Research Proposal

A NEW TOOL
A GENERAL PURPOSE DIFFERENTIATING ALGORITHM

Existing tools are essentially combinatorial: the user takes fixed elements and arranges them in space to produce plans. They are also based on Cartesian representation which is fact doesn't correspond to the intuitive human perceptions and use of space.

It is relatively easy to understand what is at stake if we compare these mechanized and ad hoc assembly routines with the other methods of form generation (familiar in biology and embryology). These may broadly be called methods of differentiation.

PROPOSED TOOL

The power of what is proposed can be easily imagined, when one remembers that a fully developed human being, emerges from a sequence of only 50-step wise differentiations. The operations, controlled by the genes, are of course context sensitive, and thus the various operations performed in different zones of the body, at, say, the 37th splitting, are different according to the context in which they occur. It is remarkable that such a finely and subtly complex and differentiated structure can be achieve by only 50 consecutive differentiating operations.

The attached paper on generated structures (first given as a lecture in the Computer science department at Stanford) using cases of natural and planned communities in India shows how mechanized layout procedures are bound to produce expensive and irretrievable errors of adaptation. Only consecutive differentiating operations can give rise to a series of adaptive decisions which are successful in meeting multi-functional requirements. This is the only way to create the necessary complexity required at any given moment of evolution and highly complex adaptive systems over time.

One intellectual challenge is figuring out the consecutive sequence. There are, mathematically speaking, only a tiny percentage of sequences which do not lead to either living with mistakes or backtracking and destroying what exists. The smooth structure preserving unfolding of space so that it is coherent and 'whole' at any point in time (like the embryo-child-adult) is what is at stake. This is the challenge I've been working on for so many years.

The second intellectual challenge is the mathematical modeling of space according to centers and properties (such as boundaries) which is how space in natural living systems is differentiated. This is far removed from the standard Cartesian representation. Leveraging computer computational speed so that differentiation and fine adaptations can be honed at a speed congruent with the speed of growth would give a design tool of great power. Providing step by step instructions which focuses on the content and context of each decision in sequential order plus computer power for modeling and honing, gives the user a tool which is not only innovative but empowering. It has not been done before.

At its most abstract, applicability of such a design approach is wide spread. As with all prototypes, we would start with the small and rather simple. SUN has an obvious interest in work stations and we have extensive experience in the geometrical configuration of successful office spaces. Working out the mathematical modeling of this problem would leave many issues untouched but would give us a prototype which which to move forward.

In the simple problem of a one person office, the user and computer interaction would be along these lines.

The initial form of the room is differentiated again and again by "splitting" just as it occurs in embryology, giving rise to new shapes, zones, and forms which are adapted and non-modular and sometimes nested or overlapping. The user simply identifies priorities in work activities (computer work, phone interviews, meetings, quiet study) and preference for placement of these activities in the room (and the sequence provides careful guidance). Rather than getting the user to combine fixed modular elements, the computer presents shapes arising FROM the form, and which can be finely adjusted by the user.

The operation has built in procedures which will tend to make it structure-preserving, and cause occurrence of the fifteen properties. For example, simplicity (all differentiation that is needed but none that is not), boundaries (that frame and protect activity centers), positive space (splitting so that all zones are alive and useful and none are dead and leftover) are in the operation itself. Thus, to some extent good form is built into the procedure, as a normal part of its operating character.

As each next step is taken in the differentiating process, the computer will enhance the minimal steps taken by the user. Thus the user will need to make no more than slight indications of what differentiation is to come next (according to the prefixed sequence determined by the computer), and the computer will immediately respond with a fluid but probable configuration, so presented that the user may then immediately modify and adapt it to make it just right, almost without thinking. Thus the step-by-step occurrence of well-adapted differentiation is essentially guaranteed by a new type of man-machine cooperation in which the computer does perhaps 75% of the work, enhancing the small indications the user makes of where he wants to go. The result is something that cannot ever be achieved by a human draftsman in a CAD system, or in a pencil and paper process, because in the new system the well-adapted aspect of the emerging form is created by a unique form of cooperation between user and machine.

I believe we can make this so smooth, that the machine's presence will seem almost unnoticeable, and the user will simply have the experience of getting a good result extremely fast, and seemingly without effort.

As each consecutive step of the differentiation is taken, the computer will enhance the minimal actions taken by the user with a supplementary differentiation injecting a new possible configuration into the ongoing emerging design (for example, the size and shape of preliminary space for a workstation with the probable space around it for comfortable moving in and out of chairs). The user may then modify and make the configuration just right by adapting it to walls, other furniture, sizes, and feeling of the space, while the computer itself also works to smooth out the cloudy space where it abuts fixed walls and objects, until the space, even if of irregular shape, is a well-formed center that feels just right to the user.

At the next step, that cloudy workstation area, is then subjected to a further differentiation which uses similar techniques to settle the size and dimensions of a regular rectangular table top that just fits nicely (and perfectly) into the growing configuration.

All in all a complex design task, can be completed in about 20-30 steps of this kind.

Comparable handling of a design task, in existing media, is well-nigh impossible. The task is feasible but arduous and too time-consuming to be practical with paper and pencil or cardboard models; and it is quite impossible with Cartesian representation systems.

The technique can be applied to a wide variety of design tasks - including, for example, design and layout of computer hardware, internal organization, and including, too, the deign of layers in a micro-chip. In a more far fetched future, it is very possible to imagine the same technique, adapted to the design of software, so that organizing principles which rely on a built-in ability to move a growing whole towards "good structure" will clean up and improve code, while it is being written.

In terms of physical production of office furniture, the potential of combining this kind of layout procedure with office furniture or equipment that is not pre-fabricated, fixed and modular but to order and adjustable (completely possible at low cost today ) would be a major move forward.

We believe that the users of this kind of software will have an experience, entirely unlike the experience they have using CAD-based tools, in which they are aware that they are creating practical and harmonious adaptation.