we shape our

tools and our

tools shape us

we shape our tools and our tools shape us

System hallucinations, imaginary countries, conversations between machines.

In July 2019, twenty-six artists, scientists and creators worked together for four days on a series of speculative projects using artificial intelligence, trying to reimagine how we work and create using machines.

System hallucinations, imaginary countries, conversations between machines.

In July 2019, twenty-six artists, scientists and creators worked together for four days on a series of speculative projects using artificial intelligence, trying to reimagine how we work and create using machines.

System hallucinations, imaginary countries, conversations between machines.

In July 2019, twenty-six artists, scientists and creators worked together for four days on a series of speculative projects using artificial intelligence, trying to reimagine how we work and create using machines.

System hallucinations, imaginary countries, conversations between machines.

In July 2019, twenty-six artists, scientists and creators worked together for four days on a series of speculative projects using artificial intelligence, trying to reimagine how we work and create using machines.

our tools

We use hundreds of products every day, in the real and digital world.

We learn how to interact with them, which buttons to press, how they work.

From the early days of mass production, we had to adapt to their design.

We want to change that.

What if the objects we use could adapt to our behaviour?

our work

The following projects explore creative applications of machine learning, where traditional interfaces are replaced by systems that could learn and adapt from humans.

F3DS

How can we reinvent our everyday interaction with the computer?

Interfaces that not only use the keyboard as input are becoming more frequent. Can the technology of machine learning offer us a new way of exploring different concepts for a human-computer interface?

Our tool enables you to control a 3D sculpting software using only your nose and a sound.

First, we use the pre-trained poseNet model from the ml5 library to track the nose position.

We train a KNNClassifier model on a sound using the Meyda audio feature extractor.

RobotJS enables us to map nose position to mouse location, and sound labels to mouse clicks.

Pillow talk

When machines meet for coffee and secretly stick their heads together. Spy on their pillow talk.

✎ source code

Posebeat

A human music sequencer that is controlled with different poses.

Argus AI

We trained our computer to help us in situations where confidentiality is needed. For example, when you work on confidential stuff or just things someone else should perhaps not see when they creep up from behind your back.

We used machine learning to train the camera to detect when another person is showing up. When that happens, a small, unobtrusive notification, pops up from the bottom of your screen.

As a way to react inconspicuously, we trained the system also to show a busy looking excel sheet, when you give the subtle command of a little cough.

BooMle it

BooMle

Do a google search with your body movement

We try to capture the movement as an image. The image will be generated by the coordinates of the body keypoints.

DreamPainter

The DreamPainter allows you to draw on a virtual canvas. A virtual airbrush is attached to your hand. The other hand decides when to spray.

The result is transmitted in real-time to RunWay and the SPADE Coco model which returns an eerie visualization.

✎ source code

AI Bubble

What if you would never have to deal with the real world anymore?! AI Bubble provides you all the news you waited for.

Never learn something new anymore, as AI Bubble helps you to keep in your unique world view.

The machine judges your interests and political orientation based on your look. Based on that result you will get your personalized news headlines.

To accomplish this we trained an LSTM network on left-wing, right-wing, fact-based and conspiracy-fake news which we then use to generate headlines. To judge your look we used KNN classification based on pictures of the authors that wrote the news articles.

dIAry

Idea: turning pictures to stories

The project is a reaction to the primacy of visual content in digital media. With image classification methods we try to create textual descriptions of events or stories that could be flavoured with the specific writing style, mood or quotes from literature.

✎ source code

Fictional Constitutions

Generate constitutions for fictional countries by sketching.

Generate your own at fictionalconstitutions.org

GANductor

GANductor is a generator of sound that enables the user to be an orchestra conductor. The Ganductor recognizes the user’s gesture to use them as inputs to manipulate the sound generation.

credits

This event was a four-day machine learning workshop for creatives, directed by Matteo Loglio and Danilo di Cuia, in collaboration with MAInD SUPSI and supported by the Hasler foundation.

If you are interested in participating or hosting the next workshop, let us know.

This event was hosted by MuDA, the Museum of Digital Arts in Zürich, Switzerland.

Thanks to all the people who participated.

A special shout-out to Serena Cangiano and Alessandro Giusti.

Website by oio.studio ↫

Copyright and intellectual property for each project are owned by the respective creators.

TAKEITEASSSS