[Feature: AI and Intellectual Property Rights] Roundtable Discussion: Things to Consider for Coexistence with Generative AI

Today, I would like to talk with everyone focusing on generative AI, which is currently spreading at a very high speed, and its relationship with intellectual property rights.

Nowadays, discussions about "ChatGPT" take place almost every day, whether you turn on the TV or open the internet. Compared to previous language processing, it produces very natural text, and it is spreading globally at a tremendous speed.

While there are positive evaluations that it is convenient, there is an ongoing controversy regarding the extent to which AI should be used in educational settings. Some argue that for a child's development, it is better not to let them touch it until they are 12 years old, sparking a debate.

AI technology has suddenly become familiar to us. Mr. Sugiura, as a researcher who has been working on the development of this technology for a long time, I would like to ask you to first introduce your own research to date and the technology that current AI is capable of.

My specialties are intelligent robotics, machine intelligence, and deep learning. Until three years ago, I was also researching speech translation at a national Research Centers and Institutes. Speech translation is a field of AI that has been researched for a long time, but now it has become so common that almost everyone knows that speech translation software is included in smartphones.

Many AIs that surpass humans have also emerged. In solving quizzes and games, they have exceeded human champions. Also, if we consider being equal to or better than the average person, machine translation and speech translation can handle more than 30 languages, so I think they are superior to ordinary people.

Image recognition technology has also improved in accuracy, with errors steadily decreasing, and by around 2015, accuracy higher than human recognition was achieved. Furthermore, there is challenging research aimed at creating an AI that can win a Nobel Prize.

As for what positive aspects the emergence of such AI that surpasses humans has for people, one example is that Go players use AI to practice repeatedly and become stronger. "AlphaGo" defeated the Go champion in 2017, and research results reported this year show that players' scores have risen sharply since then. Using AI for practice to improve human skills can be seen in various fields.

I will briefly introduce large language models. The much-discussed ChatGPT is a large language model announced in November 2022 that generates text interactively. GPT-4 was released in March of this year, and it has reached a level where it can achieve a score equivalent to the top 10% on the Uniform Bar Exam in the United States.

The mechanism is what is called a "language model," and what it does is simple: it just predicts the next word. For example, after "Once upon a time in a certain place," it is likely that "there was an old man," but "there was a father" feels out of place. In other words, the probability of occurrence in actual text is different. It reads a large amount of text data, learns the frequency, and predicts the next word with high accuracy. This is what a language model does.

The fact that a program like ChatGPT is born from this language model is very surprising even to us researchers.

Also, if you use ChatGPT, you will notice that it doesn't produce the exact same text every time; by adding a few random elements, it can produce completely different text. Language models are built within such a framework.

I understand well. Next, I'd like to ask 852wa, who creates images as an AI art director and is truly putting AI technology into practical use, to talk about the things you face in your creative work.

In terms of image generation AI models, a service called Midjourney came out in July last year and became a big topic.

This service utilizes a format called "text-to-image," which is a model where you input text and it is output as an image. For example, if you enter "flower, girl," an image of a flower and a girl is output. As the original model, there is a mass of data that has learned hundreds of millions of images and memorized their characteristics in the form of vectors. It is easy to imagine if you think of it as picking up characteristics similar to the information entered in text from that mass and generating and outputting an image.

When Midjourney came out, it became a huge topic because it generated creative images with a quality that was incomparable to previous image generation AI systems. Images of a quality that made people think, "Wait, wasn't this made by a human?" began to be output one after another.

Then, an image spread among the general public that perhaps it was just patching together and collaging commercial images created by creators until now, and it became an issue.

Basically, the images themselves are not contained within the mass of data of the AI model. The model learns conceptual features, such as that humans feel comfortable if there is a line here, or "conventions" like if there is a line here, this color comes next, as Mr. Sugiura explained with text generation. Basically, it's not that different from the mechanism of a language model.

However, since a picture is a mass of pixels, you can't tell when they are lined up, and as a result, images are created by high-precision AI to the extent that one might think existing works were cut and pasted. Even though we thought art strongly involved human originality, skill, and creative activity until now, a machine does it instantly and effortlessly.

Therefore, suspicions have arisen that perhaps images created by humans are being copied or patched together, leading to problems. Originally, there were also emotional issues, such as why my drawings are being learned without permission, and creators have complex feelings.

How do those who create using AI go about it?

There are two types of people: those who specify with text called a "prompt" and select one image from hundreds or thousands produced randomly, and those who use a method of making a stick figure (a character with limbs expressed like sticks) take a pose and then matching the AI image to it in order to output the imagined image in their head as intended.

I think these two are different as creative activities, but the output images are indistinguishable. I think it's a very difficult point to determine where the creative activity part lies. I generate images using both methods.

I looked at 852wa's art book, and they are very lovely illustrations. Some of them are presented with an introduction saying that such-and-such a prompt was used. I was impressed, thinking, "So if you put in these words, this kind of picture comes out."

In what way do creators add their own expression within their creative activities?

When instructing the AI on the image in your head, there is also a method called "image-to-image" where you provide an image as instruction data. In other words, you can give instructions by providing a similar image. That can be a finished picture, or it can be an instruction image of a stick figure saying you want it to take this kind of pose.

This is completely different from the instruction in text-to-image where you vaguely say, "Create something with the sun, sunflowers, and a girl in a straw hat."

However, there is a problem with this image-to-image: since any image data can be used as an instruction, for example, someone else's copyrighted work can be used as an instruction image. And by manipulating numerical values, it's possible to produce a picture far removed from it, or an image that is almost no different from the original.

Are the cases on the internet that cause a huge uproar, such as an AI image appearing that has changed one's own illustration, created using image-to-image?

It is likely mostly that problem.

I understand that there is original training data and algorithms, and there are methods to create images by providing instructions with image concepts or by providing instructions in language. Within that, issues regarding intellectual property rights have actually occurred and are being viewed as problematic, and lawsuits have actually been filed in the United States.

In Japan, what is the Copyright Act like regarding the use of data input into AI? Mr. Okumura, could you explain?

In Japan, there were originally provisions for limitations on rights for text mining and data mining. These provisions allow for the unauthorized use of others' copyrighted works by computer within the necessary scope for information analysis, such as when creating dictionaries or developing facial recognition technology for photos.

There was much discussion about whether these provisions could be applied as they are to deep learning in AI, but in 2018 (the 2018 amendment to the Copyright Act), provisions were established to allow the free use of others' copyrighted works within the necessary scope for information analysis, including machine learning in general.

Therefore, in Japan currently, as long as it is for the purpose of machine learning for AI, there is no problem under the Copyright Act regarding inputting—that is, reproducing—things from the internet or books. From that point of view, Japan is sometimes called a "machine learning paradise," and it is said to be the country that most clearly and broadly permits the free use of copyrighted works regarding machine learning in the world.

However, the above refers to the learning process. Cases where something very similar to an existing work is output from an AI through the generation process, as 852wa just mentioned, are a separate discussion.

One of the reasons why Sam Altman, CEO of OpenAI, visited Japan and declared his desire to open an office here is said to be because Japan's Copyright Act is easy to work with since it clearly provides limitation provisions for machine learning. In other words, I think there is a corporate benefit in being able to reduce litigation risk. If the 2018 amendment was legislated in anticipation of such a situation, it is a wonderful thing.

However, having said that, it doesn't mean that anything goes as long as it's put in as data; it is permitted within the scope of the limitations of copyright. Copyright includes various rights such as the right of reproduction and the right of public transmission, but these limitation provisions list cases that, in principle, do not constitute infringement regarding the exercise of any copyright. At the same time, they also stipulate exceptions where it must not unreasonably prejudice the interests of the copyright holder.

In other words, if it is legally determined that the interests of the copyright holder have been unreasonably prejudiced as a result of inputting data, the possibility remains that the conclusion will be overturned. How to think about that becomes important as an interpretation of the Copyright Act.

To give one example from the United States, there is a famous problem that occurred in October last year with the free version of GitHub Copilot.

A professor at Texas A&M University discovered that program code for which he holds the copyright was being generated by AI. I have seen both the code written by the human and the AI code, and they are not completely identical. However, the comments—the part where you express your thoughts rather than the main body of the program—are the same, and if you look at this normally, they are quite similar. So I think the professor's opinion is certainly valid.

Programs are also often used in a way where you output what is in your head as you go, so the whole thing isn't generated all at once; you give commands one by one with prompts. Therefore, in the case of a large-scale project, such as creating a certain game, the current situation is that code cannot be created unless the human's intention is clear.

I would definitely like to ask from where something constitutes a copyrighted work. What happens if something is generated that is not completely identical but is considered similar by a programmer?

Also, it's close to the part called the specification, but if human thoughts are used as input in a prompt, to what extent does this become a copyrighted work?

You mean the copyrightability of language written as so-called prompts or commands. Mr. Okumura, what do you think?

How to perceive the prompt itself under the Copyright Act is still my personal opinion as there have been no court cases yet, but since it is a command to an AI/computer, I think it is possible to position it as a work of computer programming.

I think a very simple prompt would not be a copyrighted work because it lacks creativity, but a complex and considerably long prompt could be evaluated as a work of computer programming.

However, even if there were a one-to-one correspondence between the prompt and the finished image, if there are various variations, it means the prompt does not directly represent the output image or text. In that case, just because you hold the copyright to the prompt doesn't mean you hold the copyright to the output.

I see. The tricky part is that it's easy to include randomness. If you do that, the appearance of the output changes quite a bit. If it ultimately comes down to judging on a case-by-case basis, will it take the form of a human looking at the output and determining that it is a problem under the Copyright Act?

That's right. There is a case in the United States where an attempt was made to register a comic created using Midjourney with the Copyright Office. This is called the "Zarya of the Dawn" case. The author applied without any particular explanation, and at first, registration was granted as is. However, after registration, the author tweeted on SNS something like, "A comic made using Midjourney has been registered," and the Copyright Office re-examined the circumstances of its creation and the previous registration was canceled.

Following that, they decided to grant copyright registration for the dialogue parts of the comic because the author created them themselves. However, they decided not to grant registration for the individual images output from Midjourney.

The reason is that in this case, the Copyright Office said the method involved entering fairly vague prompts and selecting ones close to the image from among hundreds of outputs. In this case, the degree of human contribution in creating the work was low, and at this level, it cannot be said that a human created a copyrighted work (using AI as a tool); they judged that the AI was performing the act of creation autonomously. And since U.S. copyright law does not grant copyright to things created by non-humans, the images were not registered for copyright in this case either.

Furthermore, the author also claimed to have directly added their own drawings to the images output by the AI, but since it was at the level of just painting a little color, the Copyright Office judged that to be insufficient.

Even if the prompt has a certain influence, if a large part of the output is left to randomness, it cannot be said that a human created it; the AI created it autonomously. That kind of thinking was shown. It's a U.S. case, but I think it will be very helpful in Japan as well.

The Copyright Act basically targets creative activities by humans. Under Japan's Copyright Act, the rights of an author arise for a person who has created a creative work. Systems differ by country, but the basic idea is the same. When AI is used as a tool for creative activity, whether it becomes subject to the author's rights as a copyrighted work is judged by where and to what extent a human made a creative contribution.

Even in cases where it was disputed whether there was copyright infringement of a conventional computer program work, if the program is nothing more than a combination of simple commands that would be the same no matter who wrote them as commands to a computer, or if it cannot be said that one expression was selected from multiple options, it will not be protected as a copyrighted work even if it functions as a program.

Also, the basic conventional thinking was that it becomes an infringement if a part of the expression to which a person added creativity is imitated, but it does not become a copyright infringement if only parts that are not so are similar. If this becomes a prompt, will that be applied as is? Since prompts have a high level of abstraction, I feel there will be many cases disputed starting from whether they can be called creative expressions in the first place.

Beyond that, if a generated image is created based on various commands and something that looks like a copyrighted work is produced, will that be protected under the Copyright Act? As Mr. Okumura said, will it be judged individually or comprehensively after looking at how a person was creatively involved? Basically, whether the generated product becomes subject to the author's rights is judged based on whether the person's creative contribution is reflected in the expression.

In that case, for example, we write programs, but does it mean it's also important to save the prompts so as not to be sued later?

That might be the case.

As in the "Zarya of the Dawn" case mentioned earlier, the fundamentals of copyright are different between Japan and the United States. The U.S. idea of copyright is that a work for which copyright is recognized gains copyright. On the other hand, in Japan, I think the person has the copyright at the moment they create it.

Currently, I think it's quite difficult to recognize copyright for an AI output that is just "pushed out" as is. For that AI image without copyright, for example, how much change must be given to the screen for copyright to arise? It was said earlier that just painting a little color is not recognized as a creative activity, but I think this is what all creators are most concerned about right now.

However, that's hard to judge by percentage, and in the end, I think humans have no choice but to judge it like, "this is OK, this is no good," but I think it's also very difficult for people in a legal position to make that judgment.

Even regarding expressions that do not involve AI, the judgment of whether it can be called a copyrighted work protected under the Copyright Act has been disputed in cases where it is subtle whether a certain creator's individuality is expressed in a commonplace expression. The same can be said when AI enters as a tool.

Under the Copyright Act, a work is defined as "a creative expression of thoughts or emotions." Since the conventional thinking is that such an act of expression is performed by a person, we, the people of society, perform that interpretation and make a judgment as to whether it can be called a work under the law. If it ultimately becomes a dispute, a judge makes that judgment and renders a verdict. That verdict has binding force between the parties to the dispute. This is the current situation.

When we say "copyrightability of AI-generated works," the discussion becomes about "whether it exists or not," but that is not accurate. For something completely newly created by AI, "whether it exists or not" is fine, but in cases where it incorporates part of an existing work, the existing part belongs to someone else, so copyrightability "exists."

Just because an AI created it doesn't mean copyrightability disappears even if existing parts are included. In the case of a combination of a newly created part and an existing part, it is necessary to think about them separately. This point is not a new problem caused by the advent of AI; it is the same as when a human combines existing things.

Even with the "pushed out" output 852wa mentioned, if the AI created something that has never existed before autonomously, there is no copyright at all. When 852wa adds something to this, if there is a part where individuality is expressed, 852wa's copyright exists only in that part.

If the output pushed out from the AI is something like a collage of someone else's existing work, the copyright for the existing part naturally belongs to that author. If you add something to it, you have added something to an infringing work.

Also, "what percentage" is quite difficult, and as has been said conventionally, it will be explained by whether the author's individuality is expressed.

I think homage, collage, and respect have also existed in the group of works between humans until now. Simply using an image to generate an image. So I can understand that when an AI-generated work is produced using someone else's copyrighted work, the existing copyright remains.

However, basically, the very initial stage of learning images as a model is not a problem legally, but are you saying that for images produced using that, the copyright has returned to the original?

The model itself doesn't contain the data of the image itself, but contains conceptual data or data as vectors, as numbers. When it doesn't contain the pixels or the lines of the drawing itself, I am very curious about where that copyright resides.

For example, even if you convert the expression of a certain figure into vector data, reproduction is reproduction. Whether the data is held as pixels or vectors makes no difference if the same thing is reproduced. It's the same as saying it's the same whether you hold data in analog or digital.

The question is to what extent the artificial intelligence abstracted and understood the expression.

Does it come down to how much reproducibility there is?

Reproducibility is a consequential issue. The point is the story of the process and mechanism of how the output was completed. It's whether it can be explained that the AI had something abstracted down to the style or art style of the drawing and drew based on that, so it became a similar picture.

This is a matter of evaluation from both technical and legal aspects. Even if technically it doesn't have the original drawing itself, as a legal evaluation, it is possible that the original drawing is deemed to remain in some form as a result.

For example, if you give a verbal instruction for "Pikachu" and a drawing exactly like Pikachu comes out, I think it would be difficult to prove that the AI does not have "Pikachu" as an image and merely understood the concept of Pikachu abstractly and the same thing happened to come out. That specific drawing doesn't come out from just the word Pikachu. It tends to be evaluated that the artificial intelligence remembers that drawing itself.

However, if you have it learn many "Pokemon" and instruct it to create a "Carnation Pokemon" that doesn't actually exist, and a drawing comes out. If that happens to look like a "Carnation Pokemon" created officially later, I think it would be a case where it just acquired the style common to Pokemon and happened to look similar when it drew based on that.

Even among humans, there are many people who draw pictures in the style of so-and-so, but whether that is OK or out must be considered on a case-by-case basis.

What about a concept like, for example, having memorized something like the Doraemon drawing song?

If the exact same image of Doraemon appears, I think courts tend to judge that the AI did not learn the concept, but rather memorized the image. However, if it can produce a Doraemon image that shares the same concept but is completely different from the training data, it becomes easier to explain that it has learned the concept.

Under copyright law, the basic rule is that you are allowed to imitate a style, but you must not imitate specific expressions without permission.

Therefore, no one can monopolize a style, such as an Impressionist-style way of painting or pointillism. This is because allowing such a monopoly would prevent humans from expressing themselves freely. In contrast, if someone paints a picture similar to the specific expression of Monet's "Water Lilies," they are imitating a specific expression, which constitutes copyright infringement during the term of copyright protection (unless it falls under copyright limitations such as private reproduction).

So, even with AI in the middle, the final determination of whether copyright infringement has occurred depends on whether the outputted image or program is judged to have imitated a specific expression based on current thinking.

A famous deep learning textbook actually contains an example of learning Pokémon images to generate Pokémon that do not exist in reality.

In terms of whether it is advantageous or disadvantageous for the defendant, I think it is disadvantageous if it only produces Pokémon similar to existing ones, and advantageous if it can produce new things in a Pokémon style.

In the future, legal evaluations may differ between AI that purely learns style and "pseudo" style-learning AI that appears to be learning style but is actually just storing the images themselves.

Midjourney in the U.S. is currently being sued in a class-action lawsuit, where the plaintiffs argue that the AI is not learning styles but is holding compressed versions of the training images. I believe that once a conclusion is reached in this lawsuit, a direction for what is "safe" or "out" may emerge.

In October 2022, Shutterstock, a photo and illustration provider, announced it would offer AI image generation tools through a partnership with OpenAI, stating it would create a system to compensate the authors of the original image data. I believe this is precisely an effort to avoid litigation risks.

Shutterstock's regulations state, "Contributors will be compensated for the role their intellectual property played in the development of the original model." Will the legal extent of that "role" and how it should be correctly divided be decided in the future?

I think the expression you just mentioned would be judged to apply if, upon looking at the generated image and comparing it to the image claimed to be the original, it is determined that the generated image utilizes the creative expression parts of the original image.

It means they created the terms of use in advance to prevent lawsuits from occurring. Presumably, the final layers of machine learning must contain fairly specific data, so the rights of the original authors are likely included there. Since there is a possibility of losing if sued for copyright infringement through the use of that data, I think they created a system to pay a certain amount of money beforehand.

For example, if this becomes established as legally correct, it will also serve as an incentive to develop AI that can calculate such models and roles.

In conventional judicial precedents, whether a work is being used or infringed upon has been judged by humans somewhat intuitively, determining that a creative expression of a work is being used. However, as image processing and language analysis technologies continue to advance, it has become possible, if one wishes, to analyze what percentage of an image's expression is being used.

However, the extent to which this should be done is a very large issue. Even in human creative activities, it is rare to create an expression from absolute zero that bears no resemblance to the works of predecessors. Since childhood, we have read various novels, watched movies and images, and listened to music; that accumulated data becomes, in a sense, training data, and we engage in creative activities using our brains and hands.

Given this, when comparing human creative activities to AI, must we judge AI with extreme granularity? Despite having a convenient tool, if using AI becomes more troublesome and requires payment, it will ultimately hinder the development of human culture and industry.

Therefore, to what extent should we protect the rights of original creators and ensure their financial interests, and from what point should we encourage free use and the creation of new activities through collaboration between humans and AI? This will be a very difficult judgment.

One thing to be careful about in discussions regarding "creation" using AI is that the debate over what cases constitute "creation" using AI could potentially reflect back on human creative activities. The criteria for what can be called a creation apply both when using AI and when humans create using traditional methods.

Currently, we see some very harsh critical opinions regarding creation using AI, but in fact, that blade is also pointed at human creation.

Until now, copyright law has discussed things in a fairly loose way on the premise that humans are the ones creating. However, with the advent of AI, we have to think more deeply about what creation is, what constitutes style, and what constitutes expression. Humans cannot actually distinguish between style and expression that well, so until now, we have understood them by pulling them toward the side of expression. But AI might be able to draw a sharp line between style and expression.

Also, even if we can create an AI that judges whether things are similar, to what extent should we use it? How far should we take it? I think various issues that need discussion are beginning to arise. I'm not sure if being too rigid would make everyone happy.

Copyright is a relative exclusive right. In other words, the concept of copyright law is that while it is a copyright infringement if you create something similar by relying on another person's work, it is fine if the expressions happen to be similar by chance.

While everyone engages in creative activities freely, there are countless similar paintings and similar novels. Because those just happened to be similar, we have judged them loosely in a sense, but the question is what we should do when AI enters the picture.

The topic of copyright alone is inexhaustible, but I mainly research patent law, and as it becomes possible to make inventions using AI, there is a similar discussion about whether they should be protected by patent law.

Currently, I believe the field where AI is most used to make technical inventions is programming. Besides that, the use of AI is likely advancing in the field of drug discovery.

Unlike copyright law, patent law protects ideas. It protects the specific means of solution one has come up with to solve a technical problem, which is the idea itself.

Mr. Sugiura mentioned that he uses AI for programming; to what extent can AI possess creativity? Specifically, can AI be used in terms of solving technical problems?

In January, I had students try programming using ChatGPT in my class. For beginners in programming, it is important to be taught things they don't understand one by one according to their individual level, and in that sense, I think ChatGPT is a revolutionary educational method. For deep learning programs, it is actually possible to generate simple parts, and you can ask about parts you don't understand step-by-step.

However, there are also difficult aspects. In neural networks, the number of parameters determines how complex the tasks it can perform are, but even if you ask how many parameters there are, it gives a fairly irresponsible answer, which is of a quality that experts cannot use.

I believe it is very useful under the condition that we verify the elementary knowledge, but on the other hand, the current situation is that it is difficult to generate advanced knowledge unless the idea part is included in the prompt.

When a patent attorney actually writes a patent application document, they first explain what kind of conventional technologies existed. Then, they explain the invention by saying that this invention solved the problems that could not be solved by those conventional technologies. For conventional technologies, if you perform an accurate search, you can pick up data from papers and old patent applications, and I think computers are good at the task of accurately summarizing those search results.

From there, is it still impossible for them to newly come up with a means to solve problems that could not be solved before? I imagine that if we keep training them on data of conventional technologies, an era will come when data analysis can solve problems that an average engineer would not immediately think of.

The currently mainstream method called neural networks has parts that can be explained by the concepts of interpolation and extrapolation. Actually, neural networks are very poor at extrapolation. In other words, it is quite difficult with current technology to create knowledge on the outside that did not exist before.

Therefore, human help is necessary to create something essentially new. However, if we take advantage of the fact that they are good at interpolation, if there was a middle ground between field A and field B that had not been captured as a patent until now, there is a possibility it could be discovered. I think there is a possibility that AI could come up with such ideas.

Regarding the talk of interpolation and extrapolation, if the space is very wide, knowledge only exists at the ends, and since the space in between is very empty, any number of variations can emerge. So, I don't think that just because it is only good at interpolation, new ideas will necessarily not emerge.

Another thing, which is an even older technology, is genetic algorithms, where humans provide variations in advance, saying they want a combination of this and that, and then the method determines how much to blend them. With something like that, discoveries can probably be made endlessly, and they have been used a lot for a long time to control robots.

Applying that technology to the field of chemistry will enable things like new drug discovery. I think many people are doing that now. The human who set that up is the one who is great, and I believe it can be regarded as an invention.

That's true. For example, it has always been the case that when one important study is conducted, many peripheral studies occur the following year. In that way, if a human creates just one point, it is possible that AI could fill in everything near it.

Then, if an era comes where AI handles the interpolation part thoroughly, will an environment emerge where humans can concentrate on the creative parts of the work, making it easier for more advanced inventions to be born?

Yes. For example, I mentioned the level of Go players at the beginning, and I believe a form where human inventive power increases using tools is also conceivable.

That is exactly what I expect as an educator. When we think about the task of writing papers in the humanities, we read through conventional papers written on the subject, summarize and analyze them, and then add our own considerations to produce new insights; AI will be able to help with such tasks.

If we can leave it to AI, we will be able to save a lot of time in areas where we are currently struggling to read through vast amounts of papers and precedents of our predecessors. As a result, I think humans will be able to concentrate their time and effort on the part of what to think from there.

From an educational standpoint, I always try to teach students to "master AI anyway." It is always better to be able to use tools, and at this point, it is inefficient to do everything by hand without even using a calculator. Similarly, use whatever is available. I think there is no problem at all as long as it is used to improve oneself; rather, it should be encouraged.

While it is nonsense to turn in a ChatGPT text as a report as-is, I think it should be actively used to improve oneself while interacting with AI, like the Go story mentioned earlier.

I agree; AI is a tool, just like a pencil or a dictionary. If it is a tool, one way of thinking is to educate on its proper use.

On the other hand, if AI can achieve passing scores on several U.S. certification exams, it may become difficult to judge human performance based only on reports of similar difficulty. Will it be socially acceptable in the future to only give grades based on reports for problems that can be solved by AI? I think this is something we also need to consider.

Since Google was born, searching has become a major prerequisite for writing papers. However, there are people who are good at searching and those who are bad at it; when I provide guidance on papers, there are people whose ideas are unique but whose collected materials are off the mark, perhaps because they are bad at searching. However, from now on, AI may support us and eliminate the problem of being bad at searching.

But if that happens, levels that can be immediately reproduced by AI will no longer be valued. Even in inventions, parts that seem possible with AI will be said to lack an inventive step. From now on, the difference will depend on whether or not one can do the "plus-alpha" part after using AI.

You mentioned that the level of inventive step will rise, but when we consider why intellectual property law protects works and inventions in the first place, one idea is that regarding the results of creation, a certain reward should be given to the person who produced them, and that person has the right to be treated preferentially for a certain period.

Another major factor is that newly created things lead to social implementation, which enriches society and makes life more convenient, or allows for wonderful cultural experiences. There is also the aspect of protecting them for such reasons.

Particularly in the field of patent law, which aims for the development of industry, it is not the case that society becomes rich as soon as an invention is completed. To make it usable in society, there is a process of further years of research and development, commercialization, and dissemination, and only then is it realized in society.

Then, if we only have the idea that "we protect it because a human created it," it would follow that if it can be done automatically by AI, we don't need to protect it. But it is also important to create incentives to invest time and money to implement created things in society. Therefore, the conventional thinking of not protecting something because it is a creation of AI alone will need to be reconsidered.

Even if we have AI analyze and parse conventional papers and data, humans still need to handle the creative parts. In this context, how should we educate the next generation of young people within Keio University, a research and educational institution? Or what kind of research environment should we prepare? I would love to hear your opinions.

Mr. Yako, you have seen the efforts of the AI Consortium from the beginning; could you briefly introduce it?

It is called the "AI and Advanced Programming Consortium (AIC)." About 10 years ago, AI became a topic in various places, and students who sensed this sensitively said they wanted to learn about AI.

But, for example, there are no machine learning classes in the Faculty of Law. So, we gathered students interested in AI across the boundaries of humanities and sciences and created a "place of learning within a place of learning" where they could learn from each other.

There, since the main focus is on skilled students teaching other students, we, the faculty members of the support staff, focus on creating the system rather than teaching. But as students interact with each other, perspectives and ideas unique to students that faculty members would not think of are emerging, so I feel glad we did it. We are now in our fifth year and continuing our activities.

That is a wonderful initiative. On the other hand, for places of voluntary student activity that do not grant credits like the AIC, I think highly conscious students will come, but students who are not so highly conscious and take classes just because they earn credits will probably not come there.

I think there are many students who are not that highly conscious but actually have very high potential, so how to educate them is something I always worry about.

It's difficult. The AIC has not given credits so far, and I think it will be difficult to give credits in the future. To give credits, it must be authorized as a university and recognized to some extent by the Ministry of Education, Culture, Sports, Science and Technology. This is because it is currently a framework where students teach students.

For now, I think it's fine to focus activities on "edgy" students who want to do it even if it doesn't result in credits.

Mr. Sugiura, you are actually running an AI research lab; what kind of innovations are you making in terms of education?

When I was a student, AI research was not something you could get a job with; it was a field that some students engaged in simply because it was fun. But now it is used in various applications, so many students are interested.

A major change in education at the undergraduate level is that starting this academic year, a neural network course was launched in the "Basic Experiments in Science and Technology" subject, which most students in the Faculty of Science and Technology take.

There, if we convey that there are interesting technologies in a practical format, I think the number of students showing interest will increase. While expanding to more human resources, I also want to work deeply on specialized content.

Will education in the field of law change in any way?

Currently, all U.S. precedent search services are touting AI implementation as a selling point. Also, in Japan, AI is becoming a key point in legal tech. Mastering AI as a tool is something we must also do in the field of law.

Also, we are in an era where new content that fuses content and tech, exactly like that of 852wa-san, is being created one after another. Therefore, to study and research intellectual property and copyright, even if they don't understand the technology itself, I think the number of young people who take an interest and find it fun to work on must increase.

For my part, while I cannot teach the technology itself, there are new movements in the world and new troubles occurring. I hope to attract students' interest by showing them in real-time how law responds to these things. Since young people will become able to use new technologies more and more, I want to enable them to work with ideas we cannot even imagine, even in law.

There are very many students who want to do intellectual property law because they are interested in content. Such students, even if they are in the Department of Law, have various special talents, such as liking drawing or music, or doing programming. While the basic approach to traditional legal studies is to thoroughly and systematically master the six codes, such as the Constitution, Civil Code, and Criminal Code, there are students who are not necessarily happy with the conventional education system, and in fact, they may be the majority.

It's not that such students lack talent; not at all. Having an interest in new phenomena and a great love for content and games becomes the motivation to study law. Also, I feel there are various educational possibilities, such as students who entered the Department of Law but want to go to the Faculty of Science and Technology to be taught from scratch, or those who want to do two specialties with a double degree.

When new things come out, we should leave what can be left to AI. Beyond that, I sometimes think it would be good if we could raise students who can come up with ideas on how to solve legal and social problems based on basic knowledge.

That is exactly right. Currently, AI has come to be used by a diverse range of users. However, since the side that creates AI is still biased toward information science, I think the part where we discuss with various experts, such as those in law, is important. As a teacher, I would be happy if various students took an interest.

The other day, I held this year's guidance for the AIC, and there was a very energetic second-year student from the Faculty of Law with whom I spoke for a long time after the guidance. They were already using not only ChatGPT but also generative AI for images.

The fact that there are students for whom conventional teaching methods do not fit is, in trendy terms, a kind of diversity. Considering that Keio University should properly nurture diverse students, I think we need to move toward being able to provide individual education for each person, rather than just teaching with a rigid curriculum.

To that end, I believe we should shift to a teaching method where we communicate more closely with students, grasp their hopes for what they want to do in the future, and skillfully provide the educational power that faculty members possess.

852wa-san, if you have any requests for university education, please let us know.

In generative AI communities and elsewhere, there are junior high and high school students who are technically writing programs to run AI models or actually managing communities.

I believe that strengthening the field of AI at universities from now on will be very beneficial for the future of these children. Precisely because it is a technology that can be accessed for free, it is spreading widely among the digital native generation. Even in apps like TikTok, AI conversion is very popular in the image and video fields, and it is becoming a part of their daily lives more than adults might think.

The younger generation feels it is very close to them, routinely saying things like, "This technology or this app is interesting, right? I know, right?" I also receive real voices from students who used to draw in junior high and high school but have become interested in the technical side, asking how they can go about learning it.

I believe that positively preparing the educational environment to accept this will lead to a better future.

It is important for us adults to create places where the younger generation can play an increasingly active role and freely try various things. We should watch over young people warmly even if they fail. The whole of society should teach them things like, "Maybe it would be better if you did it this way," or "Doing this part like this might hurt someone." Or we should act together with them. If we can do that, then no matter how smart AI becomes, we will have a society where we can use it to lead even more enjoyable and rich lives.

I have one thought regarding intellectual property rights. I am not well-versed in the law, but I believe the underlying philosophy is a tribute to the person who thought of it first.

I believe that respecting and utilizing the wisdom of our predecessors is rooted in respect for people. Therefore, if AI-generated outputs also clearly display which parts use a certain person's ideas, AI might come to be respected, and respect for the original creators behind it can also be expressed. I hope to see generative AI emerge that makes it visible whose ideas are behind the output.

In waka poetry, there is a concept called "honkadori" (allusive variation), where a work can stand on its own even if it borrows from the past, provided there is respect for the people of old. I believe we must properly teach students to respect the wisdom of their predecessors.

That's true. Traditional social ethics, such as valuing others or respecting the work of others, may, in a sense, become more important as technology advances.

Thank you very much for a very important discussion today.