AI and the future shape of product design

This photo shows Pierre Baqué, CEO of Neural Concept, used machine learning to design an ultra-aerodynamic bicycle. (Image credit: EPFL)

These days we talk so much about artificial intelligence and its creators that it’s easy to overlook the increasingly prolific role AI itself is playing in product creation and design. Across different industries, the technical and the creative are being drawn closely together to create a range of products that may otherwise never have been conceived.

Blowing past the wind tunnel

Take, for example, the new aerodynamic bicycle presented this month at the International Conference on Machine Learning, which was designed using Neural Concept software. By employing AI in the design phase, a small team from French college IUT Annecy were able to completely bypass the usual methods of testing for aerodynamism – a process that usually requires a great deal of time and computing power.  

Instead of engineers having to conceive and then simulator-test several different iterations of bicycle design, the technology simply took specifications for length and width, and then rapidly swept through an array of different shapes to determine which was most optimal. The CEO of Neural Concept, Pierre Baqué, told Phys.org:

“Our program results in designs that are sometimes 5–20% more aerodynamic than conventional methods. But even more importantly, it can be used in certain situations that conventional methods can’t. The shapes used in training the program can be very different from the standard shapes for a given object. That gives it a great deal of flexibility.”

In short, the AI allows design to happen more quickly, more efficiently, and with impressive results that may help push traditional thinking in new directions. In this case, IUT Annecy wound up with a bicycle that looks a lot more like a tiny racecar than those being pedaled in the Tour de France. Nevertheless, the bike’s creators are so confident in its capabilities that they are hoping to beat the current record for a bicycle traveling across flat ground at this September’s World Human Powered Speed Challenge. Creators of the AI software say that the possibilities go far beyond aerodynamic bicycles, with potential application for drones, windmills, airplanes, and automobiles.

Fashion forward

Perhaps one of the most interesting benefits of using AI as part of the creative process — one that goes beyond the obvious increased efficiencies for time and labor —  is its ability to conceive of familiar things in non-familiar ways. AI has the capacity to “remember” and analyze vast banks of data, and in doing so develops a kind of bird’s eye perspective on what has gone before. This enables AI systems to identify important trends and patterns that previously were not obvious to humans creators who, understandably, have a much more limited grasp on historical data.

A neat illustration of this is the recent collaboration between IBM, Tommy Hilfiger and the Fashion Institute of Technology. Students at FIT were given access to IBM Research’s AI know-how, including computer vision, natural language understanding, and deep learning. They used these tools to analyze not just hundreds of thousands of fashion images from Tommy Hilfiger’s product image bank, but also those from publicly available sources. Michael Ferraro, executive director of FIT’s Infor Design and Tech Lab, explains:

“The machine learning analysis gave us insights about the Tommy Hilfiger colors, silhouettes and prints that we couldn’t begin to consume or understand with the human mind. This enabled the FIT Fashion Design students to take their inspiration from Americana or popular fashion trends and marry that with the ‘DNA’, if you will, of the Tommy Hilfiger brand across those dimensions to create wholly new design concepts.”

Yet, although AI has the ability to inform and augment existing human design methods to this kind of fascinating effect, we’re also quickly learning that it can equally be used to duplicate and outwit.

The real McCoy

By understanding the tiny nuances of product design, AI systems are already being used to create truly convincing counterfeits. According to reports, we aren’t just talking about designer purses. All types of consumer and business enterprise products are increasingly being faked using technology, and the concern is that very soon human experts won’t be able to tell the difference between an original and an AI clone.

So how do we begin to counter this, and ensure that original design maintains its authenticity and integrity?

AI to the rescue, it seems. With its superior perspective and attention to detail, complex software is now being used in a kind of AI cat-and-mouse game. Companies like Cypheme, Redpoints and IBM have already built technology to spot the tell-tale signs of fakery in an ever more sophisticated environments. Writing for Computerworld, reporter Mike Elgan describes how every object, including products, has distinctive optical patterns that can be identified by AI:

“These patterns are found in the texture, color, and patterns in the materials used to make the products. IBM says the technology could be used for anything from diamonds to cash to wine to pharmaceutical drugs.”

From co-designer to counterfeiter, and — ultimately — to chief arbitrator of the latter, AI has an important role in the future of the products we buy and live with, whether they are tech devices or durable consumer goods. For the most part, AI behaves as a brilliant assistant, making dynamic suggestions and moving product development at a pace we’ve never encountered before. Yet we must consider how these intelligent systems might be used to thwart our best intentions and how we can avoid an authenticity crisis in the AI age. It’s a design challenge like none other.