Lab Report
Issue: Volume: 33 Issue: 4 (April 2010)

Lab Report

A look inside Autodesk Labs, where technologies are put through their places

What is Autodesk Labs? It is a group within Autodesk that is home to innovative, new technologies and collaborative development. Started approximately three years ago, Labs explores technology that may be commercially relevant to customers in design—from architecture and manufacturing to film and games. The goal of Labs is to take invented technologies and integrate them into early, experimental prototypes that the Autodesk Labs community can try out, with the goal of determining whether the technology could be made into a product or a function within an existing product. Basically, they strive to engage the design community to determine whether a certain technology has value.

Recently, Brian Mathews, vice president of Autodesk Labs, sat down with CGW chief editor Karen Moltenbrey to discuss what this group does and the effects of that work on designers around the world.

When was Labs formed?
There were a few different incarnations of it, but the current format started about three and a half years ago.

How many people work in Labs?
Approximately 30, but that’s misleading since my core team coordinates projects that leverage technology experts anywhere in the company. The virtual team is much larger and dynamically changes based on project needs, with people rotating in and out continually.

Where is Labs located?
Because of its dynamic nature, this question is hard to answer. If we just consider the core team, I’d say we are headquartered in San Francisco, with staff in Chicago, Phoenix, San Rafael, France, and China.

How are the groups within Labs formed?
Some projects, like Project Showroom, are created in four different locations. The project has a lot of specialty areas in it, like high-performance computing experts (people who know how to run clusters with thousands of cores), content schema experts, computer graphics global illumination experts, material science experts, Web design experts, data center experts, and so forth. So a project like Showroom spreads into several different areas and different teams globally.

What is the group’s goal?
Our primary goal is to engage a community of early adopters in the development of new design technologies. Everyone in the company has a role in innovation. Labs’ role is to sit in between our research teams and our product teams to develop inventions into innovations: We mature horizontal technologies to prove they are ready for commercialization by product teams. 

Take multi-touch, for instance. Four years ago, before the iPhone shipped or Windows 7 had multi-touch, Autodesk Labs did a bunch of “maturing” development in this area. From an invention point of view, multi-touch had been around since the 1970s, when universities were working on it. Several Autodesk researchers had in fact been working with multi-touch for decades. But, obviously, multi-touch hadn’t become a practical innovation. That is where Labs came in: We took something possible and attacked some of the technological and business barriers to making it practical. We created several prototype products and engaged our community of customers to gather feedback. We are successful in maturing a technology when we are able to graduate a technology out of Labs and have a product team pick it up.

What does Labs do, then?
The mission of Labs is to explore and validate new approaches to design technology through functional prototypes. Our vision, as I mentioned, is to engage a community of early-adopter customers in the development process. We have a lot of R&D that happens directly in Labs, but we also take emerging technology from other parts of the company and connect it to our early-adopter user community.  
In a traditional product development process, new technology is kept secret until it hits beta. In that process, customers are only involved at the very end, where all you can do is fix bugs. In contrast, the Labs process is to take technology and involve customers in development from the very beginning. For example, we’ve taken emerging concepts like cloud rendering and built Project Showroom, or 3D software as a service (SaaS) and built Project Freewheel (for viewing and sharing designs over the Web), or application remoting and built Project Twitch (running application trials over the Web). Each of those examples started out as a tangible prototype that evolved with customer feedback. Look at Freewheel: We took the idea of allowing people to collaborate using a SaaS model; early adopters people who want a competitive advantage through technology—helped us grow it into a Web service that gets hundreds of thousands of hits per year. These users want to know what is coming down the line so they can shape it, form it, even before it may be practical.

Autodesk’s Project Freewheel combines a dedicated Web site and a Web service for interactively viewing design data.

Where do you find these early adopters?
Mainly through our Web site. We have a community site ( We do a daily blog, and that gets a lot of traffic. We also have e-mails and newsletters.  People in the know and who follow technology and Autodesk the early adopters find us. We also do various trade shows, like Autodesk University.

How do you decide what to look at?
My analogy is that I am the museum curator; I don’t make the art, but I choose what gets shown. There are others who create technology that may be too risky that’s not ready for prime time, or hasn’t been productized into a tangible prototype.  Some stuff just doesn’t make the cut it’s not thought provoking or doesn’t really need an early adopter community to perfect it. Many of those technologies are handled by a product team.  I try to keep most little utilities or productivity tools out of Labs. There are other avenues for those things.

What do you look for in a technology?
We are looking for horizontal tools and technologies that would make designers in any discipline more productive.  

Are you concerned about competitors stealing your ideas?
That comes with being open, but it goes both ways. We’re dealing with immature technologies where others may copy us both in what works and what doesn’t work. We can be a catalyst that gets the entire industry involved in finding answers. By being open, we’re able to solicit a tremendous amount of user feedback that helps us shape the product.  At the end of the day, our main goal is to make a tool that works for our customers. If we make something valuable for them, how can you really lose with that?

How do you deem a technology good enough to graduate out of Labs?
Each technology is different. Once Labs has gotten rid of some of the technology speed bumps, and when our engineers say the technology has legs, and when our customers show interest, then product teams often say, ‘Hey, I want to own that.’ That’s when we graduate it and hand it off. Some technologies never make it and are killed either because the prototype didn’t work as desired, or customers were disinterested. Failures are part of the process: By getting user feedback early, we can free up engineering resources for new projects that show more promise.

Our users also pitch ideas. They usually want a new feature in a product they already have. It’s rare that they come up with a brand-new invention of how to apply technology. Our customers usually innovate in their own fields, like bridge building or airplane manufacturing, so they bring deep domain expertise that augments the software development skills we have. The advantage of a large community is that there are always people who know more than you about something.

While the goal of Autodesk Labs is to graduate a technology, not every one finds its way into a product. Above are images from Inventor Fusion, a history-free modeling concept.

How long does it take to graduate a product from Labs?
We’ve only been around for three and a half years, but if you go to the Labs Web site, there are dozens of technologies that have been built over that period, and there is a page that shows all the technologies that have graduated from Labs. A lot of the time these technologies become a feature in existing products. Some of the translators and analysis tools have shown up in Inventor, Revit, and AutoCAD. Some have become their own product, like Inventor LT.  Still others we have walked away from for a number of reasons. For instance, three years ago the picture-based visual product search concept was interesting but the technology wasn’t ready, and we withdrew it from the site.

What are some of the technologies with great potential?
Inventor Fusion, which is the idea of re-envisioning what Inventor looks like if you do history-free modeling. You have direct manipulation. You can grab your model, like a hole, and drag it and make it bigger without having to go into the parametric recipe. It is a lot simpler and has a much easier learning curve. Unlike other history-free modeling systems, with this one you get the best of both worlds since it automatically synchronizes a parametric recipe, which is technologically complicated. Many people said it was impossible or very, very difficult. It’s currently not production-perfect, but amazingly close, and that’s why it is in Labs.
Another technology was launched at this past AU, Inventor Publisher. This is a stand-alone technology preview. We have Inventor (for doing product design) and Design Review (that allows people to review, mark up, and play assembly animations of what the product looks like), but customers told us they still needed an easy-to-use tool that helps them create clear and comprehensive technical instructions. A CAD group can design the chair you are sitting in, but the technical publications department makes assembly instructions. Now with Inventor Publisher, a manufacturer can take the CAD model and augment it with other information and help publish the data in an interactive way to its customers on the Web or even an iPhone.

What about Twitch?
That is one project we have been working on for years. We saw the promise a long time ago when we did Project Freewheel. When you look at SaaS applications like Google Mail or others, they are a compromise when compared to their desktop equivalents. You can’t really do 3D graphics on the Web with current standards. So we thought about how far we could push 3D over the Web. Freewheel was a zero client; you didn’t have to install anything because it was purely HTML. We were able to do 10 frames per second of full-screen graphics on entire cities because we would render on the server; in doing so, we did some tricks with JPEGs and sent them down to a browser or smart phone, where we found we could do some simplified 3D over the Web. We then started doing complex user interfaces. When we saw what we could do, we thought, What would CAD look like on the Web? You would want perfect fidelity, zero latency, and high performance. Project Twitch came out of that. We investigated a lot of technologies in this space that just fell a bit short. So while we had the concept for a long time, it is only recently that the technology is ready.

Think of Twitch as a 1000-mile-long VGA cable. We replaced that cable with an Ethernet cable. People tried to do this before, but their solutions required tons of bandwidth, or they degraded image quality, or they buffered, which introduces latency. While many videoconferencing systems work just fine, those technologies don’t have the resolution and low-latency qualities you need to remotely run a CAD application over the Web.  We need latencies that are sub-50 msec. The best standard compression technologies out there have about 250 msec and can eat up 10mb of bandwidth for HD video. Twitch can go as low as 3mb, with latencies as low as 10 msec. So with Twitch, we have all the benefits of a SaaS application easy to implement, no capital expenditure, collaboration, real-time data access, data centralization with all the benefits of desktop software because we are running desktop software. Oh, and it runs better than your current desktop software because we have super-tricked-out hardware in a data center that is dedicated to you when you’re awake, but the price is lower since we re-use that system while you’re asleep. Before, if someone in New York and London wanted to collaborate, they constantly had to move their files and versions around. But if you can just move the video around, you can leave the data in the data center, and you have instant access to it over fiber channel. This is the first no-compromise SaaS solution benefits of the desktop with all of the power of the Web.

How does SaaS come into play?
SaaS is not just about doing stuff faster. You can now take a 1000-CPU-hour computation and run it on one machine for 1000 hours and it costs the same as if you ran it on 1000 computers for one hour. In the past, that was not the case. You can now rent supercomputers in the cloud by the second, giving you access to compute power you never had before. One example of this is Showroom. Imagine laying out a kitchen floor plan and getting a physically accurate photorealistic rendering in four seconds. Normally that would take four CPU hours, but with 2000 cores, you can squeeze that into four seconds. The Project Showroom concept is about letting you do really high quality rendering in near real time by using thousands of cores. Technically, writing software for that is different than how you write desktop software.

Project Showroom lets users mix and match product selections, textures, and colors from leading manufacturers using a Web browser.

At Labs, I look at our recently acquired analysis, simulation, and visualization technologies, and consider how to adapt all those engines into a supercomputer in the cloud. If you want to design a green building or a sustainable building, someday we’ll do real-time simulation while you are designing. Currently, people design the building, send it out to an analysis firm; a few weeks later the results are sent back, and it either meets code or fails, in which case you have to redesign it. If we can have the analysis happening in the background while you are designing so every couple of seconds you get a result back, you will design a better building.

What allowed all this innovation to occur?
A combination of many things. Like any innovation, it is evolution on a path. You have to have the Internet, search engines, content that is in a searchable universal format, rendering, and simulation technology. All this information has to be written in such a way that it can be done in parallel over multiple cores rather than just faster single cores, an operating system that can manage a large number of cores, and network implications. Most Web applications use redundant components, and it is very expensive. That is how most Web hardware is built expensive and reliable. That’s great for normal Web apps. But when you get into cloud computation, you no longer want to use the same approach; you want the software to be fault tolerant, not the hardware. In this way, you can buy much cheaper hardware. That is another innovation you get the price down, and then you need a business model. Most people cannot run their own big supercomputer centers, and that is where Microsoft Azure and Amazon EC2 come in they offer computers that are rentable on-demand and priced by the minute.

These services are all focused at the small players right now. Bigger companies generally run their own data centers at some scale, but they have to have the machines. The equipment has to cover their peak usage, and they have to buy enough for that one hour a year, for the worst-case scenario. The rest of the year it goes unused. Now with cloud computing, the bigger companies can right-size the big data centers so they can buy based on the average, and when the need arises, they can make an API call and burst the application into EC2, for example. It makes for a more efficient data center and drives the cost of computing down dramatically and that drives innovation.

Moore’s Law has computing power doubling every 18 months. Now we are seeing Moore’s Law to the third dimension with cloud computing. Moore’s Law has always been two dimensional two dimensions on a chip, a two-dimensional surface. In a data center, you can stack computers into the third dimension and you get a different Moore’s Law through Web access. So the cost of computing is going down, and that has implications across all our lives entertainment, medical, and so forth.

Is this invention or innovation?
What really counts is innovation, not just invention. Invention is great, but until you innovate about how to use it and solve a real-world problem, with a real-world delivery platform, invention is only intellectually interesting for us geeks. But innovation is where the productivity gains come from. It is how you bring sometimes-old ideas together to solve real-world problems, and you’ve got to involve the customer in the process.