Midjourney, open Pandora’s box

　　A Chinese couple suddenly became popular all over the world.
　　In the photo, a man and a woman are wearing jackets and jeans, sitting on the edge of a messy rooftop, and in the distance are high-rise buildings that look a little out of focus in the sun. The photo looks like it was done by a photographer from the 90s.
　　This pair of lovers, dressed similarly, also appeared in various scenes and different eras.
　　How is this going?
　　Many people were deceived at the beginning. This couple did not exist in real life, but came from a software, Midjourney V5 version. In other words, they are fictional characters “produced” by artificial intelligence algorithms.
　　We’re all used to AI graphics, but the image quality, texture, dynamic range, depth, and detail of this photo of a Chinese couple are so lifelike that the human eye can’t tell which is real.
　　Thinking about it again, it has gone beyond the boundaries of AI drawing. The work of painters, photographers, designers and models, all of them, are all done by one software.
　　The fake pictures not only make it more difficult for viewers to judge the authenticity of things, but also further sound the alarm for human beings.
　　In the past, our imagination of artificial intelligence was limited to the fact that technical jobs would be eliminated, and we, who were freed from labor efficiency, would invest in more creative work.

‘Trump Arrested’ photo by Dutch journalist via Midjourney

A picture made by Midjourney based on Bo Lepeng’s dream

　　We have arrived at an era of deluge of AI-generated information.

　　But now, these well-educated creative jobs, such as photographers, designers, painters, etc., may be replaced by artificial intelligence in the near future. Both your and my jobs may be taken away.
Confused

　　Even in the eyes of professionals, the power of Midjourney is amazing.
　　Bo Lepeng graduated from the Academy of Fine Arts in his early years and is now an art teacher in Guangzhou. Out of curiosity, he specially registered a Midjourney account.
　　Bo Lepeng sent a picture made by Midjourney to the reporter of “Watching the World”: a figure jumped down in the dark night, the sea water below, and a cliff behind the figure.
　　Bo Lepeng told reporters that this picture was made based on his dreams. He is very satisfied with Midjourney’s works: “After trying it, I really can’t stop it. Before that, Midjourney drew the paintings that I couldn’t figure out.” Different from
　　other artificial intelligence drawing software, the first experience of Midjourney is that the platform uses online image resources In one round of feeding, the text information entered by the user is integrated through artificial intelligence data, and a set of rough blueprints full of noise is first generated.
　　If it stops here, then Midjourney is no different from general text-to-picture tools—after all, the outline of the face and the clarity of the hands have always been the Achilles heel of this type of software.
　　However, Midjourney was able to go one step further and remove all the noise through a layer of special processing, finally generating a picture that is no different from a photo.
　　Midjourney is a generative AI developed by the laboratory of the same name and will enter testing in July 2022. The user’s operation is very simple, just send a message through the Discord robot, describe the picture you want to it, and it can be generated. Midjourney iterated to the V5 version in late March this year, and the ability and quality of image production have improved by leaps and bounds. At the same time, it has also become popular, sweeping major social media at home and abroad.
　　On March 18, former US President Trump posted a notice on his personal self-media, saying: “I may be arrested on Tuesday (March 21).” Soon, the pictures of Trump’s arrest were quickly circulated by a large number of people
　　. Forward: On the streets of New York, he was kidnapped and turned away by the special police, and was pinned to the ground by the police; some of him wore a prison uniform to participate in labor reform, and some escaped from prison, so lifelike that ordinary netizens can hardly tell them apart.
　　These decent “news works” come from Midjourney and are produced by a reporter named Elliot Higgins. He is a member of the Dutch group of in-depth investigative journalists “Bellingcat” and claims to be fact-finding.
　　Later, the reporter claimed that his account was blocked by the Midjourney platform after publishing the 50 photos of Trump’s arrest.
　　For a while, Midjourney was pushed to the forefront.
　　Of course, Trump is not the only eye-catching protagonist. It is also widely circulated that Obama and former German Chancellor Angela Merkel played and ate cakes on the beach.
　　On March 27, a group of pictures of Pope Francis wearing a fashionable down coat as a DJ at a nightclub also went viral.
　　All in all, we have arrived at an era of deluge of AI-generated information.
Digital pollution or creativity?

　　Similar generative AI continues to explode, sending a large amount of text, pictures and videos to the Internet world.
　　Are “AI works” a kind of information pollution, or do they represent a trend of graphic and creative work?
　　Not only Midjourney, DALL-E 2, Stable Diffusion and other software, but also can produce exquisite and realistic images in a few seconds according to user descriptions, and various styles.

　　On social platforms, the hashtag “#NotoAI Arts” (say no to artificial intelligence art) appeared.

　　On ChatGPT, we can ask artificial intelligence to imitate the style of a certain publication and write an article. GPT-4 has also been integrated into Office. Now, PPT, copywriting, visual data and email writing, replying, etc. can all be ordered to be automatically completed by AI.
　　Jared Spataro, vice president of Microsoft, also put it bluntly: “One hundred years from now, we will look back at this moment and say: ‘That was the beginning of the real digital age.'” Of course, video generation is no exception
　　. . For example, on Synthesia, as long as you input text or import PPT, it only takes 5 minutes, and it can help you make a video of a real person appearing on the scene, and multi-language options are available, saving a series of labor costs such as directors, actors, photography, and post-production. , as well as shooting equipment and venue costs.
　　In addition to enterprises, students, lawyers, academia, etc. have also begun to use generative AI to assist in their own work. The Writers Guild of America is also proposing to allow participation in screenwriting using ChatGPT, an AI-generated screenplay that can be polished or rewritten by human screenwriters.
　　However, the current generative AI’s working ability and quality of work are still very mixed.
　　The film researcher “Shar Pei” once used ChatGPT to conduct a subtitle translation experiment.
　　He uploaded the srt file for subtitles to the cloud according to the robot’s instructions, and the robot “promised” that it would translate the files in the cloud from English to Chinese as required, and even asked “humbly” what points to pay attention to.
　　The robot “promises” to translate as instructed each time, but ends up sending back an untranslated file.
　　Shar Pei concluded that it was just a word game. The robot made promises in “polite and thoughtful” language, but in fact it kept going in circles, and the work was still not done.
　　In the view of philosopher Harry Frankfurt, ChatGPT is a huge “nonsense” answering machine.
　　He believes that “nonsense” or “clichés” are more poisonous and harmful to language than lies. There is always a certain basis of reality behind the lies, and the “nonsense” produced by the machine answering is often completely decoupled from reality except for prevarication or perfunctory.
　　But the future is endless.
　　Previously, technological progress was linear and predictable: You figured out the steps, and the computer followed suit, says MIT professor David Otter. But ChatGPT improvised.
　　In the transformation of modern technology, no technology has impacted the employment opportunities of higher education talents. Ott believes that the current artificial intelligence may destroy the stability of a large number of white-collar jobs.

Image by Midjourney
Pandora’s Box

　　As for Midjourney, Bo Lepeng told reporters that he felt mixed feelings about its appearance.
　　He believes that his ideal of being able to restore dreams with a brush has finally come true, but since then the process of “dream hunting” has become cheap and common.
　　”I think many designers may really lose their jobs. For the art, advertising, and creative industries, it is definitely an epoch-making change comparable to the invention of the steam engine and electricity. Many people who have studied painting or design skills for half a lifetime may suddenly become different. It’s not worth the money.”
　　Many companies that develop such product technologies claim that the ultimate goal of these artificial intelligence technologies is to free people from tedious technical details and allow people to realize their imaginations at low cost.
　　The emergence of new technologies creates new problems while solving them. Compared with Bo Lepeng’s cautious optimism, the attitude of some artists seems downright pessimistic.
　　Robert Bidolph, a children’s illustrator interviewed by the British “Guardian”, believes that artificial intelligence painting is a trample on art: “In the final analysis, I think art is to ‘translate’ the inner feelings into an external form. The form can be a sculpture, a musical composition, a stage performance or a picture… and real art cares more about this ‘interpretation’ process than the result. Just one click to generate that kind of image without the ‘creation’ process is not art .”

Image by Midjourney

　　As long as artificial intelligence does not replace human nature one day, art will not die one day.

　　On social platforms, the hashtag “#NotoAIArts” (say no to artificial intelligence art) appeared.
　　Those artists who are high-profile against artificial intelligence mapping software believe that relying on a large amount of original data of artworks for artificial intelligence mapping offends the copyright of manual art creators, which can be regarded as an advanced means of plagiarism.
　　”The software is based entirely on infringing countless artists, photographers, illustrators and other legitimate copyright holders, and is not a ‘creative’ process at all.” But after all, Pandora’s box has finally been opened
　　.
　　Whether these artists welcome it or not, the emergence of this technology will have a huge impact on their industry.
　　As an art teacher, Bo Lepeng believes that after the emergence of such software, the problem is not how to teach art in the future; he is interested in what is the purpose of art education.
　　The tension between “mechanicality” and “humanity” has been the focus of people’s thinking since the beginning of the industrial revolution. This kind of tension is even more intense today. While feeling the impact of mechanicalness, we should perhaps first have a deeper understanding of what “humanity” really is.
　　Artificial intelligence may replace some creative work and artistic creation, but as long as artificial intelligence does not replace human nature, art will not die.
　　At least, for a long time to come, artificial intelligence still lacks the ability of self-understanding and self-correction. It can only create in the existing content.
　　And identifying a really interesting, new idea is something human nature is good at.