3. Our Path Forward
To move towards AI we need to
think differently from the past, in
the future software will be an
integral role in developing solutions.
We need to get better at cross
discipline integration for non-
traditional approaches.
Algorithms will begin to replace
repetitive functions, before the
more advanced decision making.
AI maturity will complement
existing skills, changing the dynamic
between Designers, Engineers and
Programmers. Multi-skilled staff will
be in high in-demand.
Engineer Drafter
Engineer DrafterDesigner
AI
80s
90s
00s
10s
Source(s): Modified from an original concept by
Craig Lamont (2016)
20s
30sEngineerProgrammerAI
Engineer Designer
Engineer Designer Drafter
ProgrammerAI Engineer
4. Parametric Design
Iterative / Codified /
Rules based / Parameterised /
Constraints based
Benefits
Can generate a large number of
options
Simplifies future design
Automation
Cost reduction
Dynamic
5. Generative Design
We’re able to deliver both
parametric and generative
design. Here we’ve applied
it to RET design.
7. How mature are you?
Level 0 Level 1 Level 2 Level 3 Level 4 Level 5
Collaborative
design
Parametric design
for single
functions.
Parametric design
for multiple
disciplines.
Automated design
for a single
discipline.
Automated
Verification for a
single discipline
Fully autonomous
design and
verification for
multiple
disciplines, but a
limited set of
knowledge
domains.
Fully autonomous
design, acting like
a human across a
wide range of
knowledge
domains.
You’re living here You might holiday
here
You could live here
tomorrow
You could live here
in 6 months
You need to be
here in 1~2 years
You need to be
here in 3~5 years
All competitors live
here
Some competitors
live here
Some competitors
live here
Some competitors
are holidaying here
Competitors are
researching a
holiday here
Competitors are
dreaming of living
here
Copyright: Alex Ferguson (2018)
8. What can you do today?
Level 0 Level 1 Level 2 Level 3 Level 4 Level 5
Collaborative
design
Parametric design
for single
functions.
Parametric design
for multiple
disciplines.
Automated design
for a single
discipline.
Automated
Verification for a
single discipline
Fully autonomous
design and
verification for
multiple
disciplines, but a
limited set of
knowledge
domains.
Fully autonomous
design, acting like
a human across a
wide range of
knowledge
domains.
Encourage teams
to look beyond
traditional delivery
approaches
towards Level 1.
Generate
awareness of the
developing toolset
Invest in
parametric design
capabilities.
Strategically hire
individuals with
this expertise.
Leverage existing
expertise to solve
repetitive tasks.
Implement design
management
processes that
require the
extensive use of
parametric design
to deliver
outcomes.
Invest in fast-
tracking research
and development
of automated
design solutions
alongside industry
expertise.
Implement design
management
processes that
require use of
automated design
to deliver
outcomes.
Invest in research
and development
into the use of AI
for design
development
Copyright: Alex Ferguson (2018)
9. What could you do today?
Level 0 Level 1 Level 2 Level 3 Level 4 Level 5
Collaborative
design
Parametric design
for single
functions.
Parametric design
for multiple
disciplines.
Automated design
for a single
discipline.
Automated
Verification for a
single discipline
Fully autonomous
design and
verification for
multiple
disciplines, but a
limited set of
knowledge
domains.
Fully autonomous
design, acting like
a human across a
wide range of
knowledge
domains.
Encourage teams
to look beyond
traditional delivery
approaches
towards Level 1.
Generate
awareness of the
developing toolset
Invest in
parametric design
capabilities.
Strategically hire
individuals with
this expertise.
Leverage existing
expertise to solve
repetitive tasks.
Implement design
management
processes that
require the
extensive use of
parametric design
to deliver
outcomes.
Invest in fast-
tracking research
and development
of automated
design solutions
alongside industry
expertise.
Implement design
management
processes that
require use of
automated design
to deliver
outcomes.
Invest in research
and development
into the use of AI
for design
development
Copyright: Alex Ferguson (2018)
10. Get Started
Problem Definition
•Understanding the
design intent and
desired outcomes
•Review existing
processes for
production and
opportunities for
parameterisation
•Identification of the
problem to be
solved through
parameterisation
•Estimate the cost,
time and resources
to create a solution
in traditional ways
(benchmark)
Define Parameters
•Definition of the
problem in terms of
the key inputs and
variables that are to
be manipulated
•Confirmation of the
intended inputs and
outputs of the
solution
•Confirmation of the
extent of control or
level of automation
to be applied to the
solution
•Confirmation of any
constraints
Define Architecture
•Identify suitable
software platforms
and capabilities
•Develop a concept
of the parametric
solution
•Agree a high-level
definition of the
standards, rules and
frameworks to be
applied
•Plan for and agree
the cost, time,
resources to be
used.
•Confirm parametric
modelling equates
to positive value (i.e.
compared to
benchmarks)
Develop Parametric
Solutions
•Commence
development of the
parametric model
•Review of the model
at each stage for
each discipline to
validate approach
•Independent
verification of the
model outputs for
given inputs
•Document the
solution in code as
required
•Document
information
required to operate
the model
Solve the Problem
•Use the model to
generate outputs
with client input
•Review model
outputs with the
client
•Confirm design
objectives are being
met
•Generate final
design deliverables
•Record details of
the problem
solution
Finalisation
•Archive the project
•Extract and
document IP created
which may be
reused in future
projects
•Publish reusable
modules and code
to knowledge
management
system
•Document lessons
learned and publish
to knowledge
management
system
You can see the strength and speed of development of AI in software like AlphaGo, which, in the span of 6-9 months, went from being unable to beat a reasonably good Go player to beating the previous world champion 4-5, the current world champion, then beating everyone while playing simultaneously! Then there was AlphaZero, which shortly after, crushed AlphaGo 100 to Zero and AlphaGo just learnt by playing itself, and it can play basically any game in which you input the rules.
It is literally, read the rules, play the game and be superhuman
It is wishful thinking to believe an AI is not going to replace us. It is capable of vastly more than almost anyone knows.
Source(s): Elon Musk (2018)
AlphaGo – better than a human in less time to train, better than an algorithm which brute forces to check every solution, 2000 years of history of “this is how the game is best played” to determine strategy and AlphaGo plays moves that humans believed to result in certain defeat, only to win. It has mastered the art.
OpenAI – 5 vs 5 competitive gameplay decision making, long term and short term memory, how might this improve our ability to manage risk, identifying strategies for delivering projects in adversarial environments.
BIM 360 IQ – about implementing AI / machine learning algorithms on top of data collected during construction.
Dynamo / Grasshopper – enabling iterative design – effectively creating search heuristics to find optimal designs that meet desired criteria
AI & Architecture – Sketch a building and have it designed for you, input a space constraint and have an AI generate a suitable layout based on your space usage requirements
Generative Design – Automatically lay out multi-discipline MEP services in a service corridor
Question.. How many lines of code do you think it takes to do the last task?