Those who watched the fantastical characters and stories created within the Marvel Universe over the past twelve years might have questioned the seemingly impossible amount of cool technology created by Tony Stark, the alter ego and genius behind Iron Man. Yes, the original suit itself appeared relatively plausible, but eventually this one-man show was creating satellite-driven, autonomously assembling, personalized fighter jets with multimodal user interfaces and complex software. A team of thousands of software and hardware engineers would struggle to deliver such output, thereby making Iron Man’s arsenal less believable than many of Marvel’s interplanetary societies and phantasmal artifacts. One man alone could not reimagine and create a multitude of inventions like that. Unless, of course, he only truly needed to create one invention that would fuel the remainder of designs: Jarvis. Using such an artificial-intelligence-companion-slash-servant-slash-engineer that he nicknamed after his father’s chauffeur, Stark reimagined and tinkered with complex technology.
All fantasy, right? Maybe not anymore.
“We were all living a lie: CAD was and is not really Computer-Aided Design but rather Computer-Aided Drafting,” states Francesco Iorio, the CEO of Augmenta AI. “The challenges and opportunities that engineering at large has faced have been dramatically unmet so far by the tools and computers that were intended to propel the human race. CAD is presently a glorified drawing board where you can hit UNDO, which at one time was considered the holy grail but is now becoming a bottleneck. We can do so much more.”
Now enters a new era of truly computer-aided tools tagged as “Generative Design”. These are unconstrained design tools where the user simply describes the problem they want to solve in the form of data, requirements and preferences. This might be weighted goal statements – in theory TRUE system requirements, which is another whole article — and the artificial intelligence looks for solutions without the compartmentalized, historical mental models of human behavior. Generative Design has found itself at automotive’s doorstep and is attempting to boldly transform the industry’s engineering. “If you keep posing the problems in the same way, you will continually try to optimize around the same constraints” says Massimiliano Moruzzi, the Head of Cognitive Engineering and Business Development at Augmenta AI. “For instance, if you start with the constraint of, ‘I must design a bumper for minimizing impact’, you will always create a bumper.” Moruzzi goes on to describe how this age-old, bolt-on mentality eventually plays-out such that Gen 2.0 Bumper is redesigned and remanufactured around a second constraint such as easier manufacturing or lighter-weight materials. “There is tremendous waste there. If instead you consider the car to be an integral part of the ecosystem of a smart city, that changes the design goals for a component like a bumper. Now those goals and requirements might include a façade that conducts connected signal data or electricity, collecting data for the next desired experience, using and reusing sustainable materials and, yes, protecting the passengers. That crumple zone might also store electricity like a battery, receive signals like an antenna and minimize impact like a bumper.”
Therein, the requirements of the entire value chain — from marketing to manufacturing to service – may be considered upfront in the design via complex algorithms weighing optimizations and compromises. The resulting, holistic widget might be a 3D printed assembly from composite, conductive material in a way that traditional casting or milling would not permit.
And, just like all artificial intelligence models, the long-term power comes from feedback loops. If the series of weighted goals were to create a more comfortable cockpit, improve visibility, increase customer satisfaction and improve serviceability, such tools can marry all of the feedback inputs. As examples, JD Power, Consumer Reports, dealership verbatims, and direct performance and usage data captured by the car itself could adjust design strategies even further. “The requirements from the different users are effectively curated and improve the fidelity over time,” explains Iorio. In essence, such instruments avoid starting with the traditional, poor ‘how’ statements, and instead the goals are a series of ‘what’ statements with the feedback loop translating those into ‘how-better’ designs with the compromises highlighted to each stakeholder. “We’re introducing a new way to even educate the engineer to learn from experience,” says Moruzzi.
The fearful engineer might examine such breakthroughs with nervousness about future job security, e.g. ‘with tools like these, who needs designers.’ “We realized that to change the game entirely, it requires enough embedded intelligence in the tools to compliment and compound the intuition of the engineers and leverage the opportunities,” says Iorio. “This is a conceptual revolution of augmenting or extending people’s capabilities — by means of software and hardware — to engineer this new human capacity beyond their intuition.” Does that mean projects will only need one mastermind like Tony Stark to create a smart car? No.
But maybe instead he’ll help engineer a smart-er car.