General assessment – What ALGODeQ achieved

Makoto Sei Watanabe (Chairman)

:Creating an atmosphere / As a single leaf

A long time ago, about 2.3 billion years in the past, there was very little oxygen in the Earth’s atmosphere. But gradually photosynthesizing organisms began to produce it, until eventually it became abundant. Other organisms used this new element to produce energy, leading to a dramatic expansion of their sphere of activity. The result is the biosphere that exists today.
It seems natural that today’s atmosphere contains about 20% oxygen. Many organisms are "designed" to flourish in such an environment. It seems so ordinary that we take it for granted. Without it nothing would be possible. But it was not always there. It was created by living organisms.

ALGODeQ is like this process of creating oxygen in a world that contains very little of it. Eventually "algorithmic design" will seem as commonplace to designers as the air we breathe. It will be something that we simply use without being conscious of it, without even naming it.
The atmosphere – the precondition for everything we do.
Algorithmic design – an indispensable method for generating our world in a better way.
Both will be things that are always there, indispensable elements that we rarely think about.

ALGODeQ is like the leaves of a young plant, generating the atmosphere that is algorithmic design.
Eventually this young plant will grow to become a tree, and the tree will become a forest.

1- Objectives of ALGODeQ

ALGODeQ is probably the world’s first international competition for algorithmic design programs themselves. What were its objectives, and what was obtained?

Algorithmic design was described and defined in the competition guidelines.

The guidelines indicated that entries would be expected to demonstrate value in three areas: the problems that the entry wants to solve (its objectives), the method used to solve them, and the results.
On the basis of the above, the guidelines described eight evaluation criteria.

There are no limits to the problems that can be solved. From a personal sense of what makes something beautiful to urban and social issues and further to the future of humanity or the evolution of the universe, any problem is valid. From among those limitless possibilities, entries were expected to select problems and share them with others as important problems to be solved. That is the "value" of a problem.

The value of a method is easy to understand. It is a new technique or a new approach, one never seen before, that solves a previously unsolvable problem or solves it in a more elegant way.
In many cases, the result will be a design. Or it may take the form of text, music, or some other creative work.

Of the eight evaluation criteria, "genetic influence" refers to the possibility of a next-generation set that may emerge from the set of "problem definition / method development / presentation of results" represented by the competition entry.
This refers to the evolutionary process of algorithmic design. This process is the outstanding characteristic of algorithmic design and the one that sets it apart from traditional architectural methodologies.
So then, what about the proposals that were submitted and selected according to this program and these ideals?

2- Jury evaluation method and process

There were190 registrations from 43 countries, and 40 entries of the programs (including ineligible ones) in ALGODeQ. In general, they were on a high level. The requirements for entry were unusually rigorous for an international competition, requiring the presentation of results produced by an original program designed to solve a self-selected set of conditions. That so many entries qualified is in itself a realization of one aim of the competition.
Judging was carried out in two stages.
Some jury members participated in both the first- and second-stage evaluations, including the final award determination, but the juries were not identical. (For the jury members, see the Jury process)
As specified in the guideline rules, jury members belonging to the same academic department or other organization as an entrant did not participate in judging of that entry. To ensure fairness, the resulting differences in the size of the juries were adjusted by applying a weighting coefficient. The first-stage jury considered the entrants' PowerPoint presentations and movies, and selected a total of 16 entries (14 in the QP category, 3 in the QA category). The second-stage jury ran the submitted programs and selected a total of 11 award candidates, from which the final award winners were selected.

The evaluation by running the programs encountered problems when many of the programs did not execute smoothly. Although anticipated, this was stressful for jury members, some of whom had to withdraw. Some programs never ran on any of the jury members' PCs. Even in this case, in accordance with the competition guidelines, other submitted materials were used to evaluate the entry. The evaluation process – including access to entry material, evaluation, voting, and exchange of opinions – was carried out on dedicated jury websites.
Another problem was the difficulty in comparing programs with different aims. Both for entrants and the juries, ALGODeQ was a very demanding competition. As the jury chairman, I would like to thank the judges for making room in their busy academic and work schedules to complete the rigorous evaluation process in a timely manner, and the members of the organizing committee for their unstinting assistance.

The results of the evaluation carried out as described above were generally compatible, with evaluations falling into similar zones for both the first and second stages. This was not the result of consultations between jury members; but was due rather to independent judgments by jury members in accordance with the aims of the competition. The congruence of results is an indicator of the accuracy of the selection process.

The ALGODeQ guidelines envisioned a single Grand Prix, for which there would be a monetary prize, and multiple honorary prizes for outstanding achievement by category (with the proviso that prizes were subject to change.)
The prize format was reviewed at the end of the second-stage judging. Before the competition, it had been anticipated that one entry would be a revolutionary proposal that was indisputably superior. However, it is clear that such proposals will be far from common, and if the general level of the competition is high then it becomes more difficult to discover a proposal that is relatively outstanding.
The level of the proposals in this year’s competition was indeed high, and no one proposal was so outstanding as to be in a different dimension. Therefore, instead of one Grand Prix, it was decided to award multiple highest prizes to outstanding entrants in three separate categories. The prize money will be equally divided between the three winners. Honorary prizes will also be awarded by category. Among them will be a Student Prize, intended to encourage student entrants.
The evaluation scores have been posted separately (see the Jury process).
ALGODeQ was an invitation to a voyage that is by no means easy. We would like to offer our thanks to all of those who accepted the challenge, and express our admiration for the insight and ambition demonstrated by their proposals, including those that were not among the final award winners.

3- Diversity and tendencies / Distribution and categories

There was tremendous variety in the submitted entries.
It would not be possible to organize their disparate sets of problem domains, methods, results, and genetic implications along a single axis. In the same way, it is difficult to compare the value sets and declare that any is superior to the others. But this was foreseen before the start of the competition. Indeed, diversity was one of the aims of ALGODeQ. The fact that such a wide range of entries were submitted was gratifying and in itself a measure of the competition's success.
When preparing the ALGODeQ guidelines, the option of establishing a number of specialized categories was considered. But this was rejected in favor of two broad categories – QP and QA – that depend only upon whether the proposal has been built as architecture. The reason why specialized categories were rejected is that the organizers wanted to leave room for entries that would surpass their expectations, without being constrained by categories or models. For the same reason, although some members of organizing committee thought it would be useful, no examples or images were disclosed in advance of the competition. Aside from a 3D virtual game on the top page of the website, there are no images in either the ALGODeQ website or the guidelines. They contain only text, so as to leave room for the free play of imagination. It was hoped that categories would emerge naturally from the received applications.

As a result, the following nine categories emerged from the competition.
In assigning names to the categories, the following format was used: "Target domain / Type of system or other target _ Action or effect".
The various proposals take on different aspects depending on the choice of viewpoints from which to consider them. They cannot be expressed on a single plane. In an attempt to convey more, we designed a dynamic site that displays multiple key term linked in 3D space (see the Award/Prizes - Linkage).
In addition to the comments below, members of the jury have selected one proposal each and written individual reviews. Reviews by members of the jury are listed below. (see the Award/Prizes - Scroll -Linkage)

3-1. Form / Space _ Translation

One domain where for exciting work was expected is form generation.
In this domain, there were two conversion programs that created new forms from other forms.
The jury was impressed by di047 "Imitation", which took an interesting approach to generating 3D forms from 2D images.
ay139 "Probabilistic Instrument" creates new 3D forms from other 3D forms.
Both of these programs "translate" one form into another. The question is whether and how it is possible to take the meaning and values expressed in a text (=form) in one language and read them the context of another world and another language. What is selected, why, and how is it translated? It would be interesting to aim for a universal translator, or to facilitate creation by translation. Further developments will be welcome.
As described in the "Points for competition entries to explain" section of the guidelines, it is not easy for programs to create beautiful forms.
That is because the meaning of "beautiful" cannot be defined. But precisely because it is difficult, it is a challenge worth taking up. We look forward to the next competition.

3-2. Form / Element _ Fabrication

ln258 "iGeo" is already a well-known program. It has been used to realize architecture through collaboration with other designers in fields such as digital fabrication, structure, and automatic form generation by agents. The jury was impressed by its maturity and wide range of application. It was considered to be a candidate for the highest award.
In the QA category, ld220 "Shift Frame Solver" generates the appropriate logic and program to realize the architect’s intention. It offers brilliant solutions to geometric problems. More will be said about this program in 3-10 below.

3-3. Structure / Physics _ Formation

Programs that discover forms and structures with rational physics.
my185 "RhinoVAULT" is a well-known application with versatile functions that allow designers to incorporate structural calculations into their own design process. More will be said about this application in 4 below.
gp028 "Agent Freeform Gridshell Generation" is a versatile and very well-implemented program for configuring dome-like structures with rational physics.
At some time in the future, functions such as these will probably be provided by CAD systems, or they may become available through seamless interaction between applications. These two entries are significant as forerunners of such an environment.

3-4. Physics / Infrastructure _ Utilization

Programs that point the way toward a general-purpose infrastructure for algorithmic design.
tu142 "Kangaroo" is a widely used and highly reputed plug-in for physics simulation. More will be said about it in 4 below.

3-5. Wiki / Market _ Adaptation

Programs that find solutions by many-to-many selection, branching and matching.
vk222 "lmn architecture" could be called a design program that proceeds according to market principles. More will be said about it in 4 below.
jl244 "Public Mall" registers the parts of existing buildings for reuse as components in new buildings. In Japanese architecture there is a tradition of physically dismantling wooden buildings so that beams and columns can be reused as beams and columns in new buildings. This is facilitated by the fact that parts in wooden architecture were connected by notching and fitting them together without nails, glue, or physical deformation, so that they can be easily disassembled, reassembled, or reworked with little physical deterioration. The same is not true of steel and reinforced concrete architecture, which use different connection methods. This raises doubts about the practicality of this proposal. But the idea is interesting, with considerable merit as a student entry.

3-6. Network / Behavior _ Visualization

Programs that specify relationships between elements and their behavior properties, and visualize the results.
kd195 "Street View" visualizes street networks under specified conditions. In the part of the human brain that is devoted to processing sensory information, the largest area by volume is occupied by visual processing functions. To understand relationships in the world, humans need to have them translated into visual relationships. This program makes a contribution to this field.
zb152 "QUE" shows the behavior of agents in the presence of attractors, repellers, and other items in the environment. Casinos, hotels, and other specific functions can be specified, including reciprocal ones, suggesting further game-like developments.

3-7. Function / Relation _Planning

Programs that specify properties and mutual relationships between functional units, in order to obtain appropriate layouts.
ir165 "3D PRINTING HOUSE" uses the rooms of a house as its basic units, and produces a good arrangement while changing their size and position. It is an ambitious program that enables an integrated workflow from facade design to 3D printing. dr144 "EAST & WEST" in the QA category shares similar aims. It has great potential, although it could not be adequately evaluated because the test program was close to auto-run.
nn196 "c-eyes", also in the QA category, will be described below in 3-10.
Programs of this type can make a direct contribution to planning architectural designs. Further progress in this direction will be very welcome. As described in the guidelines,it is important for the algorithm to evolve by self-evaluating feedback, clarifying how results are superior when compared to results that would be obtained without using the algorithm.

3-8. Function / Relation _ Generation

Programs that generate space and forms by specifying the behavior and mutual relationships of functional elements.
fw212 "Complex Morphologies" is a program to generate tree shapes from the paths taken by particle agents. It has been used to design actual architecture. The graphics are very beautiful. Results after parameter specification are intelligible, but we look forward to further development that would clarify the reasons for and objectives of parameter specifications.

3-9. Utility / Connection _Interaction

Programs that enable real-time interaction between systems with different behavior patterns.
rs230 "Boot The Bot" is a program to facilitate robot action operations. Like the infrastructure programs discussed in 4 below, interactive programs like this, which connect humans and machines, make a fundamental contribution to progress in algorithmic design. Further developments are eagerly anticipated.

3-10. About the QA category

Three entrants were selected in the first-stage QA category, which called for completed works of architecture that employed algorithmic design. All three were on a high level. Moreover, in addition to the program and algorithm, the guidelines listed the overall merit of the architecture as one of the evaluation criteria.
This made it a challenging category.
ld220 "Shift Frame Solver" generates programs for realizing designer intentions. The flow from design to program generation is very intelligible, but the same is not true in the reverse direction. To take best advantage of this advanced technology, circuits to generate design concepts from algorithms and programs would be welcome.
The creator of nn196 "c-eyes" is the designer himself. In that respect, there is a high degree of compatibility between design concepts and the program. The graphics are detailed and beautiful. We think this program might demonstrate its true capabilities when larger architectural projects with a greater variety of conditions. We look forward to next time.

4- Highest Prize – ALGODeQ Award:

4-1. Infrastructure as the grounding matrix for design – Physics / Infrastructure_Utilization

When selecting award winners, the question of how to handle existing plug-in applications was discussed. The guidelines explicitly allowed plug-in entries, as well as previously published programs. This question turns on whether to distinguish between, on the one hand, tools used by others to do something, and on the other hand methods that use tools to do something.
tu142 "Kangaroo", which scored high in the second-stage judging, can be regarded as a program that takes the broad problem of physical simulation as the problem it is intended to solve. To solve that problem, it provides an easy-to-use, effective, and innovative method that is widely used and that has given rise to next-generation descendants. It is fully worthy of commendation by ALGODeQ.
"Kangaroo" is a tool for obtaining solutions to localized problems related to design. The problems to solve are determined by the user. If one takes the view that ALGODeQ should consider new methods and results obtained by using "Kangaroo", rather than the tool itself, then "Kangaroo" might be considered to be outside the scope of the competition.
Pursuing this line of reasoning further, another question arises – whether outstanding applications for creating designs should be evaluated according to the same standards as outstanding designs themselves. Ultimately, it may become possible to compare artificial intelligence applications that are capable of design with beautiful designs themselves.
This line of thought may be interesting, but for the purposes of this competition the jury chose not to pursue it. Instead, "Kangaroo" was evaluated as an example of exceptional infrastructure that contributes to algorithmic design. If infrastructure systems are like highways and grids, and designs are like the vehicles and packets that travel over them, both infrastructure and results were regarded as within the scope of ALGODeQ.
In this sense, "Kangaroo" is a worthy candidate and appropriate winner of the highest ALGODeQ Award.
(By implication, the next ALGODeQ competition will probably welcome applications from Grasshopper and Processing. Further, Linux, Python, C++ and perhaps someday the intelligent CAD of the future...)

4-2. Structure + design – Structure / Physics_Formation

The highest score in the second-stage judging went to my185 "RhinoVAULT", a well-known application with versatile general-purpose functions. The jury expresses its respect by commending it with the highest prize, as a standard tool that adds structural processing to the design process.
Needless to say, structural design is one of the twin faces of architecture, determining its basic form along with formal and spatial design. But structural design has a large element of engineering, making it difficult for design-oriented architects working in fields that give greater weight to artistic elements. Of course, collaborations between the two are frequent for that very reason. But often collaborations proceed more smoothly when architectural designers can address structural design problems. Since 2004, the writer's "KeiRiki" series has explored the integration of structure with formal and spatial design, with the aim of obtaining what I call "aptimized" (optimized) solutions. In the "KeiRiki" series, generated forms have the lowest total material weight for the specified structural members. In "RhinoVAULT", for a given form, slight formal deformations are introduced to enable valid solutions for stress and member specifications. The approach is different, but the greater goals are similar.
We look forward to further advances to enable an even easier-to-use interface, wider applicability and expanded functions.

4-3. wiki design – Wiki / Market_Adaptation

vk222 "lmn architecture" has a dual character, with two divergent aspects.
First, it is a program for discovering optimal designs through the wisdom of the crowd/cloud or market mechanisms. This approach is encouraged by the ALGODeQ guidelines.
It calls for a transition away from decisions by the architect as isolated actor. The act of design is essentially the selection of a line. The line becomes a surface, the surface becomes a volume, and the volume determines the object. Algorithmic design liberates that first act of choosing a line. Instead of choosing one, leave open the possibility of many. Specify a "state" that can satisfy the objective.
Second, it is a program that dissolves the singularity of the actor performing the act. Instead of someone deciding, no one decides.
This double-sided innovation enables a successful transition from "deciding" to "not deciding". What is meant by "design" will change.
However, the "lmn architecture" program has not yet arrived at the stage of producing outstanding designs (or designs that seem to be outstanding). The proposal is like a test sample that makes a prediction of what will occur when the objectives are met. To validate the concept, further exploration is probably needed on a deeper conceptual level, before the programming stage. But it shows remarkable promise, in recognition of which the program is awarded the highest prize.

5- What ALGODeQ achieved / Toward a universal platform

In several significant ways, ALGODeQ is unlike conventional architecture competitions.
First, rather than designs, ALGODeQ calls for mechanisms to produce designs = programs and their algorithms.
At the same time, ALGODeQ is different form computational design. As seen in the definitions and objectives of the guidelines, ALGODeQ does not call for computer technologies like BIM. It calls for new ways to use computers as extensions of the human brain = algorithmic design.
Second, the competition sets no limits to participation. Anyone is free to enter ALGODeQ and the judging is impartial. In architectural competition, entrants with a record of built work tend to be favored when a project is to be realized; when the aim is not realization, entrants tend to be students rather than practicing architects.
It is rare for academics to participate in any competition, because their work is published at academic conferences. And structural designers have their own community. Each group has its own forums to publish and evaluate work in the field.
But ALGODeQ is open to students, researchers, architects, and engineers, and all entries are judged according to the same standards. As a forum for researchers, creators, engineers, and students to publish their results and compare them to others, it occupies a rather unique space, proposing to offer a common platform or field.
ALGODeQ is a debut portal for students to introduce themselves, a forum for well-known researchers to communicate their latest work, and a site for creators at the forefront of innovation to compare their results.
We think that in that role, ALGODeQ has been a success.

6- Toward the next stage / ALGODeQ 2

To build on this success, preparations for ALGODeQ 2 are already underway.
Details will be released at the appropriate time.

The voyage into the unknown continues.
We look forward to further adventures in the company of many intrepid explorers.