PROJECTS 

AI-BYSSAL LUMINESCENCE

by Alexandre Moura and Alejandra Oros 

Since Prehistoric ages, us humans have had a profound reliance on tools for our survival. This is in fact what make us unique in the animal world. Starting with stone artifacts from about 3.3 million years ago, we shaped our worlds through the use of tools and technological systems.  

The discovery of fire, thanks to human created tools, not only created light and heat, but also lit up human evolution. Something similar happened with sea creatures in the abyss of the sea. Living in an entirely dark environment made them evolve to be able to create light by themselves through bioluminescence.  

Light appears to be something essential for evolution. No matter where we look, we all need it, and we all have evolved around it. Sea creatures developed bio-luminescence. Us humans used stone tools or in other words, the "highest tech" of the prehistoric times.  

This made us wonder what would happen if we recreate the making of light sources with our current highest technologies, the same way prehistoric men did, but in their future (our present).

For this, we decided to feed an AI with more than 900 pictures of pendant lamps and lamp-like organic beings that emit light, in this case jellyfish. The AI then would autogenerate lamp-like objects, which would then serve us as inspiration for new luminary design.  

Having the AI generated images, we chose 3 of them as inspiration and created new light source objects by using Blender as a 3d modeling tool. The result: AI-byssal Luminescence, a collection made by Machine-Human cooperation. 

LEGO AI GENERATION

by Beatrice Gorelli and Maÿlis d'Haultfoeuille

If we feed a machine with the dataset of Legos, will it create a new lego? Yes, plenty of it, with curious and weird shapes. 

The Lego purpose is to inspire and develop children to think creatively, reason systematically and release their potential to shape their own future - experiencing the endless human possibility.

 

According to that we decided the output of our project which is really close to the images that we got, so mostly abstract shapes.

 

Might be possible to have the feeling to recognise some shapes, probably because they are more organic then the usual ones.

 

This can pique the interest of the collectors and at the same time stimulate the creativity of the kids in a curious way. 

CURRENT ANCESTOR

by Lison Christe and Maxime Magnin 

On the assumption that culture could be divided into two categories: the cultural surface (artistic, artisanal, political, social productions, etc.) and what makes it happen (particular feelings, collective memories, social, economic, political contexts, etc.), our contribution would consist in experimenting with the constitution of a cultural meta-surface by way of the Learning Machine. To extract the cultural surface; to constitute, from there, a multiplicity of data sets; and to produce, through the accumulation of objects generated by the Learning Machine, a pattern of what we perceive of culture, totally cut off from all the foundations that build it. It is not a question of producing a simulated culture, but rather of generating an object in its own right, the prism of a cultural shell. 

The methodological questions of the constitution of the data-set are then very important: for example, how not to create a pre-oriented database?  A successful data-set would make it possible to create an object that is not a pre-existing representation. All the accuracy of this new representation depends on the balance of the process. This is why the object of our work focused only on this question: how to create sufficiently balanced processes to create credible cultural material? 

To begin the experiment, we chose the mask as the object. This is a choice, above all a technical one, since there is a fairly rich photographic documentation on the subject. It is also about objects photographed most often on a neutral background and whose identification is rather simple to realize. 

Then, we chose to iterate the processes of generation of these objects:

Process 1:

To constitute the data-set we recovered a sum of "traditional" mask photography on a white background. There was no discrimination of spatial and temporal location. Neither of discrimination as for the use of these masks. Masks with too peculiar shapes (for example masks several meters high) were removed. 

We then generated our images with this data-set.

 

Process 2 :

Among the images generated from our dataset, we chose an image that opened up a wide range of interpretation while keeping a few points of references in relation to the collected images. Once chosen, we made our own three-dimensional interpretation in clay, recalling the craftsmanship in which the ancestral masks made by human hands are set.

 

process 3:

Once we arrived at this handcrafted form, we wanted to translate it into a language suitable for digitization and therefore understandable by the AI. The prototype was passed through photogrammetry in order to create a 3D mesh, allowing us to print it at will with a 3D printer too.
 

by Leyla Baghirli and Simone Di Mauro 

"Masquerade" is a design project based on machine learning whose final goal is as simple as bizarre: reimagine the aesthetic of facial masks, nowadays crucial objects of our daily life, in order to obtain new shapes, designs and color schemes that could give a totally different look to these common accessories. We trained the machine with more than 900 images with white background, starting from the usual blue disposable mask all the way to the gas ones, helmets and so on and the results were surprising. The AI combined all these kinds of masks and tried to find similar patterns between each other, trying to find a red common line.

 

The target of these unusual experiments was the fashion field, where even now, a lot of strange and weird looking masks are used as a fashion accessory, so the challenge was to make an actual 3D model of an AI generated mask and apply the final results on the faces of real catwalk models. In addition, the project focused his attention on color patterns and generated textures to explore deeper the possibilities of Machine Learning.

OPEN IRMA

by Paul Bellon-Serre, Thibaud Goiffon and Chloé Michel 

Are newspapers and magazines' horoscopes, serving us with generalities such as "go with the flow" or "you're feeling creative", really based on the complex system of understanding the natural world and our place in it that is astrology ? Would your daily predictions on Vogue magazine be really that different if it was randomly generated by an artificial intelligence ? 

Open IrmA, named after the fortune teller fictional character Madame Irma, uses machine learning to compile and learn from over 9000 horoscopes scraped from the website https://www.0800-horoscope.com/. The website then generates new unique predictions, ranging from realistic horoscopes to truly bizarre ones and addressing the topics of family, romance, friendship, career and finances.

SEX 2.0

by Manon Waneukem and Tiki Bordin 

Sex 2.0 is a speculativ research with the aim to demystify taboos linked to genitals, sexuals organs, and set off a discussion about intersectionnality and non-binarity. 

Thanks to Machine Learning, we explore the infinite possibilities of a fictional "third sexe".