Getting help from AI to generate written content, audio graphics and video is now a trend that many digital professionals have taken. So it is normal, now, that the website be full of content generated from a model of artificial intelligence. But we also see them on billboards in our cities or on event posters posted on walls. So we can make AI different uses and to varying degrees. Thus, the generated contents, their re-elaboration and the perception that people, search engines and detectors have of them change.
So it's up to you understand the role humans play in AI content generation and its purpose, to determine whether they work or not.
Content index
The Role of Human Prompts and Rework in AI-Generated Content
Whatever content an AI generates depends on its database, from his capacity and from prompt that you give. So our command, 100% human generated, is essential to go fishing from a billion-dollar database exactly what we want and stimulate actions that enhance the AI capabilities.
Delegating too much to AI sometimes doesn't pay off…
Anyone can go to ChatGPT and write: write me an article about traveling in South America.
The result released without human processing will however be an article mechanical, aseptic and banal. It will be perceived by people as not very interesting and whoever has an eye will understand that it is generated by artificial intelligence. Due to its low quality the search engine they will reserve it for last positions and AI detector various will analyze it as artificial content. Look at an example of a paragraph that came out of that prompt:
There is nothing that tells us precisely how we can explore the indigenous communities of Ecuador to discover ancient rituals and customs passed down for centuries. Nor does it suggest where to start to do so. It is imprecise and an end in itself. Just as the reference to tango is banal and vague.
In this case we are talking about a AI-generated content that doesn't work to interest people and rank on Google. Because everything has been left in the hands of an AI exploited to the minimum of its potential.
…but sometimes it can be enough for us
On the other hand, in a team that develops video games, the marketing manager might come up with an idea for a character. Then he goes up GIVE HER and writes this prompt: Draw, in vintage pixel style, a human with a dog's head wearing a wizard's hat.
It's very clear, isn't it? To write a prompt like this you don't need any special knowledge. The goal of the marketing manager in the team that produces video games is to show his idea quickly and clearly. So let's talk about a AI-generated content that works for this exact purpose.
Then the result may or may not convince those who draw the characters of the video game. It may be used as a basis to draw the same character in other styles or to characterize it better. Or it may be reproduced identically. These choices will later influence the quality of the content in terms of functioning.
AI-generated content works if the human gives good prompts and elaborates
Let's take the initial example from another point of view.
Un copywriter must write an article about traveling to South America without knowing anything about the subject. Instead of spending hours on Google searching and cross-referencing results, gradually commands the AI the information it is interested in. Then he has them given to him sources, check it out and commands the AI to do further verifications to avoid spreading trivial or false information. Once you have all the information, selects and examines them in depth to structure paragraphs and write in his own style, practicing his experience, addressing a specific audience and giving voice to the client who hired him.
In this case they become indistinguishable for people, detectors and search engines the fruits of theAI and those of the hand, of the vein, of knowledge and of reasoning human.
With the result that the generated content appears human to everyone and original… even though it can be considered a hybrid.
Because the human gave the command and reworked, but did so on an artificial basis. Even though he can ask many questions and endlessly request better and more specific answers, relies on the development of an artificial intelligence model. Both in terms of limitations and possibilities.
Then a AI-generated content really works if it starts from good prompts, if it is selected and reworked with humanity. Even regardless of how powerful the AI model is.
But beware of the flattening of standards
AI models evolve rapidly. In some cases, the written, graphic and audiovisual content generated is so well done that prompts are enough to have excellent, disseminated results. But the concept of what an excellent result is is so multifaceted that there is often a risk of flattening the standards.
Even such an image could be considered good as it is:
Generate event poster for the 2024 Ariccia Porchetta Festival.
This result at the first attempt, to someone might seem hand-drawn or drawn on a computer. Especially to those who know very little about AI and graphics. So, maybe, we are talking about a 75-year-old lady. Some strange details such as the petals of the orange flower and the shape of the cutting board, she might not notice. The error in the writing ADI ARICCIA, might seem like a typo and in her eyes those who spread this poster might seem like donkeys. Or she might not notice it and go directly to the meaning, which is clear, even unconsciously.
An example like this, even if it may seem unsuitable to you, would be sufficient for the purpose for many users. So For some, this AI-generated content might work. Whoever distributes the poster earns in time and money spent… and everyone is happy.
This is a potential risk for everyone, including professionals.
The Risk of AIs Training Other AIs
Similarly to the poster, written content generated with advanced AI models following prompts that are not even that structured could also appear human and be enough for people. And this already happens with fake news, with spammed articles around or with muzzle that interact in the Social.
The more this material spreads, the more it fills the databases of the artificial intelligences that generate the content. Therefore, the proportion of data that AI draws from risks no longer being predominantly human in origin. Therefore, it will no longer be inspired by paintings, drawings, articles, videos, voices... but by their previous and numerous artificial elaborations.
AI training AI, in a chaos that becomes an information backwater.
Whether this happens or not will depend on the professionals, who must take care not to make their work artificial. It will also depend on the clients, who must be able to distinguish and appreciate the quality and potential of content that works. And it will also depend on the people, who are all called upon to be able to distinguish content generated with AI… or at least to be able to verify it.
