Image source: Generated by Unbounded AI

The WeChat public account backend has recently been updated with a new feature.

When the article editor chooses to insert a picture, in addition to the original selection from the picture library and local upload, a new option is added - "AI matching".

Click this option, and the creator will enter an AI image generation page. You only need to enter a text to describe the image you want, and then wait for more than ten seconds, and the system will generate four photos. Based on these four photos, you can further select the image style and proportion, and then further revise the image until you are satisfied and insert it into the article.

The addition of this feature means that in the future, images for WeChat public accounts can be generated through AI, and creators no longer have to worry about the risk of image copyright.

▲Screenshot of WeChat official account backend

WeChat official account is not the first platform to update this feature. In early 2023, Baijiahao launched similar functions shortly after Wenxinyiyan was released. The significance of the WeChat official account update is that as the earliest and most influential platform in the industry, but the slowest to update, its change represents the end of one stage and the beginning of a new stage.

Coincidentally, at almost the same time as this incident occurred, information about ByteDance’s sale of Tu Chong was disclosed.

Tu Chong, a community platform for photographers to communicate, or it can be more simply and crudely understood as an online gallery, was fully acquired by ByteDance in 2016.

At that time, Toutiao had just been launched for four years, and the graphic content platform was expanding rapidly. ByteDance has only been established for two years. Tuchong is this young company’s first external investment, and its purpose is self-evident. It acquires photo galleries to provide copyright materials for graphic content on Toutiao. , thereby avoiding risks.

Nearly ten years have passed in the blink of an eye. ByteDance sold this first external investment at a time when graphics and text are gradually declining and AI is developing rapidly. Combined with the changes in WeChat official accounts , it is difficult not to ask a question - has the photo library completed its historical mission?

So is this the case? In the AI ​​era, has the photo gallery lost its value?

Confrontation and compromise in the gallery

I don’t know if it’s fate , the first thing that generative AI is known to the public is to give traditional photo galleries a disastrous impact.

On August 29, 2022, an expo was being held in Pueblo, Colorado, USA.

Since 1872, holding a fair in late August has become a Colorado tradition, even though the fair has been losing money for 21 consecutive years.

But this expo in 2022 is destined to be different, because this year a painting titled "ThThe painting "Éâtre D'opéra Spatial (Space Opera)" won an award at the exposition.

The special thing about "Space Opera" is that it is a work created by the creator Jason Michael Allen first using Midjourney and then polished it with PS. It won an award at the Colorado Exposition, making it one of the first works in history to be created using AI and win an award.

▲Picture "Théâtre D'opéra Spatial" Creator: Jason Michael Allen Source: Wikipedia

Although "Space Opera" won the award, it became quite popular on Internet platforms such as X Controversial, but from an industry perspective, its impact is no less significant than AlphaGo’s victory over Lee Sedol in 2016.

Just like the meaning of the first car to the carriage, many times the rebirth of one thing often means that another thing is about to die. The same is true for the story of AI and traditional photo galleries.

This is also true. Soon after "Space Opera" won the award, more and more creators began to use AI tools such as Midjourney and Stable Diffusion to create images and upload them to paid stock photo platforms such as Shutterstock. Earn income on.

Tuku's initial attitude was to refuse, and his response was very quick. Less than a month after "Space Opera" won the award, Getty Images, one of the top photo galleries in the United States, issued an announcement banning the uploading and selling of images generated using AI. All will be deleted.

Following Getty Images, picture websites such as Newgrounds, PurplePort, and FurAffinity also made similar decisions.

▲Picture: Getty Images and its platform iStock issued an announcement banning AI images Source: official website screenshot

As for the reason for the ban, Getty Images CEO Craig Peters said that based on Considering the legality of the content, since many AI-generated images use generated materials that do not have copyright permission, hasty sales may bring legal risks to customers and the company.

But what is surprising is that less than two weeks after the ban was announced, Getty Images cooperated with Nvidia to launch its own AI generation tool Generative AI by Getty Images.

What’s special about Generative AI by Getty Images, according to Getty Images CEOThe reason is that it is "licensed" and is only trained on the creative content of Getty Images. Because it fully complies with commercial safety standards, there will be no third-party intellectual property disputes, and we will also pay "remuneration" to the creators of these contents. ”.

Shortly after Getty Images, Shutterstock, another U.S. company that had also previously rejected AI picture headers, also cooperated with OpenAI to launch its own image generation tool.

Similar to Getty Images’ choice to pay image creators, Shutterstock has established a “contributor fund” and claims that if a creator’s images are used for AI training, they will be compensated by this fund.

In short, with the rapid spread of AI images, traditional photo galleries chose to embrace it without hesitation after a brief period of resistance.

The way they embrace AI is to tie themselves to the creators of picture content, chant the slogan of protecting intellectual property rights and the fruits of photographers’ labor, and pursue the transparency of transaction data.

But can this continue to maintain the value of the photo gallery in the AI ​​era? The answer may not be yes.

Choose AI or people?

From a commercial perspective, the core value of the photo gallery is to improve the efficiency of image content circulation while protecting the creator's intellectual property rights.

China’s earliest online photo gallery was Photocome, founded in 2000 by Chai Jijun, then photo editor and photojournalist of China Youth Daily. It was also the predecessor of Visual China.

At that time, newspapers and magazines were the most common settings for photos. As a picture editor, Chai Jijun receives about 300 sets of manuscripts every day, but the pictures that can be selected are very limited. Photos that fail to be selected are either transferred to other media colleagues for use, or they can only be thrown away, resulting in a lot of waste. [1]

At the same time, around 2000, portal websites such as NetEase, Sohu, and Sina were established one after another. The development of the Internet led to an explosive growth in demand for pictures, which was the first wave of east wind for photo gallery websites. It was against this background that Chai Jijun founded Photocome.

But no one expected that the news media industry, which put photo galleries on the fast track 20 years ago, would be the first industry to abandon photo galleries more than 20 years later.

At the end of 2023, domestic technology media 36Kr began to try to use AI-generated pictures to replace photographers’ works purchased from photo galleries. Soon after, content platforms such as Titanium Media and Huxiu also began to gradually follow suit. Nowadays, on these three platforms, AI-generated pictures have basically replaced the work of photographers.

The reason why the platform chooses AI to generate images is very simple. First, it is low cost, and second, it is low risk.

The first is low risk.According to current domestic laws, AI does not enjoy copyright. The copyright of images generated by AI belongs to the individual creator. Therefore, if you use AI to generate images, you no longer have to worry about suddenly receiving a request for hundreds of thousands of compensation from someone unknown. Lawyer's letter.

The second is cost. If you cooperate with a photo gallery website, the price of a photo usually ranges from a few hundred yuan to several thousand yuan. Even an annual agreement can easily cost hundreds of thousands of yuan.

Using AI to generate pictures, Midjourney’s unlimited number of members only costs $60 (approximately 434.5 yuan) a month. And if you choose to use the API of a third-party large model to build your own application, the most expensive large models in China, such as Tongyi Qianwen and Wenxinyiyan, have input and output prices of only 120 yuan/1M tokens.

Depending on the resolution of the output image and the output efficiency of the model, these 1 M tokens can output approximately 10,000 to 30,000 images. In contrast, using AI to generate images is almost free of charge.

From this perspective, images generated using AI can almost perfectly replace the commercial value of photo galleries. It's no wonder that all the major picture websites are rushing to develop their own picture generation tools.

But does this mean that the photographer’s work is doomed to be eliminated? The answer is obviously not.

On July 13 this year, Trump was shot while giving a presidential campaign speech in Pennsylvania. Associated Press photographer Evan Vucci took the camera and rushed to the scene immediately, risking being shot. Went in front of Trump and took a photo of a historic moment - Trump, with his face covered in blood, was supported by the crowd under the Stars and Stripes flag, with his arms raised and clenched fists, giving people a strong visual impact.

It’s hard to say how much this image contributed to Trump’s subsequent victory in the US presidential election.

▲Trump’s life photo photography: Evan Vucci

As a comparison, although AI pictures are being used more and more widely nowadays, they do not have this kind of effect at all. The inspiring power conveyed by the photos.

In fact, the current images produced in batches by AI still appear very cold and rigid, lacking the vitality unique to human creators. This may be a bit abstract, but you can clearly feel it just by looking at the comparison of the two pictures.

▲Comparison of AI illustrations and photographer illustrations

It can be said that if a picture is compared to a delicious meal, then the picture generated by AI is like a piece of whole wheat bread. Rough, dry, and has nothing to do with food other than providing energy and satiety. On the contrary, the pictures taken by photographers are like the eight major cuisines, which can stimulate people's rich taste buds.

That is, when pictures are mass-produced by AI, the meaning of pictures becomes to meet the needs of basic picture usage scenarios, and people’s pursuit of beauty, personality and humanity are naturally ignored. .

And this is precisely an important breaking point for photographers to fight against AI.

The scary thing is not AI, but the loss of imagination

Logically speaking, it seems difficult for AI to replace the work of photographers, and naturally it is also difficult to replace photo galleries, but the reality is often worrying.

On the one hand, content consumers don’t seem to care whether the pictures have soul or stories.

Because as AI images are increasingly used on various content platforms, we have not seen any content consumers raise objections to this, or in other words, there are not enough people to raise objections. Let managers of content platforms change their strategies.

On this basis, content platforms naturally ignore the needs of content consumers for content quality. After all, compared to creating a high-quality work in exchange for a bonus item that consumers may not care about, using AI libraries to bring about fundamental reductions and risk reductions is an immediate profit.

In this way, consumers’ default and content platform’s neglect have reached a consensus to some extent, and together they have promoted the rapid expansion of AI picture content. Although many people may be aware of the shortcomings, no one cares.

And this is the real challenge faced by image content creators.

Looking back, we actually need to think about such a question. When AI makes it easy for humans to obtain any picture, what is the meaning of the picture?

Those who encountered this problem earlier were creators who made a living by writing. Does writing still make sense when people only need to make a request in a dialog box of a large language model to obtain an article?

As writers, we have also thought about this question carefully. What is the meaning of the article? Is it an arrangement of words?

The answer is obviously not. We believe that the real meaning of words lies in the thoughts and emotions conveyed by the creator through words.

In 2003, Liu Cixin published a novel called "Poetry Cloud". In the novel, a high-level alien civilization exhausted most of the energy of the solar system in order to write transcendent Li Bai poems. , spent ten years listing all possible combinations of Chinese characters.

But the aliens did not write "poems that surpassed Li Bai" in the end, because the aliens discovered that although in all possibilities exhausted, there must be poems that surpassed Li Bai, but they had no way Find it out.

This is the meaning of words and poetry, that is, "technology can never replace art", and the same is true for pictures.

And if one day, AI production replaces all original content, it may not be a victory for AI, but a tragedy for mankind.