he'd actually said was factually, I didn't do anything.

I switched the machine on and the invention came out at the end.

I made no contribution.

And so in that case, it wasn't good enough for him to simply now say, but I am naming myself as the inventor.

It was contradicted by his statements of fact about how the invention had come about.

And I'm going to pause there.

YUAN HAO: Thank you Chris.

It was very informative.

I do have a lot of questions, though.

I might just ask one of them.

Because this is very important for me-- is that, it seems to me that in either UK or Europe, there's no more specific definition of what constitutes inventorship.

In the first place, leaving aside AI.

But in a traditional sense, except the situation, it has to be a natural person.

If you compare it with the US practice-- in the US, everybody knows the conception is the cornerstone of inventorship.

And to constitute a conception, there's a series of case law.

Basically, if I could paraphrase the formation of a complete and settled inventive idea in the mind of the inventor.

And if you look at China, whoever makes inventive contribution to substantive feature of the claims will be considered as the inventor.

So those are all requirements on top of the natural person.

But it's a little bit surprising to me that UK and EU-- why is that?

CHRIS DE MAUNY: Well, there's two parts of the answer to that.

The first is that the issue of inventorship and ownership is not dealt with at European level.

So the EPC does not provide for a definition and the EPO does not deal with the matter.

And nor indeed does the UPC.

So it is addressed at national law level.

In the UK, the case law is mostly derived from issues of entitlement, i.e.

competing parties saying you've filed an invention that was mine or that of which I was a co-inventor.

And the standard is that the inventor must be the actual deviser of the invention or one of the actual divisors.

And because of the nature of the case law, where it's derived principally from these kind of co-inventors or inventorship dispute type cases.

There isn't that much explanation of, if you will, what it means to be the actual divisor.

In the sense of, do you have to have made a contribution greater than a certain threshold.

The case law is to the effect that if you have contributed to the inventive concept, then you are one of the actual divisors.

But it hasn't analyzed in great detail the degree of contribution to be an actual divisor.

As regards other countries, they have their own standards.

And we-- certainly, I can't go through those now.

YUAN HAO: OK.

I think-- CHRIS DE MAUNY: Perhaps I can just add one more comment, which is that-- YUAN HAO: Sure, no problem.

CHRIS DE MAUNY: Because of-- oh, I'm sorry, I can't-- I had a slide about this, but the issue of inventorship only arises in the context of those type of entitlement disputes.

So inventorship can't be used as a free standing challenge to a patent's validity, unless the person making that challenge says they were entitled.

And that's one of the reasons why there's only very limited amount of case law.

But it also means that a DABUS-like situation-- or again, a more realistic one, where there is a person named.

If there's a suspicion that, in fact, all of the inventive work was done by an AI machine and you're a defendant being sued on that patent, for example.

There's no ready way of defending oneself by saying the inventorship is wrong.

This is a misstatement of inventorship, et cetera.

That's simply not a ground for challenging the validity of the patent.

And so, not only is there limited case law, but it's also of limited utility in patent litigation to start looking at inventorship.

YUAN HAO: Oh, it's very interesting.

I mean, my understanding-- the patent entitlement cases in the UK are somewhat similar maybe to the patent interference cases in the US, but not completely because it's less strict.

I'm sorry, just a quick-- CHRIS DE MAUNY: I mean, I don't understand the full ins and outs of interference's.

But in a broad sense, yes, it's about typically a situation where there's been a collaboration and one party to that collaboration has filed a patent and another party says I should be either the owner or a co-owner of that patent.

YUAN HAO: OK.

I mean, because of the time, we'll just continue that discussion offline.

But fascinating introduction.

Mr. Shin, would you please introduce the practice of Korea to us.

YONG WOO SHIN: Yeah.

Thank you, in changing the slides.

Yeah I introduced the Korean perspective on AI inventorship.

And I would like to introduce-- firstly, the Korean Intellectual Property Office's position.

And it's similar to the other jurisdiction.

And no, AI inventorship is admitted.

And also, the Korean Patent Office and the court does not currently recognize AI as an inventor.

On the legal basis, the Korean Patent Act is interpreted to remit inventorship to natural persons.

And there has been no official institutional discussion on AI inventorship since the DABUS case.

And the DABUS case is adjudicated in Korea, and the Seoul high court decides there is no-- in AI, inventorship is not automated.

And the case is pending in the Supreme Court.

And the Korean Patent Act indicates that the inventor is only human, and Article 33 in Korean Patent Act, that any person who has made

an invention or his or her successor shall be entitled to patent.

So in the legal interpretation, the term person is consistently interpreted by people in the Patent Office and Korean courts to mean a natural person only.

And the career follows the Civil law system.

So the court and the government should follow the statute of the act.

And I consulted the senior official in KIPO-- the Patent Office.

So KIPO is now studying the degree of human contribution required.

And if a human using AI does not make a significant contribution to the envisioned conception, the person also cannot be an inventor is referenced from the Pannu Case.

And the research goal is to establish standards for the level of human contribution needed for recognition in AI system assisted inventions.

So the study of the Korean government is an annual timeline.

So research results are expected early next year.

So I would like to share the results of this research and to you soon.

And in addition, as I said before, the patentability gap would arise if neither the human nor the AI qualifies as an inventor.

So it seems to result in that, the invention cannot be patented.

If the patent-- application is disclosed, the invention may enter the public domain.

If it was not disclosed.

Yeah.

If it were disclosed during the patent application process and if not disclosed, it would be [INAUDIBLE] patent patentability gap would arise, or a later independent human inventor

could still obtain a valid patent for the same invention.

And so.

in the perspective of practical examination, I think it is crucial to determine if a valid inventor exists prior to disclosure.

And the other issue is through examination of AI-assisted patent.

And it's difficult for patent examiners to distinguish human versus AI contributions.

So examiners often cannot separate the human input from the AI's output based on the application alone.

So we can think the temporary solution.

Some ideas from the other act.

The framework act on artificial intelligence.

it was enacted in Korea or the EU, and it will be activated next January in Korea.

And in the framework act on artificial intelligence, imposes the transfer transparency obligations on AI providers.

It requires indicating when an output has been generated by Gen AI.

And the method of indication is not confirmed in the raw, so the individual watermarking can be possible.

And this may help examiners distinguish the human's creative input during examination.

If the demand for human use the AI of the platforms of such obligations.

And the next is the other, so [INAUDIBLE] do I proceed or the remaining [INAUDIBLE] the other issues?

YUAN HAO: Sorry.

Do you what?

YONG WOO SHIN: So.

Do I proceed or?

YUAN HAO: [INAUDIBLE] in time.

Oh sure.

Maybe we'll just stop here for a little while because of the time.

But very interesting.

Maybe I can ask some questions to clarify just one or two points.

Because that's very informative introduction, Yong Woo.

But I noticed that two points very important.

My understanding is that the two points out of many required by this framework AI Act.

One is that the so-called patentability gap.

You said, maybe it's similar to the inventor inventions I mentioned earlier.

You said that, it has to be-- if it's disclosed, it enters public domain.

Was that required by the AI Act as well?

Is that a binding act or is just being proposed.

And was the act passed already?

The AI Act?

YONG WOO SHIN: Is kind the patent [INAUDIBLE] issue.

Patent Act.

Yeah.

YUAN HAO: OK.

So maybe my understanding was wrong.

The patent ability gap inventions entering public domain, that was not required by the act.

That was the practice.

[INAUDIBLE] YONG WOO SHIN: It's hard to say.

YUAN HAO: A position by the IP office or.

YONG WOO SHIN: Yeah.

Due to the IP rule and the Patent Act, if there's something of the application of the patent is disclosed in any way, so the patentability

is lost due to the Patent Act.

So-- YUAN HAO: Due to [INAUDIBLE]?

YONG WOO SHIN: Yeah.

So-- SPEAKER: It becomes prior art.

YONG WOO SHIN: Prior art.

YUAN HAO: OK.

Yeah but that act has been passed is binding or it's just still being discussed?

SPEAKER: [INAUDIBLE] YUAN HAO: Next.

SPEAKER: [INAUDIBLE] YONG WOO SHIN: No, the disclosure seems to be not discussed yet because it's the doctrine or dogmatic is the principle of Patent Act.

So the research now on process is just [INAUDIBLE] whether the person using the AI can be an inventor.

Yeah YUAN HAO: Right.

YONG WOO SHIN: In regard to the significant contribution.

Yeah.

YUAN HAO: OK.

OK.

Thank you so much.

So let's move on for now because we have many fascinating stuff to talk.

So this is for Dr. Mammen Chris You have given a lot of thoughts creativity and AI.

So the interplay between them.

So maybe you could spend six minutes or so introduce some of your thoughts, looking through this lens.

CHRIS MAMMEN: Sure.

Yes.

Happy to.

And if we could pull up the next set of slides, that would be great as well.

And while the slides are being pulled up on the screen, the discussion here is part of the output of this sabbatical project that Yuan mentioned.

We had a multidisciplinary group conduct last year.

And really, it breaks down into three principles.

We've talked so far about this doctrinal question of IP requires humans.

And as we heard Nalini mentioned in the US, this goes all the way back to our constitutional provisions as well as the Patent Act.

We've also heard from Chris and from Mr. Shin about how those play out in Europe and Korea as well.

The second point, which is a departure into more theoretical questions, is the observation that beyond the fact that the statutes and the positive law is written to require a human to-- however inventor is defined, it's been interpreted to require human

as a matter of legal interpretation.

Going a step below that and saying, well, why is that?

And what we see scattered throughout the case law is discussion about this idea of creative genius and ingenuity.

And sometimes, it's the flash of genius.

Or other times, it's a series of steps of diligent research that are not obvious.

Systemic experimentation.

But it's not just brute forcing your way through a whole bunch of permutations.

And so, we use that as our pivot from the written doctrine to a question of creativity.

The hook for this really shows up, I think, in as much in the law of non-obviousness as it does in the doctrinal requirements of an inventor being human.

And that creates some interesting tensions that are perhaps open for future explanation about the fact that inventor as human is one place in the law and creativity is another place in the law.

How do those connect up?

And then the other point that I'll introduce and then move forward on is that this idea of creativity is measured against that of a person of ordinary skill in the art, a community standard.

And I'm going to come back to this idea of a community standard as being important to creativity.

So let's go to the third of our three questions.

What is creativity?

And this was really a wonderful part for me to learn in this sabbatical project.

Because we had philosophers and physicists and neuroscientists and computer scientists all in the room together talking about this.

And we did a really deep dive in what the existing scholarship says about creativity.

And the two main points shown on this slide are the two key things that are really well covered in the literature.

If we talk about creativity of artifacts, things sometimes that's referred to as external creativity.

And there's some well-established tests saying if something's new and valuable, and sometimes if it's surprising, then we can say this thing-- this artwork or this document or this musical composition is creative.

And we can talk about it as an object.

But we also talk about psychological aspects of creativity.

A person engages in creative acts or is creative.

And there's a body of neuroscience around that.

And to the best of our understanding from this working group, there's a very interesting dance that goes on between our convergent thinking, our task positive thinking, when we're focused and working on a problem, and then an idea comes to us in the shower.

Or we sleep on it overnight and something comes out that's divergent thinking, default mode network, what our brain is doing when we're not focusing on a task.

And then the salience network is the part of our brain that says, ah, that piece is relevant to this other thing you were working on.

And it's really a back and forth between all three of these that produces creativity.

You can't brute force your way to creativity through just convergent thinking.

It's not something that just daydreaming alone is going to come up with.

And so there's an important aspect to all three of these.

And then what our working group added to that is that underlying all of this is a social or community aspect of judging creativity.

If something's new, it depends on what else exists or what the relevant community is.

Sometimes, we talk about a child drawing a picture of their cat as being creative, even though millions of children have drawn similar pictures of cats.

But there's something creative about that exercise.

So it's looking at a narrower community.

Valuable part of the aspect of the external artifact measure of creativity is a social value and salience.

Even the psychological aspects, has some aspect of social relevance.

And these are some figures that are from other sources.

They're in our white paper.

If you look at all of the words that are used in the dark circle in the Venn diagram, they're all social judgments and reflect that this is part of how we talk about creativity.

And similarly, the chart on the other side is saying, look, if something is not new at all, if it's really familiar, it's not going to excite us.

It's not going to create that kind of response.

If you move away and it's new, it'll go up the curve until it peaks at some point.

And then it becomes too avant Garde too far out there, and then we don't like it.

And those are reflective of social judgments as well.

So how do we apply this framework of creativity to patent law?

Well, clearly, we have to have an artifact.

There's a reduction to practice of the invention or constructive reduction to practice in the patent application, and new valuable and surprising tracks very conveniently with new, useful, and non-obvious.

The internal creativity-- I showed you that either the flash of genius or the methodical, but non-obvious diligent research is part of what's required.

That psychological aspect is also required as part of the creativity for patent law.

And then the social factors come into play in terms of our measure of what is the level of skill in the relevant art at the time of the invention.

So it maps very nicely to this traditional account of these three aspects of creativity.

If we then turn to generative AI, the external aspect, which is driven a lot of our conversations, it's the same.

Stuff that comes out of generative AI clearly is capable of being new and valuable.

Maybe because generative AI is focused at coming up with high probability outputs.

It's maybe a little less prevalent that it's surprising.

But that's also a soft factor for the artifacts.

Task focus.

The convergent aspect of internal thinking may be very analogous to humans.

But this dance between convergent thinking, divergent thinking, and the salience network is not something that anybody in our working group was persuaded could be mapped to what goes on with generative AI algorithms. And then the social factors are all over the map.

Some stuff humans react as if it's human made.

It's comfortably familiar.

Other things, you hit that uncanny valley.

Things are a little too weird.

But that social judgment is also present.

And so, that I invite you to consider is a framework of thinking why human creativity, and in particular the psychological aspect is really a justification for this human inventor requirement.

And there's a QR code if anybody wants to pull the white paper.

YUAN HAO: Very interesting.

I find it fascinating that actually much of Dr. Mammen research on overlaps with the course.

Rob and I co-teach at Uc Berkeley here.

And actually, it's interesting that for the first two classes of this course, we assign a lot of readings on creativity, research, and social media [INAUDIBLE] one of our students.

Hope you guys forgive us for doing that, but I am kind of obsessed with-- I think there's a lot of good stuff, as Chris just showed us.

I'm sure just the flavor of it.

He did a lot of research on this.

So one point I do want to highlight here with the limited time is that I find the difference between a product based or in Chris's words, the artifact based definition of creativity, which means, regardless of the creative process, as long as the output is

new, surprising, and valuable.

Just to quote Bowden, that's-- we find creativity there.

But there's another school of thought saying that no, no, no, that's not the only thing we care about.

We care about the process as well.

The process is very humanistic.

I find it has to be a lot of sparkles and also the creative process as fulfilling our inner vision as a creative individual.

So anyway, in a nutshell, I find this differentiation might also resonate with some deeper justifications of intellectual property as a system.

If I could refer to Professor Merges.

I think it's a seminal book, a 2011 book justifying IP.

So my question-- sorry, for all of you.

Before we quickly move to the audience is that Chris, do you think or do you think whether you are or people's deeper beliefs about IP justifications might affect how-- take these AI challenges to inventorship?

CHRIS MAMMEN: The short answer is yes, and maybe we can take the slides down so Nalini can rejoin us here.

Great.

And I think so very much.

I think in the many conversations I've had with people around this, quite often the justifications are articulated in terms of our IP system exists to support national economic and industrial policy.

Or it exists to support what Aristotle would have called eudaimonia, the pursuit of a good life of humanistic well-being.

And a lot of times, how people break on that question influences how they work backwards to answer the rest of the analysis.

Utilitarianism-- I think a lot of times gets distilled to just an economic analysis.

But take [INAUDIBLE] more broadly, a utilitarian justification could include both economic well-being as well as some of these other inchoate types of well-being.

And so how you approach those fundamental questions really does influence how you think about the whole analysis.

YUAN HAO: Beautiful.

So we have three minutes before we move to the audience questions.

So the rest of our three panelists, in your jurisdiction dd you-- OK.

So just a very quick question oversimplifying it, but I have to do that.

Is there a mainstream IP philosophy in your jurisdiction and how do you think this would affect how the courts they take the challenge?

Chris.

CHRIS DE MAUNY: I think there is, but it does vary somewhat.

And what I mean by that is that more traditionally, I think the focus in England has been rather economic in Outlook.

But the influence of European law-- and I use that term in a very loose sense, meaning both the EPC and European Union law has changed that slightly.

And if you look at continental European jurisdictions with a very broad brush, there is somewhat more focus on the rights of the author-- talking about copyright here, but moral rights.

That kind of idea is, as a generality, more important.

So I think when you bundle that all together and look at AI and IP broadly, it's a pretty open question as to where things might go.

At the moment, the focus in England at least, is more on AI and copyright and dealing with that than AI and patents.

Because I think that's thought to be the more pressing issue.

But it will need to be addressed.

And I think it will probably be looked at economically and issues like if one accepts that AI can be some kind of inventor or co-inventor, which would require a change of the law, how would you allocate that?

Because ultimately, the AI isn't going to be the beneficiary of that.

It's going to be the owner, whatever that means.

The user, the trainer, the provider, there's a whole range of options which have different economic impacts as to how you effectively designate or deem the inventorship if you accord it to an AI system.

YUAN HAO: Wonderful.

OK.

Yong Woo.

Korea.

YONG WOO SHIN: Yeah.

I think the Korean IP approach is aligned with the global legal framework, but it tends to place greater emphasis on national welfare.

And so I think-- YUAN HAO: National welfare.

YONG WOO SHIN: Yeah.

And AI inventorship in granting seems to be two aspects.

The pros and cons.

And many, many people said it drives the technology progress.

Yeah.

Because the AI invention makes the AI developers to make the greater the good performance of AI.

And other people said that if AI were granted inventorship or the AI owner would get hold of patent rights and get some benefits of the [INAUDIBLE] patent.

So it may hinder the innovation because the employee don't get the benefits of the AI to the inventing.

So it may hinder the progress or innovation of the AI inventing.

YUAN HAO: OK.

Thank you.

And Nalini, do you want to quickly-- just one or two sentences-- talk about-- NALINI MUMMALANENI: Yeah, sure.

Like you have heard about the US perspective from Dr. Mammen, this inventorship requirement, naming the inventors and the applications-- these are some of the characteristics that are very unique to US patent law, not generally shared by other countries.

So I see it as if you look at Article 1, Section 8, like Dr. Mammen talked about.

So you're trying to promote the progress of useful arts-- sciences and useful arts at the same time, reward human ingenuity.

So I think both these aspects come into play when we are looking at it from the US perspective.

YUAN HAO: Thank you.

It's very interesting.

So we have about maybe 9 minutes or so to have some Q&A from the audience.

Sorry, I wasn't that good at controlling the time, but I do my best.

Any questions from our audience.

Oh, I think Lisa.

[INAUDIBLE] SPEAKER: Her mic is coming.

YUAN HAO: So [INAUDIBLE] microphone.

SPEAKER: Her mic is coming.

YUAN HAO: Great.

LISA WILLETTE: Hi, I'm Lisa Willette from Stanford.

So thanks for a great panel.

And this follows up on the questions about how this connects to the justification for IP.

And I mean, I find a lot of the discussion about-- I mean, both how countries have doctrinally thought about this question and some of the scholarship.

Confusing in that-- I mean, at least from a US utilitarian or at least welfarist perspective.

It seems largely not focused on the key question of how does AI's role in the inventive process affect the question of whether this is the kind of invention that we-- from a social welfare perspective, want to have patent protection On.

And whether the role of the patent in incentivizing the invention and its commercialization is worth the cost that invention has for society.

So I realize you have been talking about this, but just more crisply like, why do you think that should be the way that from a US perspective, we're thinking about it at least?

Or from other countries perspective?

And if so, why does that doctrine seem more muddled than that.

CHRIS MAMMEN: Great question.

And, I'm inclined to agree with the premise that in some sense it's like a multivariable equation.

And if we can solve for one of those variables, then it becomes easier to solve the equation for the other variable.

And I'm afraid what I'm doing by suggesting this dichotomy between the industrial policy and the humanistic aspect is to unsolved for that first variable-- making it so we're dealing with the practical questions of what should the policy be.

Even as we're also caught up in the swirl of what are the underlying policy justifications.

But I think it's important to have that conversation and talk about both parts of that at the same time as opposed to postulating one of them away in order to make it an easier question to answer.

CHRIS DE MAUNY: I think-- YUAN HAO: Yeah.

CHRIS DE MAUNY: From my perspective, in English law, we always bear in mind that patents are an exception to the basic rule that monopolies are aboard.

And that needs to be thought about here.

If technology flows in a certain direction.

I think Mr. Shin was alluding to this a moment ago, that the prospect that potentially a small number of actors could suddenly hugely dominate across many fields by churning out thousands of inventions is not something which the patent system was set up to endorse, in my view.

And I don't think that most English patent lawyers would think it was either.

How one actually deals with that is really challenging, but I think it is an issue that will need to be addressed if that is how things move.

YONG WOO SHIN: Yeah.

So the role is related to policy and the benefits

of the patent system should be distributed with Justice.

So yeah.

And I think it needs to be careful to grant AI inventorship [INAUDIBLE].

Because it may result in the big impact on the economy.

Yeah, related to the technology industry.

So as I said before, the AI inventorship may be good or bad impact.

Yeah.

There are two aspects.

So we need to expect the impact of the legal system.

And so we need to be careful.

Yeah.

That's all.

YUAN HAO: So I think the lady on the right.

She raised her hand second.

Sorry.

SPEAKER: Speaking as an IP licensing lawyer.

Has anybody ever raised the question of whether it really is an invention or not if it's created by an AI algorithm?

I mean, has that ever been thought about deeply?

It just seems like there's an assumption that, yes, it's an invention because it's new.

YUAN HAO: Just to clarify the question a little bit.

So you mean the AI contributed to the invention in such a significant way that if it were a human being, she would have been found as an inventor.

The conception, US context.

Is that what you mean or?

SPEAKER: I don't know enough to ask the question well.

Let me try again.

YUAN HAO: [INAUDIBLE] that's a good question.

Sorry.

SPEAKER: Let's suppose that an AI machine operating with an algorithm creates a new result of some sort.

Is it a foregone conclusion that constitutes an invention?

Even though it may fulfill the other criteria of what is defined as an invention.

So I'm thinking not so much about the inventor as the invention piece of that.

Is it an invention.

Has anyone ever debated that?

That's my question.

YUAN HAO: OK.

Our panelists first.

Yeah.

CHRIS MAMMEN: I'm happy to jump in, but it also seems like it may be a good question for Nalini.

NALINI MUMMALANENI: Yeah.

So what we do is I can tell from the Patent Office perspective.

So what we do is once the application comes in-- once a patent application comes in, we evaluate that application on the basis of the statutory provisions.

Whether it satisfies 35 U.S.C 101, is there prior art?

Where you talk about whether even if it is an invention or not, that would play into that.

Did someone else invent this piece before-- 102 and 103?

And then is it sufficiently described?

So if you look at the patent statute, it says the person is entitled to a patent unless the Patent Office finds some reason why they should not give the patent.

So I'll start the conversation with that, and I'll let the other panelists chime in.

YUAN HAO: Maybe just one minute.

I'm really sorry, but because of the time, just quickly chime in.

CHRIS DE MAUNY: I don't think there's a legislative basis in my law for distinguishing patents, depending on how the subject matter was derived.

If we ignore the inventorship issue, I don't see where you would get a basis for distinguishing that.

It's an objective assessment of whether the claimed invention is, let's say, non-obvious to adopt a slightly different phrase from the state of the art.

YUAN HAO: I'm really sorry.

I have to call a stop here.

I know we still have questions, but luckily you guys can grab our panelists while having coffee.

So we'll come back within maybe 15 minutes for Panel 2.

And thank you so much our wonderful panelists.

Nalini.

[APPLAUSE] Thank you.

CHRIS MAMMEN: Thank you, Yuan for hosting our panel.

YUAN HAO: Oh, thank you.