Living Encyclopedia: Dream Mode
To unpack what Dataland’s Living Encyclopedia means to me, I pull back the curtain on each mode. This month is Dream Mode.
Last month Dataland announced its first offering for beta testing to the public, the Living Encyclopedia: Large Nature Model. It presents the custom artificial intelligence models that Refik Anadol Studio has developed in an interface that anyone can use to research, create, and, dream their own natural worlds. The reason Dataland, the world’s first museum dedicated to AI arts, is showcasing this is because the work that will be exhibited in the inaugural year of the museum are made with this model. As I wrote last month,1 this initiative is so exciting because it is part software product, part artistic gesture. For those unfamiliar with my journey, this fusion of art and technology is something I think about a lot. I even made a manifesto about it.2 Now, I would like to share the genesis of the project as it prepares for a general release by focusing on one of its modes: Dream Mode.
The video above is a one minute run through of the application’s Dream Mode: a lean-back experience that pans through the image latent space (or a kind of spatialized brain) of the AI model. It focuses on the overall form of the space and then dives into one point to focus, known as a label. Each label is a species of flora, fauna, or fungi found in rainforests around the world. The images and text are completely hallucinated or dreamed by the model. The studio has over several years built up this catalog and it continues to grow; hence the name Living Encyclopedia. It can run for hundreds of hours before highlighting a marker that’s been shown before. While the interface reduces the model’s multi-dimensional space into three dimensions, the same space is also seeded with sounds. As the labels change, so do the soundscapes that support it creating a kind of AI radio.
How is this part software product, part artistic gesture? The prior is straightforward. Once the Living Encyclopedia is available to the general public, you will need to register on Dataland’s website and subscribe (with your credit card) in order to experience it. This experience resembles other AI products on the market like ChatGPT by OpenAI or Claude by Anthropic. For more on how we are building it, you can click the button below to see our internal Product Requirements Document. Once in, however, what exists on the Living Encyclopedia is strikingly different from the aforementioned software products. We arrived at this interface because the team involved are approaching similar challenges to technology companies with an artistic framing. With it, we want to ask questions like how can this tool unpack the good, the bad, the misunderstood, in artificial intelligence? To frame these questions we relied on our collective knowledge of the arts. Personally, works of Doug Aitken3 and Tacita Dean4 came to mind. I wanted to give shape, pacing, and story to the world building process of interacting with AI models.
Unless you got into the beta access last month, you unfortunately cannot try this yet. But, when it does come out, I ask that you try to keep this in mind when using it. And if you do, please let me know your thoughts. In the meantime, I hope you can take my word that art lurks in software,
—Jono
Electric Earth, Doug Aitken. The Geffen Contemporary at MOCA. 2016. https://www.moca.org/exhibition/doug-aitken-electric-earth.
Works on Oedipus, Tacita Dean. Serralves. 2019. https://www.serralves.pt/en/ciclo-serralves/1901-tacita-dean/.