We are not shy here. We have no limits and we don’t hesitate to wet our jerseys to offer you the crème de la crème of the Internet. After testing the supposed artificial intelligence make us sublime, reinterpreted masterpieces pictorial art using words and queried intellectual property works generated by AI, here we are, head on, to become stars of the digital pencil thanks to artificial intelligence.
And this is the project Broadcast The Rest (part of the platform Stable Spread) which should allow us to become the Frida Kahlo of the 21st century. To do this, nothing very complicated, just sketch what we can – with the touchpad of a laptop computer and the little talent that Heaven has given us – and give a little more information to the algorithm through a textual description, here called “prompt”.
It took me a little time to get used to it and to discover several rules. Already, no excess of zeal. By embarking on my reproductions of great works, I quickly realized that it was better not to add too many elements to my tests at the risk of drowning the algorithm and causing everything to fail.
Forgotten the enigmatic pout of The Mona Lisa, the brambles around Frida Kahlo’s neck and the two figures behind Munch’s crier. Artificial intelligence goes straight to the point and cannot bother with what it quickly considers frills. Same thing for the sentences, the atmosphere is not to Searching for lost time but rather to haikus – unlike Dall-E who has welcomed the lyrical flights of my editor with joy.
The results are sometimes a little disappointing, especially when the artificial intelligence generates a photo rather than an improved drawing. I take it a bit personally. Is an algorithm telling me that I’m drawing so badly that it can’t do anything for me?
The results seem to be generated more from the textual descriptions than from my drawings, which only serve as a distant baseline – to know how many flowers I want to see in a vase for example. Finally, it’s mainly by trying to reproduce the paintings that I have fun, and not so much by discovering the images generated by Diffuse The Rest.
Disappointments and “societal biases”
During my tests, a mention indicated at the bottom of the window caught my attention. She warns Internet users that the model may “produce content that reinforces or exacerbates societal biases as well as realistic faces, pornography or violence”.
This kind of mention accompanies the vast majority of these sites. A necessity linked to the fact that technology and artificial intelligence, completely impregnated with the problems of our human societies, often make evidence of racism.
In early 2020, Kate Crawford, a researcher and professor specializing in artificial intelligence, and Trevor Paglen, artist and researcher, organized a exposure on the subject in Milan, in order to demonstrate how “Ihow humans are represented, interpreted and codified according to training data [de l’IA], and how technological systems cultivate, label and use this data”.
At Diffuse The Rest, every character seems to be necessarily white. For example, it is necessary to indicate “a black man” rather than“a man” in the textual descriptions in order to get away from the idea, imposed by the algorithm, that whiteness is the norm.
Once again, the experience above all reminded us of the importance of deconstructing content related to artificial intelligence more than it took our breath away. In the meantime, I’m going to take drawing lessons, it may be slower, but more effective.
We want to give thanks to the writer of this short article for this awesome material
We asked an AI to sublimate our drawings all cheum
Our social media profiles here as well as other pages related to them here.https://www.ai-magazine.com/related-pages/