About me

I'm Dylan Holmes, a fourth year PhD student studying artificial intelligence at MIT. I work as part of the Genesis story understanding group building models of how people think, specifically how different people reason about the morality of events in stories.

I'm interested in How to Make (Almost) Anything by way of my interest in programming and software as a kind of social literacy: programming is such a powerful force for good because it is compositional, collaborative, and empowering. I see these same cultural principles applied to the material world in this fabrication class. For this reason, I believe this class also provides a kind of literacy and a similar way of viewing the world. I bring my software-based perspective to this class, and I hope to gain an ability to share this material-based perspective with others, to help build better communities.

1 My project: The artificial skylight

In this class, my aim is to build an artificial skylight that presents the physically-realistic appearance of an infinitely distant sun in a clear summer sky.

To do so involves basically some optical trickery:

To build a bright illuminator to simulate the sun.
To build a concentrator to direct the light in one non-diffuse direction.
To (somehow) build a parabolic reflector so as to create the illusion that the sun is infinitely far away. (Placing the "sun" near the focus and hiding it from the viewer will create a mirror image that appears to be infinitely distant.)
To fabricate a material that will refract white light in the same manner as sunlight reflects in the air, producing a bright sky-blue ambient color and a warm yellow glow around the sun itself. (This requires white nanopowder which is smaller than visible wavelengths of light, e.g. 300nm. In my prelimary tests this first week, I have found a brand of sunscreen containing passable amounts of these nanomaterials.)
To build a containing chassis which securely elevates the skylight apparatus, holds everything in place, hides some of the trickery from view, and presents a windowframe-like facade.

I have some additional ideas for stretch goals, assuming this project works as planned:

To make the viewing window as large as possible. I have done volume calculations for parabolic mirrors, and I think it would be fun to push the limit on how large such a window could be.
To enable the parabolic apparatus to rotate, which given the sky-like composite would instantly pave the way for realistically-colored sunsets.
Alternatively, to build in static multiples, showing different times of day for each.
To simulate other atmospheric effects—wind, clouds, sound, smell, variable light levels as a result of simulated cloud cover, variable light levels as a result of a simulated tree canopy.
To replace the sun with a starfield, e.g. by creating a black sphere with holes surrounding the artificial sun. The trick here is to fabricate sufficiently small stars, within the tolerance of the smoothness of the parabolic reflector, and to avoid light leakage.