Documentaries are using AI, so where is the reality?

I came across a post earlier that mentioned a documentary about the development of China's high-speed rail, which used AI-generated images.

This documentary is an episode in a series, and the controversial images showcase a segment of the beam-laying process.

The entire scene is heavily AI-influenced and does not reflect the actual situation.

Even a three-year-old knows that you must wear a hard hat before entering such a large construction site, yet there are so many people in the image without hard hats, standing directly under construction equipment.

More knowledgeable netizens pointed out that the type of building structure presented in the video is not a commonly used structure for high-speed rail box girders in China.

If it were really constructed this way, safety accidents would be inevitable.

But the most direct evidence is that the watermark "Doubao AI" is directly labeled at the bottom of the image, clearly not trying to hide it.

Although this AI-generated footage can no longer be found when watching the replay of this episode on the official website, it has been replaced by an old photo.

However, on other video platforms, the original version that was preserved at the time still exists, proving that the blogger's claims are not false.

Such incidents are not isolated.

In 2021, director Morgan Neville filmed the documentary "Roadrunner: A Film About Anthony Bourdain," tracing the life story of the late renowned chef.

Most biographical documentaries use footage and audio recordings left by the subject during their lifetime as clues to connect their life story. This one is no exception.

In one scene, Bourdain's voice is heard reading an email response to a friend, which sounds very impactful.

The problem, however, is how the audio of the person reading the email was obtained. The producer stated that it was synthesized by AI.

The chain reaction this caused was that viewers began to doubt the entire content of the film, as there was no way to determine whether more sounds or images came from AI rather than being real.

After all, Bourdain himself can no longer clarify this.

This situation confuses me.

I can understand using AI for other content, like movies or variety show footage, as it is very convenient.

But isn't this a documentary?

The essence of a documentary is to record real life and capture specific real people and events. Its core is truth.

This process can involve artistic processing, but it should not involve fabrication, let alone be generated out of thin air.

As a result, now documentaries are using AI-generated images and sounds, and the generated content may even completely contradict logic and facts. Where does that leave reality?

We must remember that human memory is unreliable.

Therefore, we rely on various media to solidify, preserve, and pass on what we should remember, and documentaries are one of them.

Many documentary contents are repeatedly cited because audiences assume that the content has been rigorously verified.

Now, indistinguishable AI materials are being integrated into documentaries and spread online and in other films as "real" images, which could easily confuse history and truth.

As the Archive Production Alliance stated in an open letter published in The Hollywood Reporter: "AI-generated materials being disseminated online and in other films as 'real' images could forever confuse historical records." The more serious issue is that it can lead to a distortion of collective memory.

On the other hand, while AI itself has no stance, the people using AI do have their own biases. Therefore, we cannot judge whether the content generated by AI may be mixed with personal or algorithmic biases.

Currently, AI hallucinations are already quite severe; artificial intelligence tends to fabricate things that seem reasonable based on your needs, but those are false and can be mistaken for truth.

Once these mixed-content realities enter the documentary market, which is based on truth, and begin to spread, how can we distinguish between illusion and reality, between what is true and what is false, what is right and what is wrong?

All of this will weaken the credibility of documentaries, making the content they provide increasingly difficult to discern as true or false.

Over time, what is the significance of documentaries?

In fact, it is not just documentaries; more and more fields grounded in reality are being confused by AI products.

Every day, when browsing the internet or scrolling through our phones, we see many outrageous news stories that are fabricated by AI. They are often more creative than humans and can easily concoct sensational news that spreads widely.

Recently, many people used AI to fake videos of homeless people running into homes as a joke, which is a good example.

There are also many fields that require specialized knowledge, which have been infiltrated by AI, filled with traces of fabrication.

AI can also resurrect deceased public figures in bulk, bringing them back into the public eye.

So you can see Queen Elizabeth II wrestling in a ring, Michael Jackson dancing in short videos, and even Martin Luther King Jr. suddenly cursing in a speech.

You can also see Coco Lee, Qiao Renliang, and Gao Yixiang addressing the camera, expressing their longing for family and fans.

This also leads to some people easily using the names of these deceased celebrities to conduct misleading promotions.

For example, "tea master" Zhang Tianfu was recently resurrected by a tea company to advertise their products.

This is quite absurd. The person is no longer alive, yet they can stand up and praise your tea; isn't that just pure deception?

Consumers are deceived, and the families of the deceased suffer greatly.

They are forced to watch a machine-generated mess, using their loved one's face to say things they never said, or even words that are against their true feelings.

This is a complete trampling of life, memory, history, and reality.

There is something even more ridiculous.

Recently, the Dawo Court in Xiaogan, Hubei, heard a case regarding a housing rental contract dispute. The issue arose because the tenant did not pay the rent and utility fees for the last six months, leading the landlord to take them to court.

Upon reviewing the evidence, it was found that the photos of the utility meters provided by the plaintiff to prove the tenant's debt were all generated by AI, and they didn't even have time to remove the watermark stating "Doubao AI generated."

At this point, AI is not only fabricating truths and confusing perceptions but is directly attacking the defenses of social fairness and justice.

When AI-generated faces and false information infiltrate documentaries, news reports, and even courtrooms in highly realistic forms, we will face an era of "Rashomon," where truth and falsehood are indistinguishable Of course, the act of forgery predates the emergence of AI by a long time, but the problem is that AI will greatly lower the threshold for forgery, even significantly liberating the labor and costs of forgers.

To some extent, what AI amplifies is not falsehoods and lies, but the inherent greed and vanity in human nature.

If in the past only one in a thousand or ten thousand people had the ability to forge, now even those who haven't completed elementary school, or who can't recognize a few characters, can use AI to forge.

Moreover, we are now in an era where traffic is king.

Unqualified individuals will deliberately create various fake things to attract attention, using others' attention to exchange for traffic and money.

AI is inherently capable of creating sensational points; the content it generates encompasses current hot topics and creates various conflicts of stance, making it sensational and easily gaining traffic and revenue.

This leads to a vicious cycle.

For example, a principled person who usually writes news articles relies on their own research and effort. Even if they use AI as a tool, they will actively review and filter out false information.

Such news may be real, but it might not have enough sensational points, resulting in insufficient click-through rates, and thus no income or positive feedback.

Over time, this person loses the motivation to write.

But those without principles or quality do not care. They do not need much time, nor do they care about the truth; they only care about whether it is sensational enough.

Allowing such people to gain traffic and earn money will only encourage them to continue operating in this way, ultimately leading to substantial profits.

This is a classic case of bad money driving out good.

The result is that AI-generated fake news will become increasingly rampant, and our deep-seated belief in "seeing is believing" will completely fail.

Over time, the trust that society relies on will diminish, and the connections between people will weaken.

This is a very frightening thing.

We can certainly rely on some well-established rules to reasonably regulate AI-generated content.

For example, any content generated using AI should have complete and prominent labeling, and should not be able to confuse the real with the fake. After all, not all audiences have the ability to discern truth from falsehood.

If it involves important facts, professional personnel and materials must be introduced for review, and one cannot just listen to AI.

Additionally, images and sounds related to individuals must respect individual privacy; no one is born to be AI's material, even if the person is a public figure living in the spotlight.

However, these regulations are not foolproof.

Because as long as there are interests, there will always be people who take risks to do things that are risky but can yield profits.

It may sound pessimistic, but in the future, we might live in a world where it is difficult to distinguish between truth and falsehood. We will need to spend a lot of energy to discern the authenticity of a photo, an event, or a person, and to lonely sift through the already scarce truth.

-END-