C.Soddu, E.Colabella
Artificial intelligence and Architectural Design
Introduction
An experimental approach to the design action, building a new tool that increases the design
creativeness with the preview of a universe of possible evolutions.
The design of architecture is based on the ability in thinking the multiplicity of possible incoming
scenarios. The quality of this design is connected to the way used in checking the evolution when it
occurs; and we must do that following the increasing complexity.
The complexity cannot born ex novo, generated by an unique action.
The complexity is the result of a developmental process. The complexity acquires its peculiar
characteristic in front of the temporality of a developing path, using the stratification of multiple
choices in front of the manifold different moments and opportunities. If we want to build the
complexity, we must use a multiplicity of extemporaneous and subjective keys of approach, applied
on different and, maybe, apparently contradictory fields.
Then, the quality of a design is strongly linked to the logic approach that we adopt in the operative
action. The quality is linked to the procedures that we have activated in developing the first
hypothesis. The quality is not originated by the first idea that is, normally, strongly contingent and
axiomatic.
We must use the first idea like a catalyst to start up the process and, after this beginning, we can also
forget it.
The achieved quality/complexity of the professional results in architectural design is measured by the
presence of a concurrent manifoldness of different possible approaches, of possible ways of custom,
of possible performances that the subjectivity of the users discovers like mirror of own uniqueness.
But the quality of a project is also based on the recognisability of the idea of architecture, of
environment, that each architect pursues in the time of his design process, using, as design field, the
space of randomness generated by his discovering of a set of multiple possibilities in front of the
same request. The designer can use these spaces of randomness like evolution field to increase the
complexity and the performances of his project, and so to respond to his subjective conceptual
needs.
In this operative scenario, that identifies in the structure of the approaches the key to control the
quality of the results, the CAAD and simulation tools present, from a part, indubitable advantages for
what involves the quickness of the iterative processes and the preventive control of the results.
From the other side, these tools have increased the possibility of bypassing some steps of this
process, and then of degenerate it. This fact can happen when we use the CAAD tools like the
drafting machines and the typological patterns, whose potentiality of control is performed like
interdiction of some possible field of evolutions and not like the way to amplify the incoming
possibilities of the design.
The tools like CAAD and Rendering have definite a first substantial change in approaching the
design activity. In the beginning it seemed foreseeable that, with the handmade sketches, the
ambiguity necessarily disappear. And, with the ambiguity, one of the essential characters of the
sketch when it is used like a tool of design reasoning, to evaluate possible new evolution paths.
This hypothesis has been demonstrated false. We have considered the ambiguity of the sketch as
semantic manifoldness, as adaptive tool to discover possible design evolutions. This potentiality
remains, and the sketch on paper becomes a more effective tool because it acquires autonomy.
Not only. In this field of approach, now we can also use the ambiguity of the computerised
simulation, of the virtual reality.
The virtual reality has opened an other front that can use the ambiguity like operational tool.
It presents the potentialities to activate unexpected reflections: the manifoldness of the possible
parallel scenarios reported with the quality of the final representation, the multiple and parallel
representation of the same design idea.
Following this possibility, we are now only at the first steps. The actual CAAD follows the layout of
the antecedent tool, the drafting machine and the typological references. This is strongly coercive,
because these tools adopt integrate and optimised pre-designed forms.
But respect to these old tools, the CAAD, as all the software, has an internal availability to the
evolution that allows us to get a total overturn: from the power of interdiction to the adaptivity, from
the closing performed with static models to the ability in supporting the subjectivity and
the differences.
We must use, in the next future, these potentialities of the CAAD, its structure strongly diversified
among interface and executor, the possible metamorphosis of the interface from a set of buttons to
click, into an artificial intelligence shell. This interface must learns from the designer and must
amplifies his potentiality with the generation of a set of possible scenarios in harmony with the
designing idea of the architect.
This is not a far scenario. The experimental tools realised following this approach exist and they are
usable.
The software that we have realised (the last version of BASILICA) operates like artificial intelligence
interface to generate a multiplicity of virtual scenarios, a diversified universe of possible and parallel
reality that are the representation, in real time, of the designing idea. This tool follows and increase
the architectural idea of each designer, and his creativity. Like the ambiguity of old sketches, this
tool supports the fields of possible designing evolutions with the manifoldness of unpredictable
virtual scenarios. These scenarios, generated on real time, are always different, but they are also
always identifiable as one of the possible representation of the same designing idea. The tool works
with the morphogenetical code of the subjective idea concerning the architectural environment that
the designer builds step by step with his natural designing approach.
The incoming CAAD tools, however, are going to be characterised in two types. We must not get
confusion among software that are performed like executors (the CAD like electronic drafting
machines, or the tools for the Rendering, to create electronic maquettes) and the software that are
performed like interface between man and drafting machine.
These interface tools identify the logic procedures in approaching the design process. Today these
tools are integrated with CAD software and support only the management of the design approach
based on typological date base.
In many of these interface tools persists, in this moment, a carelessness and superficiality that affect
the advantages of using information technology in architectural design. To attract customers making
easy the interface, the CAAD producers follows simplification that inhibit the access to potentiality
already present in the tool. To get these potentiality, the ways we can follow are two: first, the
increasing complexity in the user interface (in other sectors this approach has generated monsters,
like super remote-control devices with hundreds of buttons of varied forms and colours); second,
building a dynamic interface trainable from the designer, during his "normal" professional activity.
This new type of interface must be a tool of artificial intelligence able to apprehend from the designer
using the application of relations between input and output, between requests and formal possible
matrixes. So we can leave the use of axiomatic typological references to enter in the field of the
morphological and subjective ones. Leaving also the false optimisation of "objective" data base that
pre-defines a nonexistent inferential relation among functional requests and formalised architectural
events. Defeating the most interesting bringing in of information technology in architectural design:
the ability of management of the complexity.
The interface can be a software that builds, following the logical associations selected by the designer
during his work, a subjective metadesign, an explicit and operative structure of his designing idea.
Supporting each step of the designer, the interface generates an universe of possible and parallel
scenarios, representing a fan of possible evolution paths. The designer, choosing one (or more) of
these scenarios, defines, in progress, the character, the recognisability of his own conceptual
approach. In the phases of learning and, afterward, when the tool supports the creativeness for the
realisation of the project, the interface uses the performer tool, the CAD for the technical
representation and the Rendering shell for the simulation.
In other words, the designer, developing his designing activity, can also design his drafting machine,
builds his operative tool, his subjective metadesign, making explicit (and supported by the computer)
his logical approach, his identity as designer.
Last review 02 December 1995
The Authors
Bibliography
Return to the index of the paper
Return to the first page
Mail to Prof. Celestino Soddu
Mail to Prof. Enrica Colabella