Today is Journalist Day. Not long ago, there was some teasing in the media circle that AI can also write special reports.
I fed some sample articles from special report platforms to general intelligent AI assistants for imitation, and at first glance, they looked quite decent:
<img src="https://imageproxy.pbkrs.com/https://img.huxiucdn.com/article/content/202411/08/142255030244.png/query-aW1hZ2VWaWV3Mi8yL3cvMTAwMC9mb3JtYXQvcG5nL2ludGVybGFjZS8xL3EvODU?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://img.huxiucdn.com/article/content/202411/08/142255030244.png/query-aW1hZ2VWaWV3Mi8yL3cvMTAwMC9mb3JtYXQvcG5nL2ludGVybGFjZS8xL3EvODU"/>
But upon closer inspection, it's clear that AI's imitation either resembles online articles or PPT summaries, and cannot actually be published on special report platforms.
This serves as a wake-up call for media professionals: under the increasingly formulaic news content and headline patterns, AI can easily learn human &#34;tricks.&#34;
Will AI replace journalists? This question seems to have become a cliché.
This question is too &#34;news-centric&#34; and &#34;journalist-centric,&#34; and perhaps only those who rely on news for a living would care. In the bustling world of various professions, everyone is concerned about their own industry, and most people view news merely as a tool for promotion and traffic generation.
In today's content dissemination ecosystem, the term &#34;journalist&#34; feels outdated, and &#34;self-media&#34; seems too vague. The competition for traffic has shifted from Douyin and Kuaishou to WeChat Channels and Xiaohongshu, moving from public domains to private domains. Traffic is a business, and the vast majority of traffic is part of a business model. The data from likes, collections, shares, and follows are not only for monetization and real revenue but also a craft of understanding human nature. Institutional journalists are, to some extent, also part of this ecosystem, striving for survival. In the content dissemination ecosystem, the current voices of journalists are far outmatched by the super individual IPs on Douyin, Kuaishou, WeChat Channels, Bilibili, Xiaohongshu, Twitter, YouTube, Facebook, and Instagram.
Moreover, real journalists do not care whether they will be replaced by AI. Journalists who have truly been on the ground know that it won't happen. Large language models are powerful enough to rival a university student's brain, but they are still just a brain in a jar. AI can serve as an excellent data analyst or data analyst, but it cannot reach the heights of power, traverse the vastness of the world, negotiate with elites in luxurious clubs, or share experiences with grassroots people on street corners. Such journalists do not worry about being replaced by AI. It has never been AI that replaces journalists.
My graduate thesis involved visiting the rural hometown of Vietnamese brides, finding motorcycle drivers to take me to unfamiliar farmers' homes in the fields without public transportation, to inquire about the unnamed Vietnamese brides. My teacher thought it was still insufficient and asked me, &#34;Can you stay at the Vietnamese bride's home for a few days?&#34; But I still cannot consider myself a true journalist. Upon graduation, a classmate told me that younger journalism students who took online classes during the pandemic might not even know how to conduct a social report.
People in the content industry feel that the question of &#34;AI replacing journalists&#34; is too narrow. There are many jobs that could be replaced by AI: Will AI replace illustrators? Will it replace advertising professionals? Will it replace photographers and directors?
The real question is, how will AIGC reshape the entire content dissemination industry and change people's habits of consuming content?
The vast majority of users have unknowingly consumed a lot of AIGC. AIGC is becoming increasingly realistic.
Which of the following four images is real?
<img src="https://imageproxy.pbkrs.com/https://img.huxiucdn.com/article/content/202411/08/142256845616.png/query-aW1hZ2VWaWV3Mi8yL3cvMTAwMC9mb3JtYXQvcG5nL2ludGVybGFjZS8xL3EvODU?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/https://img.huxiucdn.com/article/content/202411/08/142256845616.png/query-aW1hZ2VWaWV3Mi8yL3cvMTAwMC9mb3JtYXQvcG5nL2ludGVybGFjZS8xL3EvODU"/>
The answer is, all are AI-generated.
However, where there is a devil, there is also a way to counter it. If AI can replace the real with the fake, it can also be used to debunk fakes. By training models with real data and AI-synthesized data, the models can learn to distinguish between true and false, and even identify AI-generated images that are not recognizable by the human eye. In the performance competition of the AI Innovation Contest at the Bund Conference, the forgery recognition rate was almost above 80% on some moderately difficult tasks.
Thus, fact-checking has become one of the few skill barriers for journalists. From InVID WeVerify to RevEye Reverse Image Search, there are various tools available to detect the original source of images and whether they are synthesized.
Technically, there are always ways to debunk fakes, but the problem is that users seem to care less about the truth and more about the immersive experience and atmosphere. Using AI to resurrect deceased relatives or pets, to animate old photos, or to have conversations between people separated by time and space... Even if we know this is fake, we still pursue immersion and even spend lavishly. We become indifferent to whether the generated content is true, focusing instead on whether the content aligns with our desires and intentions. Even if the content disseminated is false, the audience's inner desires are real. Most of the time, since the deceased cannot speak for themselves and relatives obtain portrait rights, resurrecting the images of deceased relatives is permissible with the consent of family members. However, the problem lies in the many gray areas where interests grow; once AI is used for forgery, such as making a white celebrity speak Chinese or share emotional insights about female growth and male pain, it can easily drive e-commerce sales. These wild-growing gray areas still need legislative regulation.
From an industry ethics perspective, how the news industry should use AI ultimately boils down to the same principle—how to maintain objectivity, truthfulness, and accuracy. No matter how creatively AI is used, if it deviates from the original meaning of events and individuals, it cannot be used in news. David Schlesinger, former editor-in-chief of Reuters and former chairman of Thomson Reuters China, stated that the most important thing is not to use someone's image without permission and not to use it out of context. He gave an example to Huxiu: if someone obtains an image of a person speaking but uses AI to dub it in Turkish or some other language instead of the original English, that distorts the facts. A magazine once used a photo of him and his wife but cropped him out, leaving only his wife's part to convey loneliness, which completely changed the original meaning of the photo
From the perspective of content style, AI excels at refined integration but struggles with authentic rawness. AI can learn and blend various styles to generate sophisticated, high-concept content. A landscape photographer who won the global championship of the National Geographic Photography Contest stated that the more connected to reality documentary photography and news photography are, the harder it is for AI to replace them. This is bad news for advertising directors and retouching software, but good news for journalists and documentary filmmakers.
However, AI is not just a competitor in media content, as it can produce content and is itself a new type of medium. On Midjourney, you can use over 200 style vocabulary words in your prompts, such as Voxel Art, Chainmail Art, Soft Sculpture, Geode Art, Impasto, and countless other styles, to create a composite of various stylistic concepts, and then use tools like ControlNet, which are similar to brushes, to adjust the details. If painting is about using physical paint on a canvas, then AI is about painting with concepts, merging the visual styles of different concepts through algorithms. The prompt is your hand in painting.
In the view of AIGC blogger Ye Pusa, like painting, film, and XR, AI is a new medium, but it has not yet developed an expression method suitable for the characteristics of the AI medium itself. He believes that using AI to make movies is foolish, just as it was foolish in the second half of the 19th century when the &#34;pictorial photography&#34; movement tried to imitate classical painting with photography, a practice that would be eliminated by history. The expression method most suitable for the characteristics of the AI medium lies between video and games.
&#34;In the future, media will die, but content providers will not die,&#34; blogger Ye Pusa believes. &#34;There won't be as many (online traffic) entry points; there will only be entry points like Siri or ChatGPT. You won't need to go into the media to get information; you just need to tell ChatGPT what you want to know, and it will provide it directly. But content providers will not die; the key is whether what you write has style, whether it can become a concept, and whether people in the future will use your name to define a style.&#34;
Even for fact-based news, under the trend of AI, we see that news writing/videos that are increasingly formulaic and lack style are more easily replaced. Those who appear on camera in their pajamas, live-streaming while walking on bridges or streets at midnight, are merely showcasing their most authentic selves to create a style. May we all live out our own truths, remain true to reality, and create our own styles

On the occasion of Journalist Day, the media circle discusses whether AI will replace journalists. Although AI can mimic feature articles, its content often lacks publication value. With the changing ecology of content dissemination, the role of journalists seems to be marginalized, and competition for traffic is becoming increasingly fierce. True journalists are not worried about being replaced by AI because their work involves on-site reporting and interpersonal interaction, which cannot be replaced by AI

Can AI write special reports too? Do we still need journalists?