I was recently interviewed by a reporter of "Rest of World" via email and asked two questions.
1)Some Japanese politicians have suggested some specific polices about AI. One is requiring AI developers (like NovelAI and Stable Diffusion) to get permission from artists to use their copyrighted work to train AI. What is your opinion of this?
2) Another suggestion is creating licenses that allows developers to pay artists to use their work to train AI models ? What is your opinion of this?
The following text is the full text of my response.
==== ==== ==== ==== ====
First of all, I am a writer, novelist, blogger. So I am a stakeholder in the fact that my written work is being learned and utilized by AI such as ChatGPT. As one of the stakeholders, I oppose both permissions and licenses. There are 3 reasons.
(A)Because we can benefit from AI in ways that cannot be measured.
In chess and go, humans can no longer beat AI today. However, the best performer now is not a human or AI, but a "team of human and AI."
I expect the same thing to happen in the world of artists.
For example, there are many illustrators who are good at drawing but cannot make manga because they are bad at thinking of stories. Those people can now challenge the manga world by borrowing the power of chat AI.
The manga artist who wanted to draw a horror like the movie "Alien" but gave up because they were not good at designing creatures will be different from now on. They can have the design of creatures handled by Midjourney, and they can concentrate on human drama.
The manga artist who only drew manga set in modern Japan because they could not draw backgrounds, even though they wanted to draw fantasy worlds, will also be different from now on. With the power of AI, they will be able to express "what they really wanted to express."
To be honest, this email is also written using the help of ChatGPT and DeepL. Some experts have pointed out that the advent of generative AI is an event comparable to the industrial revolution or the internet revolution, but I don't think it's an exaggeration. I consider AI to be a technological innovation that rivals printing, gunpowder, and steam engines.
That is why I am opposed to setting obstacles for the research and development of AI.
(B)Because it lacks feasibility.
As above, the development of AI brings about immense benefits and will change industries, education, and military. Therefore, every country wants to lead in AI research. Thus, if regulations are imposed on AI, international efforts equivalent to the Nuclear Non-Proliferation Treaty would be required. But under the current circumstances, such international treaties are unrealistic, in my opinion.
If Japan were to regulate AI alone, Japan would fall behind significantly. Japan should not repeat the mistakes of the past thirty years.
(C)Because creation by AI is not plagiarism and is inherently similar to human creation.
There is a criticism that generative AI is just a collage of learned data, but this is not true. Stable Diffusion, for example, is said to have trained 2.3 billion images, which is only a few gigabytes; 2.3 billion images cannot fit into such a small space.
James Webb Young said "An idea is nothing more or less than a new combination of old elements." As an artist, I belive he is right.
With the exception of "The Tempest," all of Shakespeare's works are rewrites of existing stories into new plays. If you change the setting of the movie "Jaws" to space, it becomes "Alien," if you change it to a desert, it becomes "Tremors," and if you change it to a farm in the outskirts of Hollywood, it becomes "NOPE".
If even a little similarity to existing works results in being seen as plagiarism, then as artists, we can create nothing.
Of course, there is a fundamental difference between creation by AI and humans.
AI creates based only on data learned in the past, while humans create based on not only learned data but also emotions and experiences. Currently, AI cannot have emotions or experiences and requires human assistance. That's why I am convinced that AI and humans can be the best of partners.
==== ==== ==== ==== ====
Here is the article where my interview was featured.
Please let me add my opinion.
Regarding generative AI, I take a utilitarian view.
Certainly, some of my previous work may have been learned by AI without permission. Actually, I also have several works translated into foreign languages. Not only language AIs like ChatGPT but also translation AIs like DeepL may have learned without permission. But, I feel that the benefits obtained from those AIs outweigh the losses they have caused so far. (By "benefits," I mean not only personal benefits but also the benefits gained by society as a whole).
Therefore, I do not feel anger towards unauthorized learning.
However, it is understandable that those who suffer losses due to generative AI may feel angry, and I believe it is their right to express their anger.
As with the decline of the "newsreel" genre due to the emergence of television, the introduction of new technologies can change industry structures. For example, jobs such as producing large quantities of illustrations with as little individuality as possible, such as some illustrations in Game Apps, could be in crisis due to image-generating AI.
On the other hand, illustrators and painters can benefit the most from image-generating AI.
In Japan today, professional illustrators are trending towards training their own AI to learn their unique art style and create a "personal machine." For illustrators with high information literacy, image-generating AI is not an enemy that takes away their work but a powerful tool that helps streamline their work.
In contrast, those without artistic skills cannot even draw illustrations for additional learning. This is a significant disadvantage when using image-generating AI.
Only people with drawing skills can fully bring out the performance of image-generating AI.
Therefore, I am skeptical of the prediction that human artists will be ousted by image-generating AI.
I write manga scripts, but I don't think human comic artists will become unnecessary. Would a composer think that a human orchestra is unnecessary just because there is MIDI? Or does the argument that "we don't need human singers if we have Hatsune Miku" have any persuasive power?
In any case, the answer will be revealed in a few years. I think it's more meaningful to engage in creative activities than to argue on the internet. I predicted that humans and AI would become the best partners. I want to devote myself to realizing that prediction.
What future do you want to realize?
