Dictation｜Professional painters embrace AI in entanglement: it is laborious to remind words, like memorizing mathematical formulas

Source: The Paper

Reporter Shao Wen

· Some AI prompt words will pollute each other, such as "the little girl is eating a lollipop, and there are chickens on the grass in the background", the generated picture is "the little girl is eating fried chicken". Pollution like this is kind of hard to adjust, and by the time I put in the effort to adjust the words, I might have already finished the drawing.

·My colleague already relies heavily on the help of ChatGPT every day. One day before, she couldn't use ChatGPT because of network problems, which made her very anxious.

In August 2022, a work called "Space Opera House" won the "Digital Art" award in an art competition.

"The first batch of people who lost their jobs because of AI have appeared", "Some companies have laid off original painters"...Recently, news about graphic design work being "critically attacked" by artificial intelligence has emerged one after another, leaving people with the impression: the art industry Practitioners are trembling in the atmosphere that the rain is about to come.

"What a game company needs may be just one person doing repetitive things in a certain position, not everyone can make a complete picture. So when such a technology is released, the possibility of people being replaced is very high." Ban Ban, an illustrator, told The Paper (although outsiders always use AI to generate pictures to laugh at professional painters, insiders will fight each other and feel that it is disgraceful to use AI to generate pictures. "The generated picture It’s still a bit conceptual, just like a template, you always feel that you’ve seen it somewhere, you can’t say it’s your own, and you may be criticized by others if you send it out.”

Wang Xiaomi, a 29-year-old data visualization designer, said frankly that she used Midjourney, a well-known AI drawing application, to generate more than 1,000 pictures last month. a feeling of". With a five-year age gap, she and her colleagues had disagreements about how to use new technology.

Narrator: Illustrator Banban, 26 years old

The phenomenon of "scrambling corpses" appears, and everyone will be a little bit "controlling violence with violence"

The illustration circle was against AI at the beginning.

On the one hand, outsiders will directly use AI-generated pictures to laugh at those professional painters for their poor drawing. Maybe others just post what they have created to express a feeling, but some people will ridicule that there is no AI to draw well. This happens a lot.

On the other hand, there are also some people who take pictures of some masters to "pad" (pad pictures, that is, use reference pictures to generate the photos they want), or directly use other people's pictures to generate a picture, and then say it is their own , and even take it commercially. I heard someone describe this behavior as "scrambling corpses", that is, to use all the pictures in the database as their own resources, and then put together a picture, but in fact this picture is not the work of this person.

One thing was very famous in the circle at that time. A veteran painter saw too many people on the Internet laughing at others, or seeing people who lost their jobs and gloated over others, so he posted a sketch of the same person's demonstration of Elden's Valkyrie on the Internet, announcing that 30 minutes later Send the final product again. During these 30 minutes, everyone can use AI to refine this draft, and finally compare the generated drawing with the artist's drawing. Obviously, his drawing is much better than the one generated by AI.

He was fighting for those who were ridiculed, and I agree with some of what he said. He said that AI is a very new tool, and the infringement problem is also serious. We are also continuing to observe, and the international community is also thinking about how to legislate and regulate it. During this period, it is normal to use and research out of curiosity or work requirements. But what he opposes is exaggerating the role of AI, and then using a few keywords to dissolve pictures to generate some high-precision cool pictures to create anxiety.

At that time, the overall atmosphere was not good, and everyone would be a little bit "controlling violence with violence". Some people will say that since you use AI to abuse other people's pictures, then the pictures you generate have no copyright, and I can use the pictures you generated to generate new pictures. There are also people who use other people's pictures to "pad" out a more satisfactory picture and say that it is copyrighted because I worked hard to figure it out.

I myself now have the mentality of wanting to use AI, but I don't have the feeling that "the generated graph is my work". And even if I generate a better picture, I will not have a sense of accomplishment, and I will even feel a little guilty at the beginning. Up to now, I don't "pad" the pictures of famous artists. I like some pictures of independent studios, but I still don't put them in, because I feel sorry for others.

I know that technological progress is the general trend, so I also understand learning, and now I spend more than an hour a day on learning. My own feeling after using it is that it can provide some references, but it is not really useful to complete a task. For reference, for example, I drew a group of youth-themed pictures before, and then I added a picture of my own, telling it to generate a series of the same style. Then, briefly describe the effect I want. Its generated result is a very template style, but some of the colors and composition are good, you can refer to it.

But it’s true that I can’t use it directly. Apart from my own embarrassment, I think there are several reasons. One is to think of prompt words. I really don’t understand some photography vocabulary very well. Some words will also pollute each other, such as "little girl eating a lollipop, and there are chickens on the grass in the background", the generated picture is "little girl eating fried chicken". Pollution like this is kind of hard to adjust, and by the time I put in the effort to adjust the words, I might have already finished the drawing. Although I'm not that good, I feel that I can draw faster by myself at the moment. In addition, the direction of generation is unstable, and legal regulations related to copyright have not yet been determined. In addition, the generated pictures are still a bit conceptual, just like templates, you will always feel that you have seen it somewhere, there is no way to say that this is your own thing, and it may be criticized by others when you send it out.

I still have such confusion and don't know how to face this technology. The principle in my heart is that I can't use too much, because it is basically someone else's picture. But if I have to use it at work, I will be a little confused. Sometimes I also worry about whether others think that my drawings are not generated well, or that my efficiency is not as high as AI.

There is a difference between drawing a picture and generating a picture. Drawing is an eye sensation, and with muscle memory, you quickly catch where there is a problem and what color needs to be used. To generate a picture, it is necessary to clearly express the feeling seen by the eyes in words, which seems to be another ability to exercise. When I was studying, I also felt that it actually helped to clarify my thoughts, and I would analyze more rationally what kind of picture I wanted.

But I personally still prefer to draw directly, so that I don’t feel like I’m drawing when I turn it into text. It seems that drawing by hand will be more emotional. I also like what a friend said. She said that no matter how technology develops, handmade lace must be the most expensive in the end. Tools can be played and touched, and you can also stick to your own principles and not use other people's pictures. The important thing is not to forget that you should improve yourself because of the emergence of technology.

For this improvement, I think on the one hand, it is a bit like memorizing mathematical formulas to learn prompt words, including some basic knowledge, such as perspective. Because it puts a lot of emphasis on professional words, if the words are not typed correctly, the machine may not be able to understand, and using a professional term may be faster than typing a long description. On the other hand, it is the theoretical basis of art, including aesthetics. I may not have listened to the history of art when I was in school.

I will not give up painting with my own hands, and I regret that I didn't learn something in college, and now I try to learn as much as possible. But sometimes I also feel that I can't learn a little bit. When using AI, I will always pay attention to what prompt words to use. This is also a source of anxiety. I don’t know where to focus my energy if I want to improve. If I take care of both ends, it may end up being mediocre, and then I will become more anxious.

Some of my friends are anxious, and some are just outright rejecting the technique. Like a tattoo artist friend, he calculates the commission based on the drawing, which is completely hand-painted and has his own style. Therefore, when he sees someone in the studio next door using AI to mass-produce tattoo drawings, he will be very anxious. On the one hand, others will earn a lot more, and on the other hand, he wonders whether he will lose his job due to technological development in the future.

Another friend is an original artist, who is also the friend I just mentioned the word "handmade lace". She completely refuses to use this technology. But she is not an original artist in the game industry. Although no one I know has been laid off, I heard that many people in the game industry have been laid off. In fact, what a game company needs may be just one person doing repetitive things in a certain position, not everyone can make a complete picture. For example, in a large scene, someone draws the background, someone draws the main person, someone draws small objects, etc. It has a lot of similar things. So when such a technology appears, the possibility of people being replaced is very high.

Oralist: Wang Xiaomi, data visualization designer, 29 years old

The age difference is 5 years old, and the way of embracing new technologies is different

Maybe because my type of work is relatively small, I don't think AI has a particularly big impact on me at present.

My job content is mainly to find some relevant data after the editor determines the topic, and then make charts and visualizations for these data, or do some web design for interactive projects. In fact, five or six years ago, there were many websites that generated charts online, but for so many years, manual participation in design and layout was still required.

I currently use ChatGPT as an upgraded version of Google. For example, if I learn a 3D modeling software, I can ask some questions encountered during the learning process. But because the software I am learning is open source software, its upgrade is very fast, and the knowledge base of ChatGPT is still last year, which makes me lose the motivation to chat with it. However, there is a colleague of the same type of work as me. She is still in the initial stage of learning programming languages. She feels that the user experience is particularly good, and she already relies heavily on the help of ChatGPT for her daily learning. One day before, she couldn't use ChatGPT because of network problems, which made her very anxious.

Most of our group are people with strong learning self-motivation, so facing ChatGPT is like facing a newly released software, everyone should learn it, as long as they can answer their own questions or improve. I also used Midjourney to generate a picture yesterday, and I didn’t tell my colleagues that it was generated by AI. Then they praised and commented, and finally told them that they were cheated.

Some time ago, Midjourney launched the cute mode, Japanese mode, etc., and the cute mode has other modes, and it will be updated every few days. Sometimes, as my colleague said, I can't learn it as fast as it updates. However, it is difficult to say what direct help it will bring to my work, but it feels like it will broaden my horizons and use my imagination to create something else. For example, I may use it to make a poster. Like before, I had to go to the online material library to find materials for a poster. Now I use it directly to generate and sometimes get faster feedback. I bought Midjourney's monthly membership with a colleague. Last month, I used it to generate more than 1,000 images, and this month I used more than 300 images in eight or nine days. We use it like crazy every day, because it does update very quickly.

This also has some negative effects. For example, I used to learn 3D modeling software, so I learned it step by step according to the norms. After this technology comes out, it will disrupt your pace to some extent. Because sometimes it is really good, you will think that it has been done so well, should I continue to learn.

But I think the generated one is not the same as the one I made. I usually also draw, whether it is on an iPad or on paper, the feeling that this brings cannot be replaced by direct generation. Especially when you finish a painting, even if the painting is not very professional, but the feeling of the color appearing on the paper will give you good feedback, I think this is different.

In terms of the growth path of technical ability, it is not the same as before. The designer of the same type as me was born in 1999, and I was born in 1994, a difference of five years. The same way to learn code, I just follow books and tutorials in a very traditional way to learn, and follow the code. But his current learning path is to know what kind of graph I want to make, and then tell ChatGPT the data or final presentation of the graph, and let it generate code. When ChatGPT tells him the code, it will also leave comments on each line or section, such as what this step does. Through this method, good results have also been obtained.

It's just that my friend and I have some conflicts. My way of learning is more solid, but it may take longer. Others don't have any code foundation, and can use ChatGPT to make some effects, but they may be confused about why a certain part does this, and there will be a question of whether this is their own achievement. For example, I think that using code to write down the way I think step by step, and finally produce results, this is mine. If it was generated directly for me, I would feel less real, and I would not consider it to be my work, and the originality would be lost.

(At the request of the interviewee, the characters in the article are all pseudonyms)