Art generated by AI: an incredible revolution or a threat to the creative world?
Article author :
They emerged very recently and rapidly began to mushroom across the internet. Dall-e, Midjourney, Stable Diffusion and Nightcafé are platforms which generate graphic art by means of artificial intelligence. But whilst some people see in them an artistic and digital revolution, others view them as a threat to artists.
Give it a ‘prompt,’ in other words a textual request entered into an input field intended for this purpose. This request can be simple or alternatively very detailed. You can, for example, stipulate the type of location you wish it to depict, which person you would like it to include, the type of style you wish it to draw its inspiration from or even which artist you would like it to imitate. Then wait anything from a few seconds to a good minute, depending on your request. There you go! You have before you an artwork which covers all your requests and combines them with a highly impressive naturalness and fluidity.
Let us take a concrete example: let’s imagine an image which depicts the following description: ‘a meditative King Kong in the style of Van Gogh’ and ask the Stable Diffusion platform to generate it as a drawing which matches this request.
Not too shabby, is it? You can subsequently repeat the process endlessly by varying the themes or by increasingly adding specifics to the idea of the work you have in mind; the platform will perform the task and provide you with what you ask of it.
But how has such technology been developed and, more importantly, how does it do it?
Generative AI: instructions for use
To succeed in drawing each digital work which is asked of it, generative AI – as it is now being called – uses an immense database as its inspiration. And this source is none other than all of the content available on the web! It’s true; the billions of images – and the captions linked to them – to be found on the vast web are so many examples and items from which it can draw inspiration, and from which it can even retrieve small sections in order to construct its new creation!
Then, it is by means of the network of artificial neurons it consists of and the power of the algorithms integrated within its operating process that the artificial intelligence will begin to search this gigantic database for all the items you have asked for.
Lance Elliot, an internationally acknowledged expert in the field of artificial intelligence, explains the process in Forbes: ‘For example, let’s imagine that you ask what a work representing a frog with a hat on top of a chimney would look like. The artificial intelligence will in that case parse your words and generate an image which most accurately corresponds to the words you have specified. To do so it will refer to one or more images on the web which include a frog. Then it will turn to other images containing a chimney. Finally, it will turn to other images which have a hat within in them. Once it has enough ‘items,’ the AI will identify these components by carrying out calculations and try to create an image which links them together naturally and fluidly.’
A miracle tool for new vocations?
They are known as ‘prompt artists.’ They come from very varied backgrounds and often have absolutely no talent for drawing or painting. But for them generative artificial intelligence is an incredible boon. For some, it enables them to rapidly create graphic works which will serve a range of purposes; concept art, for example, in other words drawings which serve as the inspiration for the creation of more developed works such as scenes from films, video games, or certain posters or the pages of a comic book. But these images may also be used for the press, for example.
Moreover, others have become specialists in the creation of these much-vaunted ‘prompts,’ these textual requests needed by the AI to generate the required images. Because having a tool such as Midjourney or Stable Diffusion is one thing, but it is quite another to have the ideas concerning what you want to generate and above all knowing how to translate them into text.
These ‘prompt specialists’ thus make the most of specific platforms such as PromptBase to sell their textual descriptions. The principle involved is simple. On the website you find images generated by AI. If you like them you have to pay between 2 and 5 Euros to be able to discover the ‘prompt’ behind them and to use it for yourself!
For better or for worse?
As was previously stated, generative artificial intelligence needs very (very) numerous references to be able to respond correctly to the requests made of it and to generate the desired image. And the database it combs through on each new request consists of a multitude of works which have been created by and which in copyright terms belong to artists. Yet, the versions generated by the AI are often very similar to works created by humans. Or else they flagrantly mix existing aspects of certain images, without reproducing them in their totality.
And therein lies this very thorny question: is this or is this not plagiarism? Do these platforms based on this artificial intelligence have the right to take as a starting point works subject to copyright? And, most importantly, bearing in mind that their business model is for the most part fee-based, do they have the right to sell them? As Anthony Masure, head of research at the Haute Ecole d’Art de Genève (HEAD), explains in the Swiss newspaper Le Matin, the answer to these questions is not clear at this point in time: ‘The paradox, with Midjourney, […] is that in paying the licence […] you find yourself the owner of works generated on the basis of something which does not belong to you. There is a real danger. This is something which many legal experts are currently greatly concerned with, and until it is ruled on, in the meantime some artists have already asked that their works be withdrawn from these databases, as the designer Philippe Starck has done.’
Beyond this legal aspect, there is also this very low-priced and frankly unfair competition which artists are all of a sudden faced with. Because, and of this there can be no doubt, such a facility to create quality art images in a few seconds risks creating a massive imbalance. On the Artstation platform – very popular for the distribution of digital works intended for the cinema or for video games – the images generated by AI have begun to proliferate in number, obscuring the works produced by genuine artists. Much to their great displeasure, as is the case for Dan Eder, the senior artist for the popular video game Multi-Versus.
Seeing AI art being featured on the main page of Artstation saddens me. I love playing with MJ as much as anyone else, but putting something that was generated using a prompt alongside artwork that took hundreds of hours and years of experience to make is beyond disrespectful. pic.twitter.com/4p2MLDbADD
— Dan Eder (@3DanEder) December 9, 2022
The complaints have reached such a level that Epic Games, the owner of the Artstation platform, has published a long article to encourage artists to only post on its webpage works which they have genuinely created. Without actually banning users who have decided to continue posting images generated by AI. A neat exercise in performing a delicate balancing act, which illustrates very well the confused muddle around these very powerful tools.
Sources :
A story, projects or an idea to share?
Suggest your content on kingkong.
