Luxepack: Generative AI Talk.

At the start of this month, I was in Monaco, giving a talk at Luxe pack, a specialist exhibition that brings in over 10,000 design and business professionals together annually to exhibit and discuss the most sophisticated innovations in luxury packaging. My thanks to the organisers (Formes De Luxe) and the invite they extended to myself after we did a project with them in April. Thank you, Alissa Demorest , Dolores Damade , and Marie-Laetitia Luccioni .

What was it about?

I was really excited and proud to be discussing how we use Generative AI in automotive design alongside EVELIO MATTOS, a much celebrated and revered specialist in packaging design.

Perhaps one of the only times you'll ever have 'vehicular packaging & design' and 'luxury packaging & design' represented in the same presentation.

The Universal Challenge and Promise of AI.

It's no secret that there are some fairly universal challenges in design, manufacturing, and engineering. Wherever you are in the creative industries, or outside of them, the?generative AI opportunity and it's transformative applications?are potentially already disrupting what you and your peers do. They either represent the state of the art, or some pedestrian version of what was possible six months ago (which in AI is years).

Although we're not at liberty to talk about it, earlier in the year, at Vital , we worked on an OEM vehicle design programme using AI to generate concepts, bearing in mind we were limited by the state of the art at the time, that was quite a successful project commercially and conceptually. So we've invested. We've now significantly advanced our capabilities and understanding, from fine tuning models to creating our own WebUIs, to utilising alternative neural networks like Controlnet with SDXL.

What strikes me as crazy though is how 'unevenly distributed' all the developments are! One could assume that generative AI is massively resonating with brands and agencies alike and they are doing so many things with it... , but even if that was the case, the follow up question is... do they actually know what to do with it to maximise its potential? (That's a whole topic unto itself!)

If the hand-raisers in our talk and the conversations that ensued after were anything to go by, it's pure validation, that the majority of professionals at this point in time, sadly don't know what to do to maximise the impact of generative AI. I feel there's a clear line and responsibility, to keep raising awareness 'critically' and 'responsibly', around the use of these technologies and for now that responsibility partly lies with the people leading with an application or experimenting.

The opportunity cocktail

When you start to peel back the layers; regardless of the specifics, the use of generative AI is largely a cocktail comprised of four things:

1. Technologies
2. Processes
3. Contexts
4. Experimentation

New and old technologies that are embedded in existing and new applications with newly crafted processes, in the service of conducting business with humans. Sometimes the payoff is way-off, sometimes it's immediate.

You're either further down the chain or further up, you're either dealing with an idea that's forming, or a well defined idea and you need some solves, whether it's a packaging concept or a part of a car, or a bit of copy!

With progress across all fronts, use cases will evolve and there are some super smart people already doing incredible things. The point of the talk was to say, you are one of them and if you haven't started you can potentially be leading in the space in weeks, not months!

–?

At VITAL, we're jumping in the deep and challenging end, we're pioneering how we can use Generative AI (especially in customising image generation systems) effectively, with an understanding of design constraints. I think that's one of the most golden opportunities and one of the spaces we want to mine.

If there was another message I wanted to put out there, it's one of plurality, not being confined to one system, or one process and not settling for the standard output of any of the tools in their current state. Make it your own, mash up different systems and experiment.

Go from being a user of a tool to a maker of a tool. One of the phrases I keep mentioning, but it's something that was absolutely ingrained in my brain like a pattern on a stone about 20 years ago when I was still at University.

While actively experimenting, of course it's not all going to be a win, but some of it is, and that was the fourth ingredient in the cocktail. Experimentation. you need to FAFO.

FAFO : If you know this guy's meme you know what's behind him on the X axis... you need to F*ck around to Find out. Thanks Waylan Jacobs for that earworm.

From aesthetics to function

We want to analyse and figure out what the current and near future potential of generative AI is more broadly. In car design especially, there's a sense, right now, that rendering an image of a car, is the low hanging fruit. I'm sure it's the same with many other contexts, so to progress things, lets go wider, further back and further forward in the process.

Which are the combinations of AI systems that could be used to improve design and design prototyping capabilities during new product development and beyond, we don't mean the start or at the end, all the way in the messy middle too. We don't just mean all the relatively new Generative AI, that is enjoying a Cambrian explosion, NO, we mean the whole spectrum of AI and ML, the whole palette all the old stuff and the new. That in my mind is the most defensive strategy. Attack the whole of the problem first, then find your niche, as you explore what works and what doesn't work.

In the presentation I wanted to give a quick example of a relatively intermediate technique which will become embedded in every design pre-visualisation tool before long, where you can start with a sketch and quickly visualise alternatives. I asked my friend Matt Watkinson whether he would allow me to share his e-bike concept and a quick experiment I did with his sketch.

The result wasn't just the bikes you see there, it was a staggering thousand images in a short span of time.

"When the issue of skill is resolved, human judgement, curation and experience will reign supreme (for now)" h/t for that paraphrase of a Brian Eno quote from Alexander Kohlhofer

Same with this slide in the presentation:

and this...

all the way through to exploring diverse graphical styles

or a new language to describe things and a kind of digital materiality that is only possible in the latent space.

If you follow my posts or Vital 's as you might be aware, we're already using ML in our workhorse pose estimator, to enable real-time ergonomic assessment, this is an example of stacking different technologies together for a very direct benefit. It allows us to be more evidential. What's really lacking in Generative-AI right now though is explainability and transparency.

The possibilities are growing daily, whether leveraging Generative-AI in concept generation, using AI to experiment with automotive styling and aesthetics or something more advanced such as enhancing vehicle performance with AI optimisation, these are all spheres of influence and spaces we can occupy and mine.

To harness it properly though, I think we need to think more critically and uncover better problems. Rendering a car or a packaging concept, or a unique fragrance bottle idea, may only be a very small part of it.

We need to take stock of the workflows and really consider where we fit in as human designers? In which part of the process does our judgment and taste play a role? What do we outsource to the machine? and how is the outcome better / worse?

This extends to whether we're prompting or whether we try to assert more control too and obviously it extends to being multi-modal.

In a very crude top down way here are some headings aligned with how we work that don't fit the scope of this article but can be used as a jumping off point regardless of your context.

1. Understanding & Shaping the Brief (LLMs)
2. Conduit between different stakeholders / Direct Observation (LLMs)
3. Concept Development (LLMs + Image Generation)
4. Design Investigation: Leading to Insights & Prototypes (ML / AI)
5. Technology Prototype / Demonstrator (AI embedded in product )
6. Evaluation, Content Creation, Transfer of Learning / Extracting Value. (LLMs)

LLM's = Large Language Models, GPT.

AI image synthesis : Stable Diffusion, Dall-E, Midjourney

So what does the future of luxury packaging have to do with automotive?

Maybe nothing! or Maybe everything but here are some compelling reasons for why this presentation was important.

1. We're all seeking better ideas and inspiration. No one wants mediocrity. In 20 years of being in the design business, I've found the biggest inspiration points, aren't in our immediate swim lane, it's in adjacent sectors. That cross pollination really makes for better ideas, makes for better inputs you want a tapestry not a sheet.
2. I’ve found by talking about what we do to other disciplines not only do we benefit from their experience and perspectives and their advances, we get to share our ideas and give back to the global design community, it's a small world, and what we put out there comes back to us, we get exposed to interesting opportunities and possibilities for collaboration too.
3. Also on a personal note, nothing beats being on stage to sharpen your thoughts and present to an audience you don't really know. That slight mix of fear and fortune is a really worthwhile filter. Presenting with Evelio Mattos who is a packaging expert, and a seasoned professional in content creation and building and captivating an audience, was also a learning opportunity for me. I learnt a lot from the process of just putting our slides together and talking it through. Usually its a solitary affair for me, but we worked out how to do the whole thing and delivered it without notes, seamlessly switching to contextualise our experience. From the feedback from organisers, attendees and hand raisers, it looks like it landed well!

If you've read this far, thanks for coming along for the ride, and let me know how I can help you if you're mining this space.