#HTE
What Happens When AI Starts Designing Things? Autodesk CTO Sounds Off
Autodesk University is a massive annual conference aimed at “those who design, build, make and create.” It’s where industrial designers rub shoulders with architects and makers, engineers and construction foremen, scientists and entrepreneurs from all over the world.
Part of the reason to attend is the excellent classes, which we’ll get into later. Beyond practical skill-building, another good reason to go is for the packed-house keynote speeches, where Autodesk’s future-gazing wonks lay out the technology trends that will directly impact our design work.
These are not your typical BS “Wouldn’t it be cool if…” scenarios; since Autodesk is a software company that actually makes the tools that many of us will actually use, if they say something’s coming down the pike, you can lay bets. The company stays on the cutting edge by getting out ahead of these trends and designing tools to work within them.
This year’s AU kicked off with an illuminating (some might say terrifying!) talk by Autodesk Chief Technology Officer, Jeff Kowalski. We’ll print the parts we thought you’d find most relevant—on the future role of artificial intelligence doing design work and where and how we humans fit into this.
The talk has been edited for clarity and brevity.
How Our Tools Affect Our Designs
“Everything you see is a product of a person, an idea, and a tool. Throughout human history we’ve always had this urge to shape the world according to our ideas, and it’s our technology that’s given us the power to make those ideas real. We started with simple machines, using wheels and levers to stack stones on top of each other. These tools gave us the power to imagine things like the pyramids. Then we developed mass produced steel, rivets, cranes, elevators. These new tools helped us to reimagine what buildings could be, giving us the first skyscrapers.
"With every tool, there’s an upside and a downside because while they initially expand our capabilities, they ultimately also constrain our thinking. For example, the same tools that let us build skyscrapers also gave us cars that look like this:
"Boxy and boring. But then remember that year when all the cars just sort of metled around us, like this?
"Well, this change was triggered by a new tool, software that could model complex curves, allowing designers to express a whole new language. Technology has always helped us express our ideas out in the world, but our tools have always been the rate-limiting step for our creativity, a filter through which all of our best ideas had to pass until now.
"Today, powerful technologies are emerging and converging and taken together, they’re giving us not the limited expressibility of the past but an infinite expressibility that will help us shape the future. Instead of limiting our imaginations, this emerging tool set is going to help us amplify our ideas.”
The Increasingly Fast Rise of Artificial Intelligence
“Possibly the most important thing happening in software today, is artificial intelligence and machine learning. More than 60 years ago, a clever programmer taught a machine to beat humans at Tic Tac Toe. Then 45 years later in 1997, Deep Blue beat Kasparov at chess. In 2011, Watson beat these two humans at Jeopardy. For a computer, it’s a lot harder than chess. This time, rather than working from pre-defined recipes, algorithms, the computer actually had to use reasoning to overcome its human opponents.
"This year, a program called Alpha Go beat the world’s best human at Go, a game so complex it has more possible moves than there are atoms in the universe. In order to win, Alpha Go had to develop a sort of intuition about the game. In fact, at times, its programmers weren’t exactly sure why it was doing exactly what it was doing.
"If we’re looking at a timeframe for these milestones, we see something really amazing. There’s something exponential happening here. In less than a single human lifetime, computers have gone from learning a simple child’s game to mastering the game recognized as the pinnacle of strategic thought. There’s two things responsible for this acceleration. First, unprecedented available computing power for things like GPUs, multi-core and cloud. The second is that we’ve taught computers to teach themselves.
"Let me give you a simple example. Think of the classic Atari video game Breakout. How did you learn Breakout? Spending long afternoons in the family room or den playing over and over again. Let me tell you how a computer recently learned Breakout. Told only to maximize the score and to twist this one knob that controls the paddle, a system called Deep Mind learned how to play the game.
"Then it learned how to play the game better than any human ever had in just one night.
"How did it do that? It played in computer time, which means playing millions of games in parallel in the course of a single night. Compare that with how we humans learn and share what we know. Just because your buddy got good at Breakout didn’t mean that you got good at Breakout. Machines are different. Once this single machine had mastered Breakout, all machines mastered Breakout forever.”
Artificial Intelligence is “Getting More Creative”
“The machines aren’t just getting smarter. They’re also getting more creative. Up until now we’ve mostly used computers to solve left-brain-style, logic-requiring problems. Even in the case of designers, artists, writers using computers, the inspiration really only ever come from our side of the screen.
But now, computers are poised to transcend that barrier and make the journey into the realm of human creativity. The ability to grasp the unexpressed, to distill the very essence of the thing, that’s what’s going to make the computer a better creative partner for us.
"When I say creative, I mean exactly that, including things like the creative arts. We saw how one computer essentially went to video game school overnight. Why can’t a computer also go to art school overnight? Here’s one that did. It studied Rembrandt, and then it painted a brand new one.
"At Autodesk, we’re bringing this kind of machine learning to the 3D world. We’re feeding our algorithms huge amounts of 3D model data so they can grasp the essence of designs that we work on every day. With that new understanding, the software can take a 3D object, a generic chair for instance, and apply a specific style to it, make it more fluid, make it more Philippe Stark, more Cubist or more Le Corbusier.
"Playing board games, video games, creating paintings, even chairs. Computers are getting better at things like this which require human-style capabilities. Intuition, generating hunches, making creative leaps, expressing imagination. Another way to put it is that computers have always been a little bit like Mr. Spock, but today, they’re becoming a lot more like Captain Kirk.
Spock is logical and brilliant, but as we saw countless times on Star Trek, that was almost never enough to save the day. In fact, it was usually Captain Kirk who came up with the ultimate solution for whatever it was that they were facing. It was usually something that was driven by hunch, intuition, and creativity. Today, that’s exactly the kind of unbridled imagination that we need to address our biggest challenges.”
Generative Design
“Here’s another technology that’s part of the convergence that’s going to give us infinite expressibility. I’ve been talking to you about generative design for a few years now. It’s a way of collaborating with computers where we don’t tell it what to do. We tell it what we need. We can tell the computer what we want to accomplish instead of telling it what we already know.
Here’s an example of what I mean. This summer, one of our interns wanted to see if she could design a chair using generative design. Now her goal was to design a chair that was beautiful and comfortable and strong enough to support the weight of whoever might be sitting in it. She fed our generative design tool, Dreamcatcher, her goals describing what she wanted in a chair. Then she sat back and she let it explore the entire solution space on its own.
"It created thousands of options, all of which met the criteria and including many designs that she would never have come up with herself. Here’s the chair designed by collaboration between Dreamcatcher and Brittany.
"Despite her talent, there’s really no way that Brittany could’ve designed and fabricated a chair like this in just a couple of weeks. The computer augmenting natural talent, that’s what I call infinite expressibility. In fact, I look forward to see how all of you harness your infinite expressibility when we release Dreamcatcher and our generative design tools commercially early next year.
"Here’s another project that I want to share with all of you. Last month, we worked on a project with an automotive partner to redesign this thing:
"It’s a rear suspension upright for a passenger car. Using our generative design tools, we redesigned this part and ended up with two new options. The one on the left [Editor’s note: Sorry, the photo did not come out] removes weight by optimizing the geometry. The one of the right [photo above] removes even more weight by adding an internal lattice for the optimized shape. The software isn’t filling the void with the same repeating pattern. It actually mimics bone by adding material only where it’s necessary and removing material where it’s not.
"Computers are moving beyond optimizing geometry and the recent performance of things. They’re starting to understand something even more complex: The needs of people. Our team in Toronto is moving into a new building. We decided to use it as kind of a living laboratory. We’re using generative design to re-imagine what an office can be. As we came up with the process of planning the space, we knew we wanted to maximize productivity and create a really great experience for all the people working there.
"We used generative design, and rather than feeding it the forces affecting parts, we fed it the forces influencing human experience. We surveyed all of our employees and put their preferences and work habits into the system. Then the system evaluated that survey data against a set of large constraints, like the boundaries of the building, the fixed locations and the fixtures and so forth. It generated thousands of options, thousands of alternative floor plans. In this one little space, the system was looking to maximize outside views, minimize distractions, and prioritize personal relationships.
"The system is really particularly good at reconciling multiple, often competing goals. We weren’t just pushing cubicles around like Lego blocks. We weren’t stuck with the first design that worked. With generative design, we were able to create the best experience possible.”
Are These New Technologies a Threat?
“So far today, I’ve been talking about technology, but for me, that’s only half the equation. What about us? How do people fit into this vision of the future? I’m sure that some of you’ve been thinking about these technologies maybe as a threat. I want to tell you that’s 100% wrong. These technologies are not a threat. They’re more like superpowers. What’s the real threat? It’s any competitor that adopts these superpowers more quickly than you do. Look, these machines, the robots, the computers, they’re not coming for us. They’re coming for us. They’re not bringing the apocalypse. They’re bringing us beer in a self-driving truck.
"The reason the prospect of these machines and robots is so scary is actually because they’re so powerful. These tools of imagination creation are challenging our thinking. It’s not something that we experienced before, but that’s the consequence of exponentiating technology. It stretches our thinking and also our capabilities. I don’t think that’s daunting. I think that’s exciting.”
How Do Humans Fit Into This?
“Another part of this equation that we have to think about is talent, the people doing the work and using the technology. Talent used to be about stability. Now it’s about mobility. 40% of the US workforce is composed of freelancers, consultants, and other contingent workers. How does that impact all of you? All of this mobility means that you now have access to a vastly larger pool of talented people than you’ve had in the past. Imagine the flexible resources that you can now bring to bear on any challenge you face. Just as you should be embracing the technology, you should be welcoming new kinds of talent.
"Here’s my last point about what we need to do to really embrace the changes that are coming. I’ve talked about machine learning, but what about human learning? Today the increased speed of change is creating pressure on all of us to learn more quickly. You know, if you’re going to keep up with tech and talent, you’re going to need to upskill at the same pace. If your education stops when you get that one monolithic degree, you’re doomed. In this dynamic environment, you can never stop learning because ongoing learning is the antidote to fearing technology and new talent, and embracing and using it instead.
"New technology, new talent, new ideas. For millions of years, we’ve been using this powerful combination to shape our world, but never before had we had such an abundance of opportunities. So many things to learn, so much to debate, to incorporate, to create. You’ve made a great choice by coming to AU this week. Right here, right now, this morning, we are all living in the earliest moments of an amazing new chapter in the history of making things. AU is the perfect place to explore one critical question:
"What role will you play in the future of making things?”
http://www.core77.com/posts/58040/What-Happens-When-AI-Starts-Designing-Things-Autodesk-CTO-Sounds-Off