(Pluto) Naoki Urasawa: I don’t want to leave the fun to AI

Since the birth of ChatGPT, the field of generative AI has grown rapidly. Even if you don’t know how to draw, you can now automatically generate illustrations, comics and even animations with just one click.

After everyone can easily generate illustrations, should creators worry about losing their jobs? For top creators, they don’t seem to worry about this, but focus on higher-level questions: What is your original intention in drawing?

In 2023, Netflix launched the comic animation (PLUTO) and invited the original author Naoki Urasawa, who was hailed as the "manga master" by the media, and the top illustrator Mai Yoneyama for a conversation.

While the two were painting and talking, Mai Yoneyama suddenly asked Urasawa if he was using AI. Urasawa said:

“Although I have been thinking about AI, the work done by AI should be exactly what we want to do, so if it is left to AI to do it, it means handing over the fun we enjoy (painting). "

He also speculated that those who do not find pleasure in painting may use AI to work.

For Urasawa, everything about painting is interesting, from sketching to coloring and shading. I definitely want to do it myself. Leave all these tasks to AI and painting will become no fun.

However, he does not completely deny AI. In certain circumstances, AI can be a good tool. Suppose you want to draw a magnificent sunset background, which would be very time-consuming to draw by hand. In this case, you can use AI to generate an initial concept map that matches your imagination.

Mai Yoneyama also believes that AI can only be used as a tool to express one's own ideas. If there are no ideas in the mind, it will be difficult to start creating. If AI is viewed as an auxiliary tool, a world of coexistence can be achieved.

• Profile of Naoki Urasawa: Aged 64, he has won heavyweight awards such as the Tezuka Osamu Cultural Award, the Japan Cartoonists Association Award, and the Agency for Cultural Affairs Media Arts Festival Excellence Award. His representative works are: (MONSTER), (20th Century Boys) etc., the cumulative sales of all comic works have exceeded 140 million copies, and he is hailed as a comic master by the media.

• Profile of Mai Yoneyama: 36 years old, Taiwanese-Japanese illustrator and animator, with more than 1.3 million followers on X (original Twitter). He once worked for the Japanese animation company GAINAX and is currently a freelancer. He has participated in TRIGGER's productions such as "Little Witch Academia" and "DARLING in the FRANXX", and is also the animation storyboard for the ending theme of "Avatar: Edgewalker". with the show.

Why do AI pictures have no soul?

As one of the most representative contemporary Japanese illustrators, Mai Yoneyama's works have long been used for "training" by many AI generation services. Now, as long as you know how to chant, you can easily generate illustrations imitating Mai Yoneyama's style.

However, the Chinese illustrator fdodo once made a video pointing out that AI-generated paintings are still difficult to match humans, and analyzed in detail the difference between AI illustrations imitating Mai Mishan’s style and Mai Mishan’s original works. The film currently has nearly 100 views on Bilibili Millions of plays were played, and the audience discussed it enthusiastically.

fdodo takes the Yoneyama Mai original figure and the Yoneyama Mai AI style work with similar compositions as examples, as shown in the following figure:

Source: Bilibili Chinese illustrator explains in detail the differences between Mishan Wu and AI

In this picture, the top is the work of Mai Yoneyama (click here to see the original picture clearly), and the bottom is the picture generated by AI Yoneyama Mai. fdodo believes that from the perspective of painting skills and creative consciousness, there are the following differences between the two:

1. Difference in rhythm

fdodo pointed out that when processing hair, the hair bundles of AI Yoneyama Mai used the same size and spacing, while the hair bundles drawn by Mai Yoneyama were obviously divided into large, medium and small proportions, and the gaps between the hair strands also had differences in size and proportion. Every graphic is carefully designed.

He believes that this kind of rhythmic control of hair shape is actually a deliberate subjective consciousness of human beings, and the pictures generated by AI do not have this consciousness.

Source: Bilibili Chinese illustrator explains in detail the differences between Mishan Wu and AI

2. Differences in visual center arrangement

According to fdodo, a good illustrator will try to direct the eye to the visual center. Generally speaking, the visual center is usually the place with the largest amount of information and the most interesting aspects of the picture, while the visual weight of non-visual centers will be reduced, making the picture prioritized.

In order to achieve this goal, Mai Mishan used various techniques such as Gaussian blur, increasing the level of bangs, and reducing the level of back hair.

Source: Bilibili Chinese illustrator explains in detail the differences between Mishan Wu and AI

However, many AI-generated images imitating Mishan Mai fail to see such awareness. Even if the hair is far away from the visual center, the AI ​​​​is still very complex and detailed, and there is no difference from the visual center, which will make people feel visual fatigue.

Source: Bilibili Chinese illustrator explains in detail the differences between Mishan Wu and AI

3. Differences in storytelling and expression

fdodo believes that if it is just a gap in painting skills, these difficulties may one day be overcome with the development of technology, but there are still some things that are difficult for AI to replicate: storytelling and expression.

He said that although the images generated by AI are beautiful and delicate, there is always a feeling of "stopping here", while the works of Mai Yoneyama show the shining points of human creators in many parts.

Even though Mai Yoneyama herself did not describe the creative inspiration when she published the illustration, according to fdodo's analysis, her creation is highly interpretable, including the use of Japanese idol images representing obedience and rebellious girlish appearance, which brings Great contrast to create interest.

fdodo believes that the direction in which Mai Yoneyama applies lipstick with her left hand and the design that raises the corners of the girl's mouth makes a supposedly static illustration become dynamic. In addition, there are many floating stones and wounds on the girl's face, suggesting that she just experienced a fierce battle.

Mai Yoneyama also uses graphic emotions, making the girl's hair, face, fingers, clothing and even the shape of her pupils resemble triangles, using sharp graphics to express the girl's aggression.

Source: Bilibili Chinese illustrator explains in detail the differences between Mishan Wu and AI

Simply put, this kind of careful design that serves the author's expression may be the soul of a painting and the key to the work's ability to convey emotions. fdodo believes that at this stage, it is difficult to express emotions through a large number of randomly generated AI pictures through machine learning, and emotions are something unique to humans.

Will AI eventually surpass humans?

fdodo’s discussion triggered heated discussions among Chinese netizens. In addition to the differences between humans and AI in creation, some people also put forward negative views.

Some netizens believe that everything that can be described can be solved step by step, and that the public are generally common people and do not have unique aesthetics. Mishan Wu uses computer drawing tools to create, and AI is also a tool. Although AI is an empty shell at the beginning, after filling in a lot of things, sooner or later it will be much more complex than human emotions.

But some netizens believe that human emotions are not something that can be imitated and learned so easily. The AI ​​doesn't know why it draws like this, so it can only generate it based on Prompt and the original image. Human painters can express rich emotions even if they draw a back figure.

The future of human creators

In order to prevent the moment when AI surpasses human creators, it seems that creators should maintain their obsession and confidence in creation, use the imagination and emotional advantages that surpass AI, and show their own unique understanding and beauty.

Just as at the beginning of the article, Naoki Urasawa said: "I don't want to hand over the fun to AI" and "I'm afraid I will use AI if I don't find fun in painting." (Sunset Car God) film director Nicola Winding ·Hefen also highly admires the greatness of creation. He once said in the documentary "Hideo Kojima: Connecting the World":

"If you are 100 percent true to yourself as an artist, your heart will always beat for the future."