In automotive design, generative AI is at its earliest stages. Designers and engineers face a unique opportunity to shape AI’s development, rather than passively allowing it to shape their work, writes Shay Moradi
Our attitudes towards emerging technology matter, but I think we are still very much in the early stages of implementing any meaningful introduction of generative AI in automotive design.
At Vital Auto, we have a reputation as people who experiment with and implement emerging technologies in prototyping and design. Our calculated bet on 3D printing paid off, and has since become a staple of our prototype manufacturing work. We were among the first to be asked by OEM clients how generative AI should be approached in early-stage product development, for image synthesis in particular.
In a sense, we de-risked their use of this technology, by explaining its limitations and issues. We accelerated their inhouse capabilities, by showing them a series of workflows that deliver fantastic, controllable results. We couldn’t have done that without experimenting with purpose.
From experience, we know an optimal path forward involves a great deal of curiosity and a balanced, pragmatic approach. We should not blindly adopt any technology such as AI, just because it is novel or computationally powerful.
Instead, we should ask targeted questions about its suitability for specific design tasks. And we should think about where it may end up taking us. Are we outsourcing our actual thinking or complementing it, being inspired by the possibility space?
To explain the business case, we also want to look for some form of evidence that it can provide value beyond human expertise alone. Can it boost creativity or hinder it? Is it actually more efficient? Answering these questions can help shape internal attitudes and policies in a more mature way.
AI should complement the work of designers and engineers in a nuanced way — a synergistic relationship where human creativity and computational capabilities enrich each other
At Vital, we think AI should complement the work of designers and engineers in a nuanced way — a synergistic relationship where human creativity and computational capabilities enrich each other.
Unique complexities
Automotive design has its own unique complexities, as does any branch of design, really — but there’s this sense that automotive is very engineering-led, very exclusive, very specialised. A programme must integrate performance considerations, mechanical and material constraints, aerodynamics, efficiency, safety and more.
As a multi-objective task, this is challenging for any experienced human or team, let alone a generative AI system. And this is why AI, right now, is only affecting small portions of development. Its function is still quite singular and its behaviour still a little ‘black box’-like.
I see one of the biggest challenges in automotive design to be maintaining the emotional resonance and desirability of automotive aesthetics and function. In some product areas, we’ve definitely seen a homogenisation or ‘blandification’ effect, because we’re trying to make things that are guaranteed to function a certain way.
As our use of AI develops, and the technology matures to perhaps encompass and be aware of engineering constraints, I figure it could go one of two ways.
First, it could absolutely risk overoptimising to increase utility — but without their beautiful imperfections and quirks, our designs may lack charm and fail to generate interest.
Second, generative AI’s ability to endlessly remix at scale may actually level the playing field and allow in some outsiders, bringing out the best designs that fit diverse tastes and functions.
Shaping AI like a tree
Although not everyone has the bandwidth or capability to do this, I think we can shape things more when they are just forming, like nurturing a sapling as it grows into a tree. Early adopters also have a commercial first-mover advantage.
Experimenting extensively with AI tools to understand how they work and customise them using existing tool kits for your needs is advised before trying to build something custom. The more hands-on experience you get, the better. Plus, the likelihood is higher of finding something that gets you 90% of the way to where you want to go.
If you really want to commission or get involved in making something new, get intimate with technical details like model architectures, the quality of source training data, ways of fine tuning. In order to shape it, you have to understand how the AI system arrives at its outputs.
Have an attitude of combining outputs from different systems with your own unique human perspective and creativity. Ultimately this should be the purpose: to catalyse and supercharge human design intent. This leads to using AI more imaginatively than prescriptively.
Consider open-sourcing and sharing your augmented workflows, so that other designers can build on your contributions.
The key principles are taking an active role through extensive customisation, retaining human creativity, pursuing AI expertise and advocating for transparency. This should empower you to shape AI to your purposes, rather than be passively shaped by it.
About the author:
Shay Moradi is VP of innovation & experiential technology at Vital Auto, where he’s applying his expertise in the implementation of emerging technologies in digital product design to the mobility sector.
On X (formerly Twitter), he’s @organised at @vital_auto
