Sauron And “Friggin’ Lasers”: Autodesk’s CTO Prepares For The Next Century

Jeff Kowalski’s home is run by Sauron—an all-seeing eye that keeps track of everything he’s got going on in the house. It makes instant decisions for him so he doesn’t have to waste brain power remembering what or when he has to do something like water the plants, pay the bills, or check up on his kids’ homework.

Kowalski is the chief technology officer at Autodesk, one of the world’s largest software companies. Over the years, he’s perfected Sauron, what you might call his own personal operating system, so that it stores every piece of paper that comes across his desk, keeps his email inbox clean and simple, and reminds him of even his most remedial tasks, both at home and at work.

“This is where I live,” Kowalski said of Sauron one day, sitting behind his desk in Autodesk’s San Francisco offices. “This actually tells me everything I need to know . . . every possible thing that I could ever need to do is here. Forever.”

When you visit his office, one thing you notice is how incredibly clean it is. My desk is covered with paper, but there’s hardly any to be found on Kowalski’s. Yet Sauron, which he can access on any computer or connected device he owns, stores almost every piece of paper he’s ever had. A drawing of the interior of a TV from his childhood? Check. Every receipt? Absolutely. A free-cookie-per-visit card to a department store from years ago? Why not. It’s all digital, there for the finding when he needs it, or just sitting there, a collection of ones and zeroes that he may never again call upon.

His fastidiousness is well known at Autodesk. Among the people who work for Kowalski, who revere him and see him as an irreplaceable leader, it’s also something to tease him about.

“He hates paper,” Maurice Conti, the head of Autodesk’s Applied Research Lab, and the director of strategic innovation in the company’s Office of the CTO—or OCTO, as Kowalski’s sprawling domain is called—told me recently. “I love driving him nuts coming into his office with, like, a 100-page printout and plopping it onto his desk, just to mess with him. You can tell it drives him nuts.”

Software companies like Microsoft, Adobe, and Oracle may be household names, but Autodesk probably isn’t one most people know—despite the $13.3 billion company making tools essential to the lives of everyone you know. That’s because it develops design software like AutoCAD, 3ds Max, Maya, and others that are used to design almost everything we experience in our homes and in the world around us, from office buildings to vehicles to furniture, airplanes, computers, lamps, and more. Every visual effects Oscar-winning film since 1997, for example, has used Maya.

While Autodesk’s many product groups are focused on the present and the next few years, Kowalski’s OCTO is looking down the line, 10 years, 20 years, even 100 years into the future. That’s why OCTO is home to groups doing work on things like generative design, telling computers what we want them to accomplish instead of drawing upon what we already know: robotics, or finding ways to get robots to work alongside humans; BioNano research, or molecular visualization, genetic design, DNA nanotechnology, molecular design, and more; machine intelligence; augmented reality and virtual reality; and more.

Autodesk won’t talk publicly about the direct economic impact of OCTO, but its executives point to work being done by Kowalski and the many people under him that will one day be very important to the company and its many customers.

As Mark Davis, Autodesk’s senior director of design research, put it, “The largest financial impact of [Kowalski’s] leadership has been [developing] strategic relationships and engagements with leading customers—Nike, Airbus, NASA—as well as a host of net new customers.”

Conti recalled talking to Airbus about robotics recently, and being told that the aviation giant was very interested in how Autodesk was thinking about getting robots and people working together. Kowalski may not have specifically come up with that approach, but it was his encouragement of Conti and his team that freed them to pursue all-new ways of thinking on the topic.

“We’re surprised, wait a minute, you’re Airbus [and] we’re just fooling around in the lab, are you really interested in what we’re doing, like, for reals,” Conti explained. “They go, ‘No, no, because of the way you’re approaching this, because you haven’t been doing this for 30 years, and you don’t have the ‘right way’ to do it’. . . For instance, on the assembly of aircraft, which is very manual, with our approach, you can actually put a robot into that space, whereas without it, you need to automate the whole thing. The human isn’t exactly where the human is supposed to be at every given millisecond, which is what traditional robots need. But things are a little bit more loosey-goosey. If the robot can perceive that and correct in real time, you can do stuff you couldn’t do before.”

One of the keys to Kowalski’s approach is to give his people a mandate to look into the future and figure out what the company doesn’t yet know but will need to in the next five years and beyond.

Conti, who had worked at Autodesk in the past and then left the company, remembered sitting down with Kowalski when he was thinking about returning, and listening to his soon-to-be boss explain what his job would be.

“He says, ‘Okay, what I need you to do is go look in our blind spots.’ And I said, ‘Oh, by definition, you can’t tell me where to go look?’ And he said, ‘Exactly.’”

Kowalski himself might put it differently. He would say that OCTO’s charter is to figure out the things that Autodesk isn’t organically set up to pursue. The company’s core product business is “a big battleship that is set up to go somewhere, so we are intentionally not doing that stuff,” he told me. Rather, the point is to look well into the future, to try “to figure out how Autodesk survives the next 100 years, not the next 100 days.”

That’s tricky business. It means identifying things that will be important to Autodesk’s future and putting them in the context of the present so the company’s product groups can prepare for them.

“We often liken it to imagining the early Victorian explorers, going to the deep jungles of Kenya, and trying to come back and explain a giraffe . . . like, it’s a horse, but with a long neck,” Kowalski said.

Often, it’s not just finding what Autodesk and its customers will need. It’s also about helping them understand what they don’t want, even if they think they do. For example, the types of futuristic computer interfaces from movies that let users control things in the air just by moving their hands around.

“Almost every customer conversation we get into,” Kowalski said, “They say, ‘I saw Iron Man and I loved how they do that thing in mid-air and they design something,’ or ‘I saw in Minority Report, the sliding and glove interface.’ And it looks appealing for just about every one of our products, from Maya to Inventor until you get to the actuality of that thing, and it turns out the reason we use keyboards and mice . . . is it’s super fatiguing to design in mid-air, and you lose precision.”

Kowalski sees that his mandate is to find smart people and set them loose synthesizing information, taking ideas from one domain and seeing if they work in others, like figuring out if there are lessons that can be learned from the entertainment industry that can be applied to building bridges.

A Different Kind Of R&D

Every big company talks about R&D—research and development. But though Kowalski is in charge of research, he prefers the term “risk and determinism” and how to make that work for Autodesk’s product groups.

“Known inputs go in the front end, and super high-quality stuff comes out the back end and it’s impactful,” he said. “The machine works best when you don’t put junk in the front. What we’re trying to do is figure out in that risk-absorbing front end, how we take the crazy ideas and make them viable for the deterministic side of the business.”

The idea isn’t even to try to bring these ideas to known products or technologies or solutions to problems. Instead, it’s about throwing futuristic spaghetti at the wall.

“We’re not even asking the ‘what is it for’ question just yet,” Kowalski said. “We’re just asking, What can we do? Literally putting two and two together. We try to go out and find as many twos as we can and bring them back into the lab . . . and try to assemble different twos and see what happens. We might take something from the entertainment industry and bring it all the way over into construction. You might imagine that those two fields don’t really marry together, but that doesn’t bother us.”

As those at OCTO smash the twos together, they’re looking to evaluate the results on three main criteria: Is it desirable—does anyone want it? Is it feasible—can it be done? and is it viable—is it worth doing? When the answers are yes, they devise something that looks like a business plan and hand it over to a product group.

Designing in the cloud is a good example, Kowalski said. In the past, most design work was done on a single workstation under someone’s desk, meaning there was literally no outside-the-box thinking.

“We started noodling on that problem, what happens to design if you unconstrain some of the things we thought were constraints,” Kowalski explained. “What happens if you have more computing power than you would ever need to explore all the solutions?”

That experiment led to Autodesk’s Dreamcatcher initiative, a cloud-based design system that takes criteria and automatically generates countless designs that meet them. Most will be useless, but a small number may be better, more interesting, and more efficient than anything a person could come up with on their own.

Here’s an example: a chair. Existing tools could tell you if a design was strong enough to support a person, and how much it would cost to build. But they couldn’t tell you if the chair will be comfortable, which is more than a little important. That’s because no software could adequately model a human.

“Which is a really crazy thing when you think about it,” he said, “because all the design on the planet is meant for some kind of human consumption or interaction. You’re in a building or a car, yet we have no idea how a human would actually [interact] with that design output.”

So Kowalski’s people proposed the “crazy, audacious” plan of building the “most robust parametric human model ever.” That’s now a real project at Autodesk, aiming to understand everything a designer might want to know about people, and how to apply that to the design of products like clothes, chairs, cars, shoes, and so on. In other words, the businesses of countless Autodesk customers.

“I actually think Autodesk is in one of the best positions to change outcomes on the planet,” he said, “because we supply the tools that make every damn thing, everything. There is not a single [manmade] thing on the planet that doesn’t come through a design tool. Nothing. So if we want better stuff on the planet, we’ve got to have better tools, better ways of making and designing things.”

That may be a stretch, since there’s doubtless many things that people design without software to guide them, but you get his point. As well as his contention that Autodesk is trying to become as much of a foundational company to people in the future as General Electric is and has always been.

Growing A Car

A lot of what goes on inside OCTO is definitely not ready for general consumption. Yet it is very much about showing where things could go, given logical technological advances.

An example is the Hack Rod project, almost certainly the world’s first AI-designed car. Led by Autodesk Research fellow Mickey McManus and Mouse McCoy, the CEO of the media production company Bandito Brothers (yes, the project was led by Mickey and Mouse), the initiative was about learning how form and function work in nature, and how that might be applied to product design.

The project began with the most basic car chassis. McManus and McCoy covered it in sensors, put it out on a race track and began testing it, collecting data to feed back into Dreamcatcher.

“The data is literally all the stresses and strain in the vehicle as it’s being driven over time,” Kowalski explained. “This tube in the structure of the car’s actually under stress as it’s going over a bump, and you get to see which areas are experiencing those stresses and which areas are not as it’s driven under those conditions . . . We literally use that data as the target for this form of directed design. We tell Dreamcatcher: Make a car that supports those forces.”

It turns out that by connecting purpose to form, you can replicate the way cells in nature grow and transform. That’s why Autodesk, a design software company, supports research in synthetic biology—because it’s interested in how to use biology as a programming platform in years to come.

“Today, our life sciences work does not directly connect to automotive,” Kowalski said, “but in the fullness of time, I know it’s going to.”

In Conti’s group, work is currently under way on what he’s calling Meshbot, a new way of looking at additive manufacturing. Using two robots working in conjunction, the process allows for building much larger and more accurate three-dimensional physical products than even the most sophisticated 3D printer. And because it works with metal and carbon fiber, it means it’s possible to make architectural elements, furniture, and even high-performance, lightweight structures for automotive or aerospace uses.

Meshbot also has a virtual element: Visitors at Autodesk’s annual conference, Autodesk University in November, will be able to immerse themselves in a virtual-reality simulation where they can build a bench using meshbot’s principles. They’ll be able to grab, pull, and push elements of the bench, but the system is designed to be smart enough not to let them create anything that would be impossible to make in real life.

“The stretch goal is that we’d like to have some parameters in there that keep you from making something that’s ugly,” Conti said. “So if we can teach the computer some very basic aesthetic rules, if you try and pull on this thing past a certain point, it’ll go, ‘No, that’s too much overhang. That’s just ugly.'”

Conti’s lab is also working on developing a robotic additive manufacturing system that actually watches its own progress and adjusts on the fly, meaning that instead of creating something where small errors compound as successive materials are put in slightly the wrong position—as is the case with standard 3D printing—the lab’s robots would know immediately if something was placed incorrectly and would adjust.

These are the kinds of projects that groups within OCTO are encouraged by Kowalski to pursue, even as he himself remains largely hands-off.

“Jeff doesn’t impose himself on [a high] level of detail,” Conti said, “unless he has a great idea.”

Kowalski is extremely popular with OCTO’s staff of 101 people, Conti said, so much so that if he left Autodesk, “you’d have an exodus, because a lot of us are here because of him. Very plainly.”

Besides the fact that Kowalski’s a good manager who is also good at business, Conti said, he’s got other attributes that appeal to the many big brains inside OCTO.

“He’s wicked smart,” Conti said. “Like, scary smart, which I think for a lot of the people in OCTO, who sort of pride themselves as reasonably smart folks, is an inspiration.”

Conti said Kowalski is both able to synthesize high-level information and lots of data coming in from different places, and able to understand technology at the level of the people working for him.

“He’ll go into GitHub and look at anybody’s code and be able to critique it,” Conti said, “so he has the technical expertise that is deep enough to really understand what the state of a technology is, all of the BioNano stuff, he actually understands what the geneticists are talking about, and that’s the case with all of the different groups that report to him.”

That’s important from a business perspective, too. Conti argued that Kowalski is able to evaluate in real-time what a startup might be telling him in a meeting, and understand whether a technology is less than impressive or worthy of being acquired.

All of this, of course, is time consuming, and would probably occupy most people’s minds day and night.

Kowalski, though, sleeps just fine, unencumbered by worries about things he has to do, or worse, hasn’t done. That’s all thanks to Sauron, which allows him to work on exactly the things he needs to during the day, and be free to focus on life with his family in the evenings.

Yet there’s also a big disconnect between his work life, where he focuses on the big ideas of the future and talks to the biggest brains at Autodesk, at other companies, at national laboratories, and elsewhere, and his home life.

“For us [in OCTO], it’s like friggin’ lasers in the morning, followed by ‘Let’s resurrect the dinosaurs in the afternoon,'” Kowalski said. “The hardest thing actually is not the stuff I get to see at work, it’s coming home. My wife has to continually remind me, ‘Look, at work, you live in the future. At home, you live in the present.’ When I get home, I’m slightly more frustrated that the house doesn’t work ideally. You know, it’s not all friggin’ lasers already . . . At home, we’re still working on putting the toilet seat down.”