Inspired on Turing’s proposal about the "computing machinery and
intelligence”, in our post-doctoral investigation, we propose to
consider the question, "Can machines design?". Our investigation is
still very early stages. We have very simplistic models of Design at
this time.
To test our hypothesis, we assumed that the design activity involves
two big actions: [1] the definition of WHAT will be the object; and
[2] the definition of HOW it will be. Our investigation is focused on
the WHAT.
Then, we translated the WHAT action in two computable actions: [1] the
data mining action, when designers seek for opportunities that
integrate needs of people, technology possibilities and business
viability; and, [2] the decision making action, when designers must
integrate the information collected with his experience to propose
objects that addresses the opportunities.
In this sense, we propose an abstract machine that uses two groups of
intelligent agents: [1] the first group with skills to searching the
Internet and build a database of opportunities; and, [2] the second
group of agents that make decisions and propose innovative artifacts.
To test our abstract machine, we needed to build an ml format named as
DTML (Design Thinking Markup Language). It has information about the
needs of people, technology possibilities and business strategies
extracted from real market data.
Then, we made an IOS app that uses the DTML to design. Our first data
base is focused on app market.
Like Turing’s original paper, we propose an imitation game as an
experiment to validate our hypothesis. We use the IOS app developed to
design four apps, each one to a different opportunity. In two of these
projects, a human assisted the agent and in the other two it worked
alone to propose the apps.
The four projects were sent to seven designers to be evaluated. For
each project, the designers given a note between 1 and 5 for a
pre-defined list of criteria (creativity, feasibility, viability and
suitability).
The projects' evaluations did not show big differences, the projects
still was on the same scale (more than 3 and less than 4). The most
interesting data is the suitability criteria, where the projects made
by agents had better notes than others two. We attributed this to the
fact that the agents did not has preferences, its projects was focused
exclusively on the needs of the customer.
In this experiment, the agents passed the test, although computational
simplicity, the results are very impressive. All designers that
participated in the test did not suspect that there was some
Artificial Intelligence in the process.
To the next steps, we expect to test agents to others domains, less
virtual than the app market. We are developing new databases to
fashion and furniture domains.
This is just a initial investigation, but with a interesting
possibilities. If anyone has curiosity, our app is available for free
on the App Store.
https://goo.gl/XjSWd9
--
André Neves, phD
GDRlab - UFPE
Master Artisan of Bits
Artisans were the dominant producers of consumer products prior to the
Industrial Revolution.
Artisans of Bits are the dominant producers of consumer products after
the Industrial Revolution.
http://www.designthinkingcanvas.com.br/
-----------------------------------------------------------------
PhD-Design mailing list <[log in to unmask]>
Discussion of PhD studies and related research in Design
Subscribe or Unsubscribe at https://www.jiscmail.ac.uk/phd-design
-----------------------------------------------------------------
|