Edited on 4 May 2012 for minor readability improvements.
Readers may be interested in the legacy of FilmsFolded, so I have written here some brief notes about where the project started and the various episodes leading to this point.
The short version: I am a student of how we represent complex and soft constructions of concepts. I have always been interested in both how these present logically (so it can be modeled and understood by a machine) and how they can be expressed by visual grammars. My interest in concept representations-via-form include visual shapes, origami, string figures, a few other things using laws of symmetry, but most of all I'm attracted to cinema.
So I initiated a personal study, one which grew here to impressive obsession. It was swept up into an ambitious internet startup that failed, Framethrower. I took some time for recovery and have restarted with the personal study you see here.
Some MIT History
Strictly speaking, this project has its origins in speculations more than 40 years ago. MIT had a loose collection of projects named the ‘Architecture Machine,’ which in one form evolved into the current Media Lab. A related project, by a somewhat circuitous route led to this work. Generally, what bound the projects was the challenge of modeling the architectural design process. Certain facets of such design are simply engineering processes. Other facets, concerning style, are computationally tenable as well, with difficulty addressed even then by what has since become called ‘artificial intelligence.’ But one corner of the process has to do with how a space feels, just feels by the people inside it.
This is quite a different matter than how the outside shapes of the building register as a sculptural object; it instead concerns how the environment — in terms of perceived form — influence individuals, groups and activities therein at a psychological level. We knew that such a phenomenon existed, that in some crude way some effects could be observed and that some talented designers could intuit key principles somehow. The question was whether such a thing could be captured with the logical tools then available.
The answer was clearly no. This is equivalent to asking a machine to fall in love, be perhaps disappointed in passion and manipulate subtle theatrics to make art about the experience. In some ways, the goals of this project grew to be as large as the challenges of all AI and cognitive science. But the promising pathway for exploration seemed to by constraining the vocabulary to form, and the focus on fixing or extending machine logic. The project left MIT and was picked up by an intelligence lab, still unacknowledged.
Adding in Film
In my life, that context got mixed with my discovery of film as consciously designed cognitive manipulation. A turning point was a viewing of Woman in the Dunes in room 10-250 of MIT in 1966. That film contained obvious, almost condescending metaphor in the story. But it also had something else: a visual power, a cinematic grammar of form and manipulation of perspective that was at once less obvious and more powerful. My film viewing since has been informed by insights from the project in that lab, using what were then radical ideas in geometric logic — a sort of visual physics of what we have around us, from which we can build new abstractions.
As you will see, those abstractions simplify the world into a few elements. Where ordinary science reduces the world to objects with properties, forces and influences (with associated information exchange), we abstract the world to urges and functions of generative creation. The result are systems in which objects are embedded in and possibly result from their behavior rather than the other way around.
In other words, we have agents (which can be verb-like in addition to noun-like), causal dynamics that generate and or are caused by those objects, and environments (the systems and situations) in which they work. Those environments are characterized by form in such a way that relates to the shapes of architectural spaces, as well as cinematic narratives. But that form also has some logical properties in an extended mathematics of logic.
To put that in movie terms: the narratives in film have form, some of which is abstract, some real. Collectively, those narratives interact by certain dynamics (many of which we can understand) so that the elements of the narrative (characters, story, cinematic devices) generate the narrative's future from its past using certain causal logic (which with some cleverness we can also understand).
The bottom line: the FilmsFolded work is completely in terms of movies themselves, with their components described in ways a filmmaker or intelligent viewer would understand. Those components are tied together in ways inspired by a rather clever, different way of parsing the world: one of form, its structure (that is, its folds) and the causal dynamics which govern it.
Some DARPA History
The NSA-related work segued to a collection of projects sponsored by DARPA, in which I also played a role.
The target application of the DARPA work was to help distributed organizations be more effective in designing and building complex manufactured objects, like advanced aircraft and computers. The elements were new: product and process features from the nitty gritty of engineering and management science. But the dynamics were the same as explored in the film and intelligence studies: make every company, process, technology and management practice an agent (in terms of neutral ‘value features’) and ask those agents to build good narratives.
In business terms, this meant creating a complex product or service, while at the same time creating the organization to deliver it, from a large pool of small companies, including competitors. These could be companies that are not known to each other, have no central coordinating authority (like a prime contractor) and will be expected to do things they never have before. We called this the agile virtual enterprise.
This magic is accomplished by forming each of the features into an agent that would opportunistically collaborate with others to build (and continuously modify) the virtual enterprise. (The usual way — which would never work well — was and still is to make each small company an agent, scrapping in the marketplace.)
The narrative in this project would be the product: its design and manufacturing plan plus all its constituents, not too far from the common notion of a story. The (substantial) funding was important for two reasons.
One was that it allowed us to build and validate some systems along the development path, refining the notion of collaborative narrative dynamics.
Another is that it sent us on a quest to find existing examples of virtual enterprises, which started with the whaling industry and ended quite unexpectedly in Hollywood. We were provided with profound access to film organizations and we formed insights not available anywhere about how the actual production of films works, and what the enablers are. Some were quite unintuitive, others subtle. Still others led us to the outer edges of the creative side and the subconscious communicative channels among the artists.
Many of the some component structures used to weave collaborative creators together in the process of filming are the very same narrative structures used to engage and carry the audience. Some of this is obvious, and captured in the concept of genre. But it goes much deeper.
This is what triggered the study proper, resulting in what you are now reading.
IMDB and FilmsFolded
I started the FilmsFolded project in earnest, and by that name after the DARPA efforts ended. The original goal was purely personal: I would watch a film per day and extract some of the causal dynamics I observed.
I believe that we have the new logical tools to understand and model the dynamics I found; I write about that elsewhere. But each dynamic has to be discovered from nature. An analogy is the periodic table of elements or the standard model of elementary particles. The order that relates the elements is known, but early in the game the qualities of each element have to be discovered or at least validated in the wild. (The analogy is imprecise because with film we are concerned with experiences, not objects.)
The initial process was that I would watch a film and enter the type of folding observed in a database, adding descriptive fields as necessary. I would then try to say something insightful about the film in 500-900 words, possibly without regard to foldingdetails. These would get posted to IMDb in the ‘user comments’ section (now sadly renamed by IMDb to ‘user reviews’). The comments were for my own reference, to help me remember the film when I went back to write the essays on general principles; this is essential for many of the films.
They also served as a sort of test; if the most interesting thing about the film (to me) was not about some discovered ‘fold,’ then the question was why not? Was my approach incomplete, my notions too sterile? Over time, I found that I had a readership for these IMDb comments, and at one point I was getting 15 messages a day in feedback.
I was also getting a lot of hate mail. This was of two kinds, the most numerous were from Christian fundamentalists who took exception to some perceived ‘liberal’ tone of a comment. Early in the IMDb process, it was very easy to have a comment deleted. All you had to do was to complain and the process would be automatic; you'd never know when a comment disappeared. An apparently concerted campaign killed a few hundred of my earlier — and I think best — comments.
The other sort of hate mail was from apparent teenagers remarking on my arrogance, often wrapped in some sort of gay epithet.
One reader created an RSS feed, adding a significant number of readers because it fed a social network conduit. Another created a website, TedgDB.com, that collected all my IMDb comments as they were posted, entering them into a handily searchable database. I was able to talk him into moving it to FilmsFolded.com (since reinvented) and adding the initial outline of these essays. The community expanded.
My plan at that time was to use the comments as a placeholder for a database of elements from each film — perhaps gathered from the community, and to link them to essays on key insights. That plan was waylaid by the Framethrower project (described below), and is again the basis of what you have here: essays on basic principles linked to explicit examples, advised by an expert community.
The narrative dynamics are sometimes sophisticated, but generally understandable. There are many kinds, and a comprehensive study of the dynamics is supported elsewhere. The FilmsFolded project focuses on a subset of the dynamics called ‘folding.’ As a result, in this collection of essays the term ‘folding’ is used as a generic word for the dynamics.
This project, the part of the website you are currently visiting, is a revival of the original vision before Framethrower investor worries and the rush of coding.
That original vision allows me to simply write some essays about what I have seen with the expectation that over time the insights will grow, be coherent and be linked to useful examples. There are some clever technologies planned for this site, and some already built in. The progress will be slow, and I will doing it outside of income-producing activities.
But we will get there, beyond the Framethrower.com vision for sure.
The strategy is to build the essays here, with examples from individual films. All essays will be temporary as new material is added. Community participation is welcome, via the comment area at the bottom of every page in this area — or by direct email.
The blog at blog.tedgoranson.com/filmsfolded will report significant progress when it occurs. An RSS feed from the blog can advise you when I post progress.
A companion site at tedgoranson.filmsfolded.com collects all the original/rewritten comments, plus remarks on folding. The essays can be viewed from that site as well.
redframer.com is where we will be building and demonstrating the new site, successor to Flamethrower.com
Much Older Sites
Those who have been following me for decades will know of a previous Sirius-Beta site that referenced two projects that could fairly be considered historically related. For completeness sake, I'll note them here. They are predecessors of the underlying frameworks that made it possible for us to consider a fast track for the Framethrower.com project, so bear more on the formal methods and code than on the narrative dynamics.
An archived version of that legacy site is available (as of 7 May 2012) here.
As it happens, I could not get permissions to disclose that work, so was unable to finish that site.
There were two main projects, ALF and ALICE. They have since evolved to the system we are working on now.
ALF was an ambitious, distributed agent system that would organize distributed, dynamic, heterogeneous knowledge (and half-assed data) into structured concepts. It was the ’serverside’ back end of the system. A specific version of ALF was to support an ambitious predecessor to Framethrower.com called Fusecap.com.
ALF did not yet have the narrative structure we now rely on, and it was lacking some of the coding techniques now available. But it worked and did things that still are beyond the reach of any commercial system.
The name came from a sponsor who was interested in the path Alfred Nobel was able to take, a sort of weapons merchant to philanthropy model that would have balanced a ledger.
This was the client-side complement which should stand on its own. It was named much earlier with a reflexive acronym for Alice (as in Alice Liddell) Lattice Integrated Conceptual Environment. It also was instanced in a stateless Ada as the Ada LICE. It has evolved into the current client-side environment, the one that will host the results of our kutachi work.