Carl Bass has been making stuff for forty years, from wooden furniture and granite benches to makeshift rafts built from discarded navy pontoons. These days, outside of his day job running design software company Autodesk, he keeps right on making stuff. Like an autonomous electric go-kart powered by transplanted drone hardware (currently under repair). Last week, Gizmag checked into Autodesk's pop-up gallery in Tokyo, where Bass offered his thoughts on the mildly unsettling notion that sometimes a computer's ideas might be better than ours, an emerging concept known as generative design.
"Up until now, my belief is that you've barely ever had a computer aid us in design, it is really computer-documented design, a computer really doesn't do much for you," says Bass.
It seems a slightly odd statement from the CEO of a company who's flagship product trades on that very phrase. For more than 30 years, AutoCAD (Auto-Computer-Aided-Design) has been the software of choice for many architects, engineers and designers working on everything from New York's Freedom Tower to Tesla's electric vehicles.
But just as the terminology might have changed in that time, so too has what exactly these machines are able to do. From a mother robot that builds each robot child better than the last to mind-bending new architectural designs, today's powerful computers can burn through one complex algorithm after another, discarding undesirable outcomes and building on the desired, learning as they go until they arrive at the optimal solution.
"We've mostly treated computing as a precious resource, that it was scarce, that we didn't have enough of it and couldn't afford it," says Bass. "But if you change your mind to say computing is cheap, it is broadly available, it's the cheapest asset we can use to solve any problem. And if you combine that with what's going on in the cloud you have access elastically to as much computing as you want."
In Bass' view, this newfound abundance of processing power is opening new doors when it comes to computer-aided design and is redefining the term altogether. Earlier this year, Autodesk launched software called Autodesk Within that when given certain parameters, like desired weight and maximum stress, for example, spits out optimized designs that meet those requirements. Bass had a rendering of strange-looking engine block on-hand to demonstrate its capabilities.
"You may not think it's beautiful, but I think it's beautiful and that's the subjective thing," he says. "But the way it is not subjective is that it is nearly optimal in design. We know it's nearly optimal because of the technique used to generate it. The way it does this is it tries out an answer and after getting one answer it tries another one. Like an evolutionary technique of trying one after another, it marches toward an answer."
Bass then wheeled out another example of generative design, a 3D printed motorcycle swingarm, that is, the suspension arm that joins the frame to the rear wheel. The swingarm in question was designed for the world's fastest production motorcycle, an electric widow-maker called the Lightning LS-218. Taking into consideration the aerodynamics, stress loads and everything else that goes with crafting the world's fastest two-wheeler, the generative design software ended up with a peculiar, bony looking object that would look more at home in a dinosaur museum than the back of a motorbike.
"It's a very biologically inspired form," says Bass. "If you take the pelvis of a cat of a dog, it looks virtually identical. The idea that it actually mirrors real evolution is incredibly interesting, that after billions of years of experimentation, nature ended up with almost an identical form."
It might seem far-fetched, but one Dutch designer has already enlisted Autodesk's software to produce a 3D printed, metal bridge in Amsterdam. Just like the optimized swingarm for the Lightning LS-218, the bridge was generatively designed, with the team specifying the problem and the software taking care of the rest. The bridge will be built by robotic arms that essentially print the steel structure as they move across it, pointing to a future where both design and fabrication of real-world infrastructure might be a whole lot more accessible.
"Back to the point of how things are made is changing, here's a small group of people who were able to use the most advanced manufacturing to do things," Bass says. "And I think that's really upsetting the apple cart, in terms of small companies and small groups of people, who are empowered in ways to do things that used to require huge amounts of capital."
The thought that these Darwinian machines can crunch through countless possibilities, mutating designs until they produce something resembling the perfect solution, is a fascinating idea. Bass does acknowledge that the technology won't provide a one-size fits all approach, however, that there will be problems where generative design is an appropriate solution and problems that are better solved by human minds.
And as for the fabrication of future designs, both our own and those conjured up by computers? While 3D printing technology is advancing all the time, Bass is of the view that it will only ever complement existing techniques like subtractive manufacturing, rather than completely snuff them out.
"Does 3D printing replace all manufacturing? It's another tool in the toolbox," he says. "There will be times when 3D printing is awesome, there will be times when manufacturing is awesome. What I think the future of making things is, is this combination of having powerful design tools in order to make them, and the powerful fabrication techniques to realize those designs."
Disclosure: Nick traveled to Tokyo as a guest of Autodesk.
It would be cool to see it design a bike frame made of various materials (aluminium, carbon fiber etc.). manufacturing to the design it creates might be pretty difficult if it ends up creating a frame that's hallow in parts but the design it creates would at least be useful information.
It's interesting how much the swingarm resembles a pelvis but at the same time it makes sense.
I would like to see an exhibition of problem solving by this technology in whatever realm that Autodesk believes it would best be suited to outperform mankind, presented in an open contest format that allows entry by anyone who cares to compete with the machine to see if they're able to produce an objectively and quantifiable better solution.
I personally would like to participate in that challenge.
Note: I've been using Autocad for almost 20 years and I have written many Lisp and Visual Basic programs which, similar to what this article describes, will perform many iterations before selecting the ideal output based on the perimeters that I define. However, everything my software considers, was first considered by me, who then coded the programming to automate certain aspects of construction engineering. But those are merely tools which cannot, and will never replace my experience and creativity.
Define your terms Autodesk, let's see who does it better. :)