These projects aim to provoke public discussion about the capabilities of state-of-the-art Artificial Intelligence algorithms. For example, we show how AI algorithms can manipulate human emotion (for good or for bad), learn extreme biases from data, or produce creative literature in collaboration with people.
A central idea in machine learning is that the data we use to teach an algorithm can significantly influence its behavior. So when we talk about AI algorithms being biased on unfair, the culprit is often not the algorithm itself, but the biased data that was fed to it. This project, launched on April Fools' Day, highlights the role of data in algorithmic bias by pushing the idea to the extreme. When given an inkblot image (a kind of psychoanalytic test), a Deep Learning-based image captioning algorithm can see very different things, even sick things, if trained on the wrong data set. This project aims to stimulate public awareness and discussion of these issues.
Web site: http://norman-ai.mit.edu/
Selected media: BBC, CNN, NY Post, The Times, La Repubblica, Fortune, Fast Company, Wired (Germany), Rolling Stone, Vice, USA Today, Tech Crunch, The Telegraph, Fast Company
SHELLEY: Human-AI Collaborative Horror Stories (2017)
To mark Halloween 2017, we presented Shelley: the world's first collaborative AI Horror Writer! Shelley is a deep-learning powered AI trained on 140,000 eerie stories from r/nosleep. Like Mary Shelley - her Victorian namesake - Shelley takes a bit of inspiration in the form of a random seed, or a short snippet of text, and starts creating stories. But what Shelley truly enjoys is to work collaboratively with humans. Starting October 25, and leading up to Halloween, she responded to the stories she would start every hour on her Twitter account, and she write the first AI-Human horror short story anthology ever put together! Check out the stories here.
Deep Empathy (2017)
Deep Empathy is a collaboration between the Scalable Cooperation group and the UNICEF Innovation Office to pursue a scalable way to increase empathy. We use deep learning algorithms to learn characteristics of Syrian neighborhoods after the war (for example, Aleppo, Syria), and uses these features to transform images of cities all over the world, simulating how they would look if they suffered disasters like those in Syria. We used these simulated images to investigate whether AI can induce more empathy.
Interactive: Deep Empathy web site
The Nightmare Machine (2016)
To mark Halloween 2017, we present the Nightmare Machine. Since centuries, and across geographies, religions, and cultures, people have tried to innovate ways of scaring each other. Creating a visceral emotion such as fear remains one of the cornerstones of human creativity. This challenge is especially important in an age in which we wonder what the limits of artificial intelligence are -- in this case, can machines learn to scare us? Towards this goal, we present Haunted Faces and Haunted Places: computer-generated scary imagery powered by deep learning algorithms.
Interactive: Nightmare Machine web site