Mentoring the Machines

Daniel O’Connor | TCT Magazine

As I was ushered into a private room in the Ritz Carlton in Berlin, resplendent with fruit platters, pastries, French press coffee, and premium bottled waters, the interview that I’d arranged with Autodesk’s Chief Technology Officer began to feel more like I was interviewing Jeff Bridges than Jeff Kowalski. Those initial intimidations were quickly washed away when I was introduced to Jeff, an unassuming man dressed in sneakers and comfortable clothing.

Jeff speaks of complex subjects with a clarity that makes you believe he would have been the greatest teacher you never had. Instantly as I left the interview with the man so passionately behind Autodesk’s push towards generative design, I felt more intelligent, such is Jeff’s ability to explain concepts like how slime mould networks maintain contact when moving from food source to food source and how that can be applied to manufacturing today.

This is what makes Jeff such an engaging speaker; hours after our interview the man who joined Autodesk in 1993 and assumed the role of CTO in 2006 is due on stage presenting his keynote on “The Shape of Things to Come” at re:publica 2016. Not that you’d know, Jeff is man comfortable with his subject, even in the face of questions like “Can you talk about the negative side of technology in the context of Adorno and Horkheimer’s Dialectic of Enlightenment?” (the first question from the audience he was asked post-presentation in front of an auditorium of over 1,000 people), Jeff fielded this question, clearly designed to trip him up, with a smile and a passionate response:

“Because on Star Trek Captain Kirk was able to call the Enterprise from his hand, before we had any such technology, when that technology started to materialise we already knew what to do with it and we welcomed it. I think that the stories that we tell have a power to manifest themselves in the careers that we pursue and the things we put on the planet.”

Divide and Conquer

Before that grilling and before I sat down with Jeff I had the chance to talk to Bastian Schaefer, Innovation Manager at Airbus. Bastian’s work on designing an entirely new 3D printed partition wall is pride of place on the Autodesk booth at re:publica. After debuting at Autodesk University the design has been featured everywhere from Wired Magazine to Fox News, such is the vision behind a mundane part of the aeroplane that divides one us passengers from Cabin quarters.

The “Bionic Wall” is 45% lighter and just as strong if not stronger than the current iteration of the wall flying in every single A320 in the world. By applying this design throughout the cabin on the backlog of A320s, Airbus estimates a saving of up to 465,000 metric tons of C02 emissions per year, the equivalent of taking about 96,000 passenger cars off the road for one year.

The partition structure is entirely 3D printed but due to the fact that it represents one of the biggest individual components to an aircraft it was printed in 162 individual parts across eight builds using a mixture of the EOS M400 machinery and Concept Laser’s M2 platform. 122 of the parts were printed in Airbus Group’s second-generation aluminium magnesium-scandium alloy, Scallmalloy, with the further 40 being printed in titanium.

Although the assembling of the partition does away with one of 3D printing’s key benefits Bastian assured me that they are working with OEMs in order to make a machine that would be capable of building this in one shot. Bastian says there is also a company, Arevo Labs, who claim that they can make the partition in one shot using an extrusion technology and composite materials.


Next Gen Design

3D printing is only one step of the process; the key to the partition’s ability to be both lighter and stronger than the traditional counterpart is in the design. David Benjamin, the Founder and Principal of The Living design studio in New York, which was acquired by Autodesk in 2014, met with Bastian Schaefer at Autodesk University’s exclusive gathering, the Design Symposium. David’s research into the aforementioned slime mould networks and bionic design intrigued Bastian, who really wants to revolutionise the way we traverse the globe.

“They’re both passionate about changing the bulk head design that is used for aircraft, submarines and pressure vessels, which has been around for centuries,” explains Jeff Kowalski in the plush hotel meeting room. “Moving past that design concept and mental constraint by re-envisioning the aeroplane of the future. At the same time at Autodesk we had a desire to reapply what we found to be an overabundance of computing and experiment with the idea of generative design, using the computer as the tool for exploration as opposed to just documentation and analysis, those two seemed like a natural fit.”

Another natural fit is generative design and additive manufacturing; I asked Jeff if he felt that the two were creating a perfect storm towards Industry 4.0. “We’re trying to get to forms and design objects that really exploit the power of additive manufacturing. Every area of the volume can be addressable, objects don’t just have to be the solid shapes that you could have done on the mill. The types of things that generative design has been creating seem to be pretty well exploitative of that.

“Once we start getting into bespoke materials where we can have material gradients as opposed to isotropic or monolithic materials we’re really going to be able to exploit the generative design aspect,” Jeff begins to coil with excitement. “Take the mechanical hinge of my glasses, that is just one way of achieving that type of functionality I could equally have had software come up with the need for flexibility realised in a flexible material that maybe was only flexible in the horizontal direction and not in the vertical direction and be able to actually materialise that.”

It is clear from Jeff’s forward thinking why he makes such a good CTO, why he’s entrusted with Autodesk’s long-term technology vision. He is equally adept to seeing current limitations as he is to seeing the future solution. Take the Airbus design for instance; it has been mooted before that as much as 80% of an aircraft could be 3D printed, if that were the case could we build it with generative design?

“Today’s generative design software solves structural problems very well,” explains the CTO. “Issues of structure, strength vs. weight vs. cost vs. manufacturability are in the realm of trivial for generative design to now solve, computational fluid dynamics, the aerodynamics of the plane aren’t something I would want to be tackling using the software today.”

Build of partition parts in Scallmalloy from EOS M400-2

Though according to Jeff the software’s ability to compute aerodynamics is just a matter of time, one issue that is causing a bottle neck for the adoption of new technologies particularly in the realms of aerospace is that of qualification. Only perhaps in the medical world, particularly in the US where FDA approval is notoriously difficult to attain, is it more difficult to get a part to market, but difficult isn’t a word Jeff shies away from.

“It’s one of the reasons why we picked working with Airbus as a partner, There’s a lot of qualification that has to go into Scalmalloy as a material and then the coupons having printed it using one particular process, and then taking those end results and testing that, then taking the entire structure and testing that. It is a rate limiting step but when aerospace pioneers it gets adopted more quickly in other sectors.”

Devious Design

As Jeff says, the fact that “you can pull a car to the side of the road” means that the trickle down effect of certification from aerospace to automotive and even further down the industrial hierarchy is a lot quicker. So once we have cracked it in the sky, everything on the ground will be generatively designed, right?

“One of the biggest challenges we have is a little bit more,” Jeff pauses for thought. “…sinister. We can create objects that engineers no longer have a way of mentally relating to. When we make things like this chair leg out of a solid bar, you and I both know how a solid bar performs; it clearly performs, as I’m sitting on this chair, under compression, it also happens to perform well, by accident, under torsion. If I’m (generatively) designing a chair I’m going to tell the software that it needs to perform well under compression but I might not ask for it to perform well under torsion, generative may very well design me a chair that as soon as I rock back on it, it folds underneath itself. It would have a design that is complex enough as I look at it that the torsion issue is not immediately obvious to me as an engineer, there’s something insidious in that, something we need to overcome.”

Jeff on stage at Re-publica-3

Autodesk is pumping funds into researching how they close that gap between what designers ask for and what that objects needs to be but it is a third technology at the junction of innovation and application, that generative and additive are patiently waiting at; machine learning is on its way and according to Jeff it will have exponential ramifications.

“Every single analyses that has been run by any engineer on the planet up until now has fallen to the floor,” Jeff bemoans. “Every one of those was a training set for computers to understand what’s really going on inside a particular design object. If that knowledge could be shared by one generative design programme it would already know how to design what we struggle to do today.

“Last summer we generated synthetic bodies, just shapes and subject them to computational fluid dynamics,” Jeff continues. “We just made object and stuck it in the wind tunnel, made an object and stuck it in the wind tunnel, we kept doing that over and over again but we did that under the supervision of a machine learning system. After 10,000 iterations we were able to show a novel object that the machine learning system had never seen before and it would simply give us the flow field within 2 per cent. That’s pretty damn good!”

So with all that in mind I asked Jeff that with his current knowledge of the convergence of these three paradigm shifting technologies, what would he be studying if he were a to become a student now? “It’s a good idea to study what has come before and understand that precedent but not to be limited by it. I don’t think there’s any problem for students to continue to be exposed to, but not reliant upon, drawing things by hand or using the computer.

“We’re going to end up as mentors to these tools because they’re going to be remembering what kind of work they’re performing. In the same way that you go to art school to learn different kinds of genres, why wouldn’t it be the case that a computer that has a memory of its interaction with humans shouldn’t also be taught about different kinds of classifications of things. In the fullness of time I might want to say ‘Design me these glasses as if Philippe Starck did that’ and it would know what I meant. Or we could do some kind of semantic transfer where I say “I really like my glasses and I wish I had some cutlery that was of the same kind.” the ability to be able to say make this like that is still something that is purely human but I think that it can be learnt by a computer.”

To get the latest updates straight from key stakeholders, sign up for Additive Innovation 2016!

This post was originally published on, and has been reprinted here with permission