Engineering.com interviews Autodesk CTO Scott Borduin about resuming his role in the position.
Scott Borduin, Autodesk’s current CTO, has been here before. Borduin started with Autodesk in 1993. He was the CTO before Jeff Kowalski took over the position in 2006. Although Kowalski left Autodesk a few months ago, Borduin and his group, OCTO, or office of the CTO, have not skipped a beat and are quite busy.
Scott, can you give us some of your history at Autodesk?
I started with Autodesk in 1993. I came in as part of the Woodbourne acquisition. We were a third-party developer and had created the first parametric feature-based CAD system on top of AutoCAD (formerly Mechanical Desktop).
A History at Autodesk
What have been your roles at Autodesk?
I was the senior architect of Inventor. When Carl [Bass] left to do Buzzsaw, I replaced him as CTO in 1999. I was CTO under Carol Bartz from 1999 to 2005.
What version of AutoCAD did you start and end with?
In 1999, it was R14. Carl was the one who shepherded that through.
That was a good release. You missed the entire AutoCAD R13 fiasco?
R14 was a good one. I was at Autodesk building on top of R13. I think we helped that team figure out some of what it meant to have a complicated object-oriented program on top of the base of ARX.
Can we credit you with rescuing the company from AutoCAD R13?
No, no, no. That’s not me. I came in the internet era. Right in the middle of the dotcom boom. I was spending a lot of time in the early years figuring out Autodesk’s relationship to this internet thing and what it meant to our customers. Like, where do we take the products?
AutoCAD was still Autodesk’s biggest product, and most users were still AutoCAD users?
Yes. The verticals were starting to play a pretty significant part by then. Mechanical Desktop and Architectural Desktop were both really starting to take off as verticals. Right about 2002, Inventor really hit an inflection point and started to take off.
Inventor began replacing Mechanical Desktop?
Actually, we bundled Inventor with Mechanical Desktop.
That’s what led to its acceptance. You were letting customers transition at their own pace.
Yes, when they were ready.
While Mechanical Desktop was built on AutoCAD, Inventor was a whole new codebase?
Inventor was brand new.
The Inventor team came to Autodesk from where?
Some of the people were from Novi [Mich., near Detroit]. The early Inventor team had a couple of us from Woodbourne: me and senior architect Raul Riviera, who is still at Autodesk. Then, we brought in a number of people from the outside. We got a user experience designer from SDRC and an advanced data object-oriented database designer from Mentor Graphics. We had a number of pools of disciplines pulled together as the initial core Inventor team. We then branched out. For instance, the whole assembly modeling environment was done in Novi. Drawing Manager, the initial drawing management program, had a team in San Diego, though, it was a distributed project by the time it hit market.
That explains what you were up to between 1993 and 1999 when Inventor launched. Then what happened?
I moved over to the CTO role about the time Inventor launched.
A Detour to Save the World
You left Autodesk in 2005. Why?
I decided I would leave and go try to save the world with a nonprofit.
Inspired by Robert Putnam’s 2000 book on the disintegration of American society, Scott Borduin launched a nonprofit, Democracy Talking. (Image courtesy of BowlingAlone.com.)
That didn’t work. The world still needs saving.
I failed miserably. In fact, what I tried to do was… You will find this amusing… I was dismayed by the state of…
Technology?
No, the state of our political institutions. I knew what it was like to be part of a high-functioning team. I felt like there had to be some things we could do to help our political system work better.
You formed a nonprofit to affect a change in the political systems?
Yes, we formed Democracy Talking, which was all about trying to create civic engagement.
Read Robert Putnam’s “Bowling Alone”. This was late 90s early 2000s. It was all about the decline of the civic institutions, the sort of glue that brought people together in the middle of the century. Everything from bowling leagues to the VFWs. These institutions brought people from diverse points of view together and made communities function. Anyway, we tried to recreate that. We worked with other nonprofits doing things like campaign finance reform. We tried to fix the whole political system.
It may need some fixing right now. Do you want to go back there?
I do not, actually. It’s difficult now. The less said, the better.
Back to Autodesk
As you were winding down with the nonprofit, you got a call from an old friend?
Buzz Kross called me up in early 2012 and said, “Do you want to come and have lunch?” That turned into an interview, which turned into my returning to Autodesk.
You came back to a different position?
I was called a group CTO. I had the job of looking out for technology across the manufacturing division. In general, bringing a certain perspective to those sorts of things. I’ve always been interested, not so much in just technology, per se, but the intersection of technology, business and people. That’s where things get really interesting, being able to see where things were going and helping guide the technology strategy and, to some extent, the business strategy of the organization. I did that for a number of years, intermittently running different groups like the user experience group, core modeling and components group. Those kinds of groups.
When Buzz left, I went to work for a few different people, and that brings me to the present.
How do you feel about following Carl Bass as CTO, who may be regarded as the prime technologist of Autodesk, and Kowalski?
You don’t get to be executives at their level without having the technical facility. When I decided I would take this position, I brought along an interest in technology, imagining where technology would go and telling stories around it. The current CTO organization is good at that. They’ve done a terrific job, I think, of changing the company’s image, moving its brand forward, turning it and helping transform it from more of a fast-follower company to a thought-leadership company.
That’s the Office of the CTO [OCTO]?
OCTO’s been a big part of all of that. I think I can bring some additional value from the last six years I’ve spent at the top level of the product organizations as well as a perspective on how the work we do in the CTO organization can integrate and make more of an impact on the rest of the company and our customers.
Jeff Kowalski, ex-CTO of Autodesk (Picture courtesy of Autodesk)
What would you say is Jeff’s legacy?
I said this to both Jeff and the team when he left: he was by far the most consequential CTO we’ve had in the company. Jeff, when he first took the job, did it somewhat the same way I did, with a relatively small team. It was more about imagining where technology would go, telling stories around that, communicating that internally and externally, and so on and so forth. Then, what he started to do was build a whole team that wasn’t just about examining the future from an external perspective but helping build the future.
Any examples?
Jeff starting to do things like generative design, interesting user interface enhancements. Some of the things that we’re investigating now around AI (artificial intelligence) knowledge and learning, and a whole bunch of other things that are going on in the organization.
AI and generative design, right? You mean merging the two?
AI and generative design are very much on my mind, like two sides of the same coin. If you want to talk to the purists about this, generative design is about using algorithms to synthesize multiple different possible solutions to a particular given set of design constraints, and then providing tools for people to curate those solutions and choosing which ones. Generative by itself does not imply classic deep learning, AI algorithms.
However, those algorithms can play a major part in making generative design more functional.
They can be used to make things?
AI algorithms can be used to do things like help classify the outcomes of generative design synthesis into shape categories or help determine the outcome of a simulation based upon previous similar simulations without having to run the entire deep algorithmic solution, etc. There are many things that generative design can do.
Is that Autodesk’s biggest focus now, AI and generative design?
We have teams in both those areas, so it’s certainly an ongoing area of interest.
You see them actually making usable shapes? You talked about the functionality of these shapes, but what about aesthetics? Do you think AI would ever make a part you can use in production?
Some things we are working on I can’t get into detail about, but let’s take a step back and consider the business of synthesizing alternatives in the generative world. You start with a model and determines its sensitivity to a particular set of design constraint and inputs. The models that readily fall to hand when we think about it are things like geometry and stress, strain and other results. Those are all things we know how to model today.
When you start to talk to engineers and designers about what else they’re interested in, they tell you more things. They are interested in what it looks like, the aesthetics, and the cost, what is it going to cost to produce?
Hold on. Cost may be an easy result. But is aesthetics too subjective?
Aesthetics are subjective. However, when you start to look at your favorite automotive company or other products that have a strong aesthetic component, you see each one of their companies have their own design language, such as companies like Mercedes and BMW. You recognize their products almost immediately because they have a certain design language. There’s this synthesis of different design elements. You can start thinking, “Why can’t we capture a design language like that?” You start to get into the realm of learning.
AI: Does It Write Itself?
CTOs traditionally have been keepers of the code. Their people write code and correct code. It’s all about code. AI goes beyond creating code. It learns. Am I correct?
You’re correct. The fascinating thing about AI is that the behavior of an AI system is not deterministic. Code is supposed to be deterministic. A given set of inputs should always yield the same set of outputs and a set of outputs that was envisioned by the coder. AI is beyond that.
The whole point of AI is to actually, in many cases, yield results that you didn’t anticipate.
Like generative designs does?
Like generative design does. This is a pattern that I talk to people about frequently. We’re going through this inflection point. When you go through inflection points, you may not even know it. We’re going through a major inflection point in the relationship between people and technology. Since the stone age all the way through the industrial revolution to the present, technology was primarily a tool. That means at best, technology did exactly what you wanted it to do. No more, no less.
We’re now getting to the point where technology is moving beyond that and becoming a true collaborator in human endeavors—in the greater endeavor, in the endeavor of physically manufacturing something or constructing something or those kinds of things. It’s becoming something that was only in the realm of human intellect before.
You’re speaking now strictly as a technologist. Looking at it from society’s point of view, are you not concerned about the risks? Another technologist, Elon Musk, has issued a warning about AI.
I think like anything, technology can be used or misused. That is why you see so much of a focus on this from Andrew [Anagnost]. Who to the core of his being…
Is a technologist?
He is also incredibly concentrated on the future of technology, work and society. At Autodesk University last year, he shared that as a company working in this area, we don’t only have a responsibility as a business but also a moral obligation to society to make sure that the technology we provide does the right things for society. You hear this mantra from Andrew all the time about more, better, less. There’s the inevitability of more in the world. The population is growing, but you have some phenomenal number of people—400,000 every day—joining the middle class. That creates inevitable demand for more things – food, water, housing, products, transportation, infrastructure, etc., and a way to make those things better.
At the same time you’ve got a fundamental constraint tied to the planet. We have to be able to do all of that with less. One of things you have to decrease is the number of people per unit volume. Each person has to do more.
It’s ironic that some are talking about the threat of technology to jobs and others job types are suffering shortage like construction workers, truck drivers and skilled laborers of all kinds. We have this fundamental challenge that we don’t have enough people to fill the jobs that we’ve got today.
What about the upheaval caused by new technology? There are the lower strata of people who are going to be washed by technology changes, right? The people that don’t fit. Think of the industrial revolution. There were many people who didn’t get and didn’t fit into the new scheme of things.
I think people adapt over time. The classic example is farming. We had nearly half of the workforce in 1900 doing agriculture. Now, it’s 2 percent. If you ask most of the people who were displaced by technology on the farm if they actually would like to work in the fields doing that level of manual labor, you wouldn’t find many people who would want to do that.
I think it’s always a question of doing things in a responsible way. It’s a fundamental challenge of technology companies to address the question of the impact of technology change on society. I’m proud that Autodesk has been more aware of this than most technology companies. We are trying much harder to be engaged, whether it’s on the education side of things or scaling the workforce. Some of the research we’re doing in the office of CTO is about the notion of continuous learning and continuous scaling as opposed to needing to go back to school every two years or every time something new comes along. We’re spending a lot of time thinking about knowledge. What do we need to know? When do we need to know it? How are we going to learn it? There’s a lot of work going on in that area.
Those are all things we had our intern group here work on over the summer. When we reviewed their projects, we were amazed by what they did. That was one of the challenges. There was also the notion of human-machine collaboration that I talked about earlier.
One of the main challenges we have with the human-machine collaboration can be used to prevent the planet from being covered 3-feet deep in useless junk by the end of the century. It’s our challenge, not just as members of the voting public but as members of a responsible society in the work we do every day, to make sure we use technology the right way. I honestly think the public institutions are going to be challenged to keep up with the pace of change we are seeing now.
From Philosophy to Technology
When I was here with Andrew [Anagnost, CEO, see interview], we talked about some of the advances Autodesk is making using databases rather than files. Do you have anything to add to that?
One of the things that I was, and still am, is a big advocate for are the changes we’re doing around the cloud as a platform—the cloud as a data flow, a workflow, process automation platform.
I’ve been very close to that work and still am a huge advocate of it. Getting back to technology as a tool, the whole industry, not just Autodesk, was intent on making specialist tools for specialists. You had people good at turning design into geometry. You had people good at doing simulation. You had people good at systems analysis, as well as people who were good at CAM and so on. The natural bent always was to make that tool deeper and deeper, and more and more functional, for the person user using it.
Design and simulation were getting more specialized?
Right. When you talk to customers, which we did, and those at higher levels—the engineering managers, business people—and ask them, “What is your biggest problem as a business?” You virtually never hear about this or that missing feature in one of the vertical tools we supplied. It is always about the process and how there are bottle necks. There was a lot of that. There would be data loss or there would be no value-added work being done. Historically, we didn’t do a great job, quite honestly, as a company and industry.
Was there an ease of use problem or something else?
It’s really about data flow and workflow. Here’s an example. I need to do a simulation. The first thing I must do is translate the model into a model I can use. Three quarters of these tiny little pieces of geometry may not matter. I have to first go through the model and rid it of all the details. Then, I run the simulation and send it back to the designer. They revise the design and send it back to me. I do the same thing all over again. We’ve taken a high-priced person and had them wasting half their time or more doing this. This is what you hear from customers: the high percentage of work that adds no value. It requires no creativity, no judgment. It’s like you are in your garage workshop with a bucket of fasteners.
Every time you need one, you pour them all out on the floor and sift through them. That’s what we made our customers do every single day with our tools.
We’re moving in a direction where, when we finish one thing, the data that needs to go to a person for that purpose is just this data, no more, no less. I’m going to make sure every time that design is revised, that data’s going to flow there automatically. If I need to run some pre-preparation processes on those, I’m going to queue those up to run automatically too.
Just-in-time information?
Yes, it is all about just-in-time information. It’s kind of funny. There’s a video of me taken at Autodesk University 2003 that just resurfaced talking about just-in-time information and these exact kinds of concepts with a prototype Microsoft tablet PC. We were never really able to make it happen fully. We’re now getting to where we can make it happen. It is really exciting.
Is this because the internet and mobile devices are now pervasive?
It’s certainly because of mobile devices being pervasive, as well as being sufficiently powerful enough now to deal with what we need to do. It is also because networks are available like cellular networks and Wi-Fi at the construction trailers, on the shop floor and everywhere else. It’s pervasive networking. It’s the combination of things we’ve been able to build in the cloud the last few years. It’s a set of callable services that have our core IP behind them. Honestly, there are some pretty unique things we’ve done in terms of managing data modeling and data flow. It’s what we are doing and what we’re enabling people to build on top of it. All of that is going to come together over the next couple of years. You’ll see more of this coming to fruition.
I was at AECOM and saw them working on an airport. Everybody on their team, wherever they were, were all working off the same data model because the model was on the cloud.
Interesting that you say one data model. It only works that way. The only way you can get that is in the cloud because in practice, many different data stores are federated together in such a way that they look like one data model.
The cloud is ultimately transforming this industry. I don’t mean taking all of those old vertical tools and making browser versions from them. It’s about automating the customer’s whole process from design through construction, from design through manufacturing—automating those workflows and then building intelligence on top of that automation with things like generative design and AI.
When I started this, I remember reading white papers in the late 70s, early 80s, about solid modeling, feature-based modeling and feature-based CAD. I think the authors of those papers are still around. They must be looking at the state of the industry today and saying, “That sounds familiar. That was what we envisioned 40 years ago for this industry.” Ten years from now, I don’t think that will be true. I think things will have transformed so much that they will have gone well beyond the visions of what the early pioneers of CAD envisioned.
Do you see people not using CAD?
I think people will still use CAD. But, generative design obviously is one of those transformative things. The idea that a computer can design…
Instead of an engineer creating the shape, it’s the software?
And dealing with system level abstractions, like the ability to model a whole end-to-end process and be able to have that automated. The idea that your tool is now a collaborator, that it is smarter every single day. Every time you use it, it’s a little smarter about what you’re going to do next, who you’re going to collaborate with, what data you need and what you did last time, all of those kinds of things. All of that is going to be a quite different experience from what was envisioned back in the day.
Like my phone tells me how long it will take to get home without me asking?
Yes. I start walking toward the parking garage, and my phone buzzes and says, “Take this exit because that route will be faster today.” We’re embedded in this change of technology as a collaborator. We don’t necessarily think of it that way, but that’s what’s happening.
Security and Privacy
With everybody storing their data in a cloud, what thoughts does Autodesk give to security and privacy?
It’s not really an Autodesk problem. It’s everybody’s problem. You see brand new roles in organizations called chief security officer. We have one, Reeny Sondhi, who was recently appointed. She’s here in San Francisco.
We take security and privacy very seriously. We all go through training on protecting people’s privacy and securing data.
In a recent interview, you talked about we live in a state of hyperconnectivity. You can detect a person moving around in an online forum. I’m not quoting you properly, but you can actually follow a person around, even in an online world. Isn’t that a violation of privacy?
I think things like [General Data Protection Regulation (GDPR)], if not GDPR specifically, tell us we can’t do any of that, unless the people specifically permit it. As our CEO says, we’re not a company that is in the business of collecting customer data to sell it to others. When we use data, it is for improving customer productivity and experiences. And, it is always with their permission. We’re very careful about anonymizing data. Our point of view on data is almost completely different than Google’s or Facebook’s for an obvious reason: their business is basically selling data.
But collecting user data, Autodesk has been doing that for some time. Keystroke data for example.
Yes. Particularly in things like Beta programs. We collect usage data. We learn a lot about how to improve customer experiences that way.
Parting Thoughts
In conclusion, anything you want to add or emphasize?
One of the things I find fascinating and potentially enormously powerful is this notion of knowledge. I talk to customers on a regular basis at customer briefings. It seems like in almost every customer briefing, the same subject comes up. It is the notion of knowledge transfer. We have all these bright people in our engineering and technical worlds. Every time one of them walks out the door, they take 20 years of knowledge and experience with them that we have to somehow recreate. It seems the reasons we made that product the way it was are lost to history. The same thing happens in construction. Huge amounts of information gets collected about a construction site that impacts how well that building came in on cost or time, the safety risks and everything else. Once again, all that information is lost to history.
Also true of manufacturing that went overseas. If it didn’t, many of those who still had manufacturing jobs have retired. How are we ever going to bring manufacturing back?
That’s another one of those kinds of things that they would talk to customers about. Twenty years ago, we would go downstairs to the shop floor when we wanted to know something. Now, we have to ask a supplier, who may not want to share that information with us.
This notion of being able to accumulate knowledge through time and in a way that you don’t have to require people to use your proprietary systems to be able to accumulate that knowledge is one of the classic problems with PLM. Nobody wants to spend all the time needed to enter all the information into PLM. They do it because it helps the company enforce its processes and rules. If you have to capture knowledge through that same sort of thing, most of the interesting knowledge won’t get captured because all of that will happen in person-to-person transactions that are below the radar screen. You have to figure out how to take email, Slack feeds, pictures of whiteboards, specification documents, notebooks, engineering notebooks and so on. We’re going to have to figure out how we take all that stuff and turn it into a body of knowledge that customers can use to figure out.
That would be done with AI?
Absolutely. That will involve many applications of AI.
More than just design and manufacturing knowledge, you are talking about process knowledge.
It is absolutely process knowledge. In fact, it’s the process knowledge more than anything that is very interesting. It’s not going to be one source of information from one company. It’s going to be multiple distributed sorts of information and being able to do that in the domains we know well. In the domains of design, manufacturing and construction, that’s going to be a powerful emerging thing.