A Blog by Jonathan Low

 

Dec 6, 2018

Artificial Intelligence Enables the Self-Designing Machine

Generative design optimizes the variables required to produce the most effective outcome. JL


Pam Baker reports in ars technica:

AI-driven generative design software makes it possible for humans and AI to work together to consider every design option and test before choosing one for production. In an AI-driven generative design paradigm, humans input design goals and parameters. Software will control ecosystems of suppliers and producers because the more transactions and participants on the platforms, the greater the optimization. Software explores infinite design permutations that are stronger, lighter and use less material to save money, increase scalability and raise efficiency while enhancing form and function.”
Manufacturing is in the early states of a state of disruption brought on by technologies such as artificial intelligence (AI) and 3D printing. "Additive manufacturing" has already worked itself into companies such as Porsche and Bugatti, and aircraft builder Airbus is experimenting with UAV THOR, a drone made entirely of 3D-printed parts. At the same time, AI is coming into play in a number of ways, in everything from analytics to manufacturing robotics. So the "factory of the future," as envisioned by projects such as the Defense Advanced Research Projects Agency's Adaptive Vehicle Make program, is one in which software drives the manufacturing process and the factory can be rapidly reconfigured to change what it makes.
AI has increasingly played a role in designing products in the form of generative design software. AI-driven generative design software makes it possible for humans and AI to work together to rapidly consider every conceivable design option and to test them all before choosing one for production.“In an AI-driven generative design paradigm, humans input design goals and material parameters,” explains Avi Reichental, the CEO and founder of XponentialWorks (a venture investment, corporate advisory, and product development company specializing in artificial intelligence, 3D printing, robotics, and digital transformation). “The software does the rest—exploring nearly infinite design permutations based on existing design concepts. This includes designs that are stronger, lighter and use less material than would be used otherwise to save money, increase scalability and raise efficiency while enhancing form and function.”
In this increasingly connected manufacturing chain, a product’s form and features don't even need to be finished when it ships. With just a little Internet access, products themselves are beginning to participate in improving their overall design long after they’ve left the factory. They can in effect “phone home” changes to their own design to improve efficiencies or to overcome new or unforeseen obstacles.
Thus, human designers have started seeing their role change: they're increasingly becoming co-designers with a thing for a product's entire lifespan.

Manufacturing outside factory walls

The freedom for Internet of Things (IoT) devices to co-design, and eventually self-design, on the fly came about because no one told AI that its work had to be confined to an industrial setting. Without such a rule, taking manufacturing on the road became an enticing option in design evolution.
Modern generative design works to continuously improve efficiency, sustainability, and resilience at ever faster speeds. It can use any mix of AI/machine learning, edge computing, the cloud, and additive manufacturing (3D-printing) to quickly develop myriad manufacturing design alternatives. These technologies can also be used to help enable things to execute changes to their own design even when they are far outside their creator’s facilities.
“For example, a robot on Mars might detect very loose sand and determine it cannot move about efficiently to complete its mission,” explains Ben Schrauwen, co-founder and CTO of Oqton, an autonomous manufacturing platform. “The robot could learn to suggest different modalities on how to move in that environment and with 3D printing technology and some local robotics, it's very conceivable that the robot could reconfigure itself at a distance to continue its mission unimpeded.”
Interplanetary travel and space missions aside, there is plenty of motivation to enable things to co-design or self-design here on Earth, too.
“The most exciting change is the embedding of sensors within manufactured items, to create a design system that is a self-improving circuit, where the sensors provide feedback to the design to cause it to respond and improve,” said Tod Northman, partner at Tucker Ellis, a law firm with a specialized practice in intellectual property and liability issues concerning autonomous vehicles and other artificially intelligent devices. “Such a system will become a self-improving loop, with better products resulting without human intervention."
Sometimes that self-improving loop will be adding to the thing’s design to upgrade its functionality and features. But sometimes, it will suggest a design change meant to affect a real-time repair. For example, if a robot were to break a leg or a vehicle to blow a tire or break an axle, a truly AI-driven thing could adjust its design to create a different form of mobility and continue on its way.
Simply put, the day is coming when manufacturing becomes autonomous—a time when anyone can manufacture anything on a home or regional 3D printer by selecting base models from a store and adding customizations as desired. Or, soon customers may simply give the AI a general description of a design or a product’s function they want made. The AI will then take care of everything else from material selection and supply chain activation to design optimization and manufacturing processes.
“You won't have to tell it that your design needs to be this material or that material or it needs to be printed or machined. You're only interested in the end performance of the part, the function it needs to do in the world, and that's what you tell the system," says Schrauwen. "The system then figures out whatever material and tools are available locally and how it can deliver on that requirement."
While machines will do this work, don’t expect conformity in their outputs. “You might order the same part from five different small factories and you might get five different parts, but they still have the same function in the world,” Schrauwen added.
When presented with a variety of things that all meet your requirement or desire, you’re bound to have preferences among them. And with preference comes branding so that you can easily find that preferred make again. In that way, an AI-future for manufacturing may not look so different: “AIs could develop their own brands and fan bases,” as Schrauwen puts it.
While some AIs will become more popular than others, none of those will be a competitive threat to human designers. It’s far more likely that humans and AI will become co-designers as opposed to machines taking over these jobs. That’s partly because humans are still more creative than machines, and partly because things are manufactured for human use and preferences. AI is a long way from being able to discern what humans want and like beyond the basics.
“Human designers are not looking for more inspiration on how to shape parts or make things,” explains Schrauwen. “They’re looking to the machines for help in determining whether it’s possible to manufacture the design, and if so, the optimum way to do so. Designers are looking to AI to deal with the complexities in manufacturing thus setting them free to focus on the creative process.”
What they may lack in creativity, today's  machines do excel at computing logistics, costs, and myriad details which in turn magnify human creative expressions. The combination of machine and man is thus more productive than either is alone.
“AI is permitting designers to generate rapid, precise variations in ways that were previously unthinkable,” said Northman.
Today, this symbiotic relationship is more complicated than simple efficiency and expedience. As it gains new experience in design work, AI is beginning to contemplate more than human designers are likely to consider.
“Using AI to generate designs has enabled the system to push through human limitations—producing designs that would be impossible for a human to conceive, let alone produce,” Ellis said.
Again, the human limitations AI is pushing through are not the creative aspects of the design, rather AI finds new material and production possibilities which in turn can affect the creative characteristics of the design. By testing data loops back into the design process, AI speeds up iterative improvements beyond anything experienced in manufacturing earlier. In other words, the human designer initially uses the AI-enhanced generative design as a production assistant, but the two still effectively become co-designers in the end.
Perhaps even better for certain industries, AI can be a human partner in planning, too. When combined with improved 3D printing, adding AI to manufacturing can also address tariffs and regulations.
“With 3D printing, companies can manufacture parts on-site without the costs, tariffs, or time delays associated with shipping. AI-assisted generative designs can be created in one country with the additive manufacturing happening in another,” said Reichental. “More effective additive manufacturing will enable faster and more easily-deployed manufacturing supply chains that circumvent international borders and lower consumer costs dramatically."
In a nutshell, humans and machines become partners rather than competitors because designers don’t want to be engineers, and engineers don’t want to get lost in the production grind. Conversely, self-designing things are focused on solving problems and not aesthetics unless instructed by humans to follow functional improvements with cosmetic work.
Thus, the division of labor is distinctly drawn: Humans create. Machines build. Things change the world.

AI’s effect on tooling design

For machines to build in the most efficient way, AI must also analyze the tools and innovate or invent accordingly. Historically, manufacturing involved cutting, stamping, or pressing shapes from myriad raw materials. Those processes are collectively referred to as subtractive manufacturing. Today, additive manufacturing, aka 3D-printing, adds layer upon layer of specialized “inks” to create a 3D object faster, with no waste, more precision, and fewer steps.
Of course, there are also drawbacks, such as in poorer surface quality. Therefore, many manufacturers prefer to use both methods today, leveraging the strengths of each in hybrid manufacturing processes.
Sai Nelaturi is a scientist and area manager of Computational Automation for Systems Engineering at PARC, a Xerox Company. He specializes in using scientific breakthroughs to create industrial and government technological innovations. He said that designing artifacts that are enabled by advancements such as hybrid additive and subtractive processes requires “searching very high dimensional spaces and evaluating tradeoffs between performance, material properties, manufacturing, and other engineering and design criteria.”
Nelaturi welcomes the emergence of AI-assisted software tools. “The complexity enabled by modern materials and manufacturing processes can truly disrupt the way we design tools,” Nelaturi added. That is the nature of AI tools: they analyze every process, every tool, every design, function, feature, and human whim—and change them all.
Among the initial industry shifts AI has driven, this technology is largely pushing companies towards additive manufacturing.
“Generative design is the killer app for 3D printing/additive manufacturing,” said Richard D’Aveni, a 3D printing and generative design expert at Dartmouth. “For two reasons: First, many companies are slow to adopt additive due to resistance from veteran engineers and designers who don’t want to learn a new manufacturing approach all over again. Second, Generative Design promises a substantial boost in product performance and quality. Once product developers realize its power, they’ll insist on working with it—and will insist that their companies shift to additive to realize the full potential of Generative Design."
AI-based generative design is fueling the adoption and improvement of 3D printing due to that technology’s versatility, efficiency, and lack of waste.
“Additive manufacturing has the potential to decentralize manufacturing because it's basically a single machine with a single supply of material that can create an infinite stream of different parts,” explained Schrauwen. “It really doesn't matter what they are. You don't need to make molds, there's no upfront investment to produce a new type of part. Many machines, spread across the world, can make different parts all day long.”

Democratization, commercialization, personalization

For all the machine-driven disruption being discussed in manufacturing, interestingly, it wasn’t technology that sparked this moment. Humans did; specifically, the Maker Movement.
“The Maker Movement really was at the forefront of all of this. What they did was make it much easier to use these types of machines and processes," Schrauwen said. "Once the tools are there, people who are not an expert can use laser cutters and entry level FDM printers, and suddenly everyone could start producing parts. Once you commoditize the engineering aspect, manufacturing can happen anywhere. That's what the Maker movement showed us.”
The Maker Movement showed that anyone can make something, though ultimately creativity in design remains a rare skill among humans and nonexistent among machines. So even as manufacturing's future gets more tech-driven, the demand for designers is expected to escalate. Contrary to the common concern, all this new tech may thus create jobs within these communities.
“An obvious benefit of AI-assisted design is the shift to digital manufacturing, which will permit DIY and small maker communities to compete with large-scale manufacturers on equal footing,”said Northman. “Working cooperatively, groups will be able to pool resources to spread the initial capital cost of digital manufacturing equipment across many, since all users can get the custom result they seek. Production volume will no longer be a competitive advantage, which will emphasize skill and artistry as the distinguishing factor.”
Not everyone thinks AI and distributed manufacturing will favor the little guys like this, however. While Dartmouth's D’Aveni agrees that AI-based Generative Design will promote additive manufacturing (which will in turn shift manufacturing to a distributed model closer to customers), that’s all he agrees with. Because the AI-enhanced software digitally optimizes production and supply chains, he said, it is likely to encourage the formation of software platforms to coordinate and optimize manufacturing in general. D'Aveni predicts that will lead to highly diversified, giant pan-industrial corporations and networks as opposed to a small business-driven landscape.
“These software platforms will end up controlling broad ecosystems of suppliers and producers because the more transactions and participants on the platforms, the greater the benefits of optimization,” D’Aveni said. “Maker communities can still benefit from AI-driven Generative Design. But they’ll optimize on far fewer variables than pan-industrials will, so they won’t make full use of its power.”
No matter who ultimately benefits the most, it's clear AI is changing manufacturing in all its aspects from design and production to business practices and the tools themselves. And as those changes slowly begin to mainstream in the near-future, one thing is clear right now: your next co-designer is very likely to be a self-designing machine with a few good ideas of its own. Just add art.

0 comments:

Post a Comment