Shulou(Shulou.com)11/24 Report--

On March 15, AI drawing tool Midjourney released the fifth generation V5 model, by which a picture of "Chinese lovers" is described by some AI art lovers as "perfect" because its effect resembles that of a real person.

In less than a month, the Internet has been flooded with images generated by netizens using Midjourney, including portraits, commercial advertisements, city street views, and even run-down images of celebrities such as Trump and Obama living on the streets. Do you still remember the Space Opera House, which won the award last August? That painting was also created in Midjourney. At this time, it was only one year before Midjourney first appeared in March 2022.

What is surprising about ▲ 's "Chinese couple" photos generated by Midjourney is not the emergence of AI, but the rapid iteration of AI technology that is getting closer to real photography. In the eyes of its creators, its iterative level is equally exciting. "Midjourney V5 now feels like I finally put on my glasses after a long time of poor eyesight," said Julie Wieland, a graphic designer using Midjourney. "all of a sudden you see 4K, and it's incredible but amazing."

This whirlwind has also spread to the photography industry, and a large number of professionals have begun to test the waters of AI creation. On social media platforms such as Wechat, Xiaohongshu and bilibili, some photographers posted pictures or videos documenting the use of Midjourney V5, which also sparked a lot of discussion about whether AI technology will replace photographers.

01. Superb AI is not perfect. What is the effect of AI technology?

Wu Xiaolong, a photography teacher at East China normal University, tried out two versions of Midjourney V4 and V5 and felt that "the effect of V5 is really more realistic."

The image generated by ▲ Wu Xiaolong through Midjourney V4 "the above picture was generated by V4 at that time, using black, retro, indoor, listening to music, lighting and other key words, except for six fingers, everything else is perfect. The portrait generated by V4 is basically a person's head shot, relatively single, to V5 is very diverse, there are all kinds of scenes."

In another test by Wu Xiaolong, the Midjourney V5 showed amazing imitation ability. As long as you enter a series of simple instructions in Midjourney: photo by xxx, where xxx is the name of some famous photographers, AI can accurately identify and present the photographer's creative style and texture, and render four images to choose from in less than a minute.

The image generated by ▲ Wu Xiaolong through Midjourney V5 shows the performance of Midjourney V5. Wu Xiaolong gives four words: "surprised that I simply typed a few words and it came out with a complete picture. The good news is that although it is not exactly the same as I thought, it also has a sense of beauty and is richer than I thought."

Long Erhuan, a blogger who is interested in AI technology, is the first to use Midjourney V5 to generate two groups of AI images of "Beijing Hutong 10-year-old" and "Pink Doll stereotype". She believes that the performance of Midjourney "can even be on a par with real photography".

"although the V5 version is slightly poor in hand and text performance, the precision of the character portrayal is unexpected. Before I used it, I imagined that it could almost depict a person's face, as expected; unexpectedly, it could depict a person's facial expression more precisely than expected, and the relationship between it and the background was more harmonious."

Midjourney V5 presents superb painting skills, but still can not achieve perfect interaction. The experience of photographer Z after trial is that keywords have a great influence on the result of AI image generation.

"people who are proficient in English and understand the technical terms of photography will get better results by using AI software such as Midjourney. If you know the basic words of photography, such as light, scene, scale, color temperature, etc., and know the art genre, professionals who are well versed in photography aesthetics, using AI software will get better results."

In this regard, photographer Z also gives a further example: if you want to create a creative picture, if the instructions given to AI are news headline descriptions such as "Weekly funding and fighting against Confucian scholars", it may not be able to directly produce the ideal results, and AI must get more accurate descriptive instructions.

In addition to the strong skill required to "feed" keywords to AI, some photographers will also find clues of "stereotype" in AI.

"I asked AI to generate five female photographers holding cameras to take pictures of a man in a pink T-shirt. As a result, AI generated a man holding a camera and a woman wearing a pink T-shirt. No matter how you change the keywords, you will not get the desired results." Dragon second lap encountered such a "speechless incident", "let AI produce a lot of Chinese children, these children often do not look so 'sunny', clothes are often shabby, a kind of sadness, until I have to add the keyword'a rich Chinese child'."

Behind the "stereotype", is the database obsolete and the photo material subjective? Or is there a stereotype in AI's logic? The photographers couldn't give an answer.

02, AI will not become such 0.01% although the interaction of Midjourney V5 is not perfect at present, its phenomenon has led to a large-scale debate about "AI replacement photographer". There is no lack of netizens to comment on "photography no longer exists" and "AI will replace photographers" under the generated works.

Whether AI technology will replace photography may have to be discussed separately, and the more common view in the industry is that "commercial photography is the most dangerous under AI technology."

In long er Huan's view, part of the stereotyped commercial photography will be replaced. "it is conceivable that a skilled'AI trainer 'uses AI to generate images at a lower cost, in less time, and more in line with the needs of customers who want to change to satisfaction over and over again."

The image generated by the second circle of the ▲ dragon through Midjourney V5, while Wu Xiaolong further proposed that commercial photography will comply with the law of "rigid demand maintenance, low-end elimination, and high-end change" in the AI era. " The rigid demand is certificate photos and wedding photos; some low-end uncreative photography will be eliminated by AI; high-end ones, such as star ads, also need to change, otherwise, as long as face data is authorized, it can also be generated by AI. "

News photography, documentary photography and art photography are more "optimistic" in the face of AI technology. In the eyes of photographers, AI technology is more equivalent to computing power, or just a possible auxiliary tool, while photography is still people-oriented. "people will need good news photos more, and the demand will increase because of AI, and those who see too many fake ones want to see the most real ones." Wu Xiaolong analysis.

Photographer Z takes a similar view, arguing that AI cannot achieve news timeliness, human creativity and spirituality. "machines ultimately imitate human beings, but they cannot replace human creativity. When a photographer takes pictures with mood and character, he shows not only pictures, but also unique human compassion, care and love. As a Google engineer said, human spirituality can never be replaced."

At the same time, AI does not seem to be able to reach the level fully recognized by photography professionals, although the creation is amazing, but people in the industry can still see that it is not "real". Wu Xiaolong believes that "AI is still short of breath, only the surface is similar, there is no core, can not stand in-depth research."

Especially in art photography, we only want AI as an auxiliary. "Art photography requires the creator to have the feeling of the world and rebellion against mediocrity, but AI has no emotion, and the images it produces are more based on machine learning and probability than the feeling of the world. In art photography, AI is more suitable to help artists gather information." The second circle of dragon said.

Photographer Z gives a similar point of view, thinking that AI may bring some inspiration to contemporary art. "machines must have seen more than people have seen, so machines may give people some sources of thinking. In the future contemporary art and commercial creation, people can actually use AI to draw imaginary scenes for themselves as a verification before actual shooting."

Overall, it seems that people are not worried about AI replacing human beings. As photographer Z said, it will take a century or two for a master of art like Tarkovsky to be born. They greatly expand the artistic boundary. AI will not become such 0.01%, but can only bring some enlightenment to people.

03. The era of picture and truth has passed. With the "high IQ" AI set off a new round of AI innovation, people seem to have high expectations for AI in various fields. Nvidia founder CEO Huang Renxun called this round of technological innovation "the iPhone moment of AI". He believes that AI can help human beings open their "brains", stimulate human creativity and improve production efficiency.

In the field of photography, it is true that there have been many discussions that take a positive view of AI. As a university teacher, Wu Xiaolong is more inclined to actively guide students. "AI will not take away their jobs, but will create more possibilities and should take the initiative to welcome change." Recently, photographers have also been thinking about how to apply AI technology to the profession.

But even if the photography industry takes a positive view of the creative changes brought about by AI, they are equally concerned about another "risk" brought about by the evolution of related technologies. As one netizen wrote in a comment, "seeing is believing'is getting farther and farther away from human beings." when the photographic works produced by AI become more and more realistic, how can people tell what is real?

▲ August 2022 Chongqing Mountain Fire, there are a large number of AI-generated fire pictures on social media, and the photography industry in the future seems to need more tools to verify the emergence of AI. "I think there must be some rules to identify to prevent AI generation from being real. Maybe we can develop an App to distinguish whether it is a real photo, such as a 'photo authenticity discriminator,' to determine whether the photo is generated by AI," long said.

"the era of where there is picture, there is truth is over, and we really need to face these problems." Photographer Z also believes that the pace of technological change is unstoppable. "once Pandora's magic box is opened, people can't stop exploring. The only thing we can think about is how to create confrontational tools to curb the legal and moral hazard caused by AI to generate images, and how to avoid people and organizations who master AI math, use AI channels to influence public thinking, and control technology to have a say."

Today, it is still impossible to judge how much impact AI technology will have on photography and its related industries.

But predictably, various industries in the torrent of AI will begin to think about the question of "how to live with artificial intelligence".

This article comes from the official account of Wechat: slow play (ID:manfangsd)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.