The pros and cons of AI in design

For decades we have been hearing promises that robots and artificial intelligence would liberate us all from dull, dirty, dangerous and boring work. Until now, this has not fully materialised, but what has become completely practical is utilising computational assistance and digitised technology to improve the workflows and potentially the outputs of engineering design.

In our design work, computational assistance is becoming of major value in the process and technical analysis aspects of design optioneering and design development. Parametric design for efficient iterations, modelling for embodied carbon, modelling of energy and climate, all of these things are more efficient using digital technology. Because the processes are so much faster than manual calculations by a human, it is possible to embed these value-adding approaches in our standard design without imposing a major time/cost burden on the client or project budget.

But are these technologies truly artificial intelligence?

Strictly speaking, many of the computational approaches we are using are not AI (artificial intelligence) as they do not make independent decisions as such, although there can be some elements of machine learning involved, for example, adjusting parameters based on the result of prior calculations. Parametric design generally though should be classified as automation, rather than AI.

For engineering design, the goal is to leverage the digital tools to undertake as much of the more routine parts of the calculation work so we can be more focused on the creative part, the elements that require informed judgement and evidence-based invention.

We also continually look at how we can use code to develop new tools or improve the functionality of existing ones – the embodied carbon calculator plug-in developed in-house for Revit, for example. Now we are working to introduce elements of AI into computational design, so the AI can for example find the most economical location for columns within a design scheme.

If AI can do that part of the design iteration, it means we can put more human time into the high-value aspects.

But I don’t think we will ever be able to have AI undertake the full process of structural design without expert human oversight and verification. Also, there are limits in terms of AI delivering the type of holistic design we need for Zero Carbon Design approaches. While it can calculate embodied carbon it is not able to understand some of the trade-offs between individual high energy consumption items and overall asset performance and whole-of-asset energy and emissions.

The sheer amount of computing power required – and the energy and emissions footprint of that computing power – makes holistic design via AI almost unattainable for the foreseeable future. Also, there are aspects to design that cannot be translated into parameters and concepts AI can manage as they are not numerical.

AI can manage mechanical properties such as structural qualities, levels of lighting, thermal comfort benchmarking and so forth, but how do we build a mathematical model for beauty? For delight? For the sensation of touch, or the view from a window? Atmosphere and feeling as perceived by humans are something AI cannot understand or factor into a final design

Where this is taking us

Currently some industries are quite advanced in the use of AI, for example, manufacturing and industrial processes and some forms of knowledge economy services such as administration. We are currently also testing the use of AI-enabled robotic construction approaches utilising 3D printing to build a design.

If this type of methodology becomes more ubiquitous, we will find that more human construction and engineering professionals will need to move up to roles that require more design thinking, expert judgement and quality oversight.

This is not something to see as a risk but to regard as a positive opportunity for human knowledge to grow and work to become more creative and interesting. And there will always been a need for human design knowledge and expertise to provide a layer of oversight for AI-enabled design and construction. Just as currently designs are peer-reviewed by another human or a body such a Building Authority, to reduce the risk of errors and non-compliance with codes and standards.

People need to be certain structures are safe, and it is human judgement and knowledge we continue to rely on for that. There is also the question of liability to consider – if AI was to deliver a result that caused injury or damage, who would be liable?

Much like if a crane accident occurs, the crane is not the entity that faces legal penalties, but the humans who had oversight and control of that equipment. So too it must be for AI – we must always have humans hold the ultimate responsibility for the outcome.

For any business in construction and engineering, the challenge right now is to become an AI future-proof company – understand its potential and its limitations and start engaging with it and using it to the most suitable level of potential.

That means thinking ahead and ensuring AI can be applied usefully across every aspect of a company, because the future is coming quickly and progress in the evolution of the technology is accelerating rapidly, rather than taking an orderly, linear path.

We also need to ensure the process of AI adoption is controlled by humans, and that both the beginning and the end of any specific project involving AI are also managed and controlled by humans.

Finally, the pace of adoption needs to be human-centric. The recent Hollywood writers’ strike should be a warning to us that trying to devalue human creativity and effort because there are cost savings in using AI instead is not an ideal way to proceed. The unique talents of the human heart and mind to create can never be replaced, nor should they.