Marcelo Coelho

Large Language Objects (LLO) are a series of objects and physical interfaces for language models that extend the capabilities of artificial intelligence into the physical world. By combining models such as ChatGPT, DALL·E, and custom trained discriminators, they reveal a host of new applications for artificial intelligence, from new ways of experiencing music or creating and telling stories, to new forms of playing and communicating.

Developed as part of MIT’s 4.043/4.044 Interaction Intelligence course, these Large Language Objects also provide a glimpse at a new kind of creative process that leverages the capabilities of generative AI from the early stages of concept development to coding, fabrication, and deployment.

Consumer electronics rely heavily on materials that are often resource-intensive, high in carbon emissions, and difficult to recycle. As the demand for embedded computing grows, so does the need for sustainable alternatives that can support electronic functionality while reducing environmental impact.

Geolectric is a design and research project addressing this challenge by advancing the formulation and application of geopolymers—a class of inorganic materials that form covalently bonded, non-crystalline networks. Unlike traditional ceramics, geopolymers do not require high-temperature firing, allowing electronic components to be seamlessly embedded during fabrication. With properties similar to ceramics but a significantly lower carbon footprint, geopolymers present a sustainable alternative for consumer electronics, product design, and architecture. Their low-temperature synthesis, use of locally available raw materials, and electrical properties enable whole new applications in embedded computing, by combining the rich material experience of ceramics with interaction technologies.

Light Craft is a research project to investigate how deep learning and new hardware technologies can be combined to make film lighting easier to learn, create, and communicate. By creating predictive and on-set lighting assistants, we are making film lighting more accessible, flattening pre- and post-production processes, and using artificial intelligence to put humans at the center of the creative process.

NNN is a dataflow and visual programming environment that leverages the power of artificial intelligence for creative applications. Using a node-based graphical user interface, NNN abstracts away some of the complexities of machine learning and allows users to train and create ‘networks of neural networks’, scaffolding simpler models into increasingly complex behaviors and interactions. By creating interfaces that leverage what humans and machines do best, we can create more powerful, generalizable and accessible tools.

Beyond Vision is a lighting performance created for the Rio 2016 Paralympics Opening Ceremony. As a poetic representation of the sense of sight, a mass cast of 400 dancers were equipped with illuminated walking sticks collectively forming a large-scale 2.5-dimensional display.

Sticks were outfitted with a row of 128 programmable, high-intensity LEDs, which were triggered by a radio transmitter to play a series of pre-scripted animations and visual effects while dancers defined the position and orientation of each light.

Six-Forty by Four-Eighty is a large-scale lighting installation that breaks pixels away from the screen, physically immersing viewers into an interactive computing experience.

Individually, pixels can be rearranged in space and change their color in response to touch; however, they only communicate their state to each other by using a person's body as the conduit for information. When grouped together, the pixels create patterns and animations, becoming a medium for composition and expression which evokes a form of digital graffiti.

Sandcastles is a series of photographs that explore material representation across architectural and nano-scales, by traversing pre and post-photography technologies.

In collaboration with visual artist Vik Muniz, we drew a series of castle drawings onto a grain of sand with the use of a focused ion beam and scanning electron microscope. The final images are exhibited as large scale prints.