Artificial intelligence is now flooding our social media with publications that are more or less pleasing to the eye. The generative AI is responsible for these productions. They can take the form of text, pictures, sounds or videos.
The algorithms create these works by using databases. If you want to order a picture of a cow it is derived from the database.
The generative AI operating mode.
The AI’s working model is what explains the concern of people and companies that produce original online content that it will be plunder or used to fuel AI.
A cascade of lawsuits to defend creators
There is a growing number of lawsuits. In October 2024 the Wall Street Journal or the New York Post from the USA filed a lawsuit in New York Federal Court against Perplexity AI. These three creators sued MidJourney, an image-generation platform, in Californian court (New Window), in 2023. They claimed that the platforms allow for works to be created at will, “in the style” these artists thanks to prompts, including their names. These trials continue to multiply across the globe.
We must first take into consideration all of the existing works, which have been recorded without the creators’ knowledge in the documentaries of AIs that generate art. It’s difficult to determine that a part of a fresco created by algorithms is your own work.
Finally, it is possible to make legal and technological provisions that prevent online publications from being automatically picked up by automata used for mass collection. It is not known how effective these measures will be over time.
Technologies to help artists?
Researchers at the University of Chicago promise this with their software, Glaze (New Window)They suggest that artists protect original creations by subjecting them an algorithm which will imperceptibly alter their work. They can do this by changing the style. This is not visible to human eyes.
It is a form of model poisoning. Only the algorithm that generates AI images, which has been modified specifically for this image, will be attracted by it.
We can’t always assume this technique is going to work. We are seeing a constant adaptation of the weapon and shield, both in terms of their defense capabilities and attack abilities. This form of protection is therefore easily circumvented over time.
Artificial intelligence malnourished by other AI
Studies have shown that using artificial intelligence data to improve an AI model will lead to poor results.
The rapid degradation of the images produced by an AI that is fed only artificially created productions, as demonstrated by a study published in July 2024 by the academics at Cambridge University (Great Britain) and presented by the journal Nature shows the impact on the image.
Rice University, USA, has also made similar findings . Their analyses show the negative effects that results based solely on algorithms can have.
All the arts are exposed to the AI wave
Generated AIs can now elaborate text – including poetry – and images. They also reconstruct voices. The CNRS was able to combine the General de Gaulle’s voice, which has not been recorded, with animal sounds, natural phenomena such as water runoff or thunder, and create a new version of his June 18 Appeal .
Websites like Suno.Ai (New Window)– offer contextual elements to create a song of two minutes with lyrics and music that are personalized.
You can change the voice of the singer and style, (rap, Folk, Metal …),) as you please. Subscriber formulas let you download several hundred songs every day.
Get a discount on your purchase of music.
A reconsideration of the artistic approach?
Technology will not stop people from creating. The productive ability of AIs that generate content obliges both artists and the general public to be interested in how this content is created. Appreciate works in full knowledge.
Customized workflows using Make.com automation