Nicholas Childs PART 2: From Analog Roots to Digital Frontiers

NICHOLAS CHILD And I know nobody wants to hear this. Sometimes it's okay to slow down to go faster. This is the world right now where if you are not paying attention, it's not an if. It's a when are they going to pass us up? When is this going to happen? Who's going to come in and eat our lunch? And I just want people to take the time, look at their business, say, "Hey, this is what we do. This is what I want to accomplish. These are the things we do really well and let's go try to figure out how we can optimize that." And once we start doing that, you're going to see gains that you never thought were possible. And I'm seeing that in the market through other organizations, really, really cool.

GREG SIMPSON Hey gang, Greg here and welcome to another episode of The Built to Suit Podcast, the show Built to Build What's Next. This was a guest that was so big he grew over two episodes, a fantastic conversation we had with Nicholas Childs of Autodesk last time, but he comes back today. We're going to talk about AI and automation and digital twins and what's next right here on the Built to Suit podcast.

GREG SIMPSON I first saw Digital Twin probably in the context of an airplane engine, something that was built and singular and you could look at meantime between failure data on the way the engine performed. And I think sometimes we extrapolate that to buildings and I probably think that's an unfair extrapolation. So what I want you to do is I want you to thin that down just a little bit and talk about Digital Twin and the confines of the built environment and how you and Autodesk view it.

NICHOLAS CHILD Yeah. I would say 90% of the people I talk to are pursuing Digital Twin. The first problem that they have is the model doesn't have to be perfect because the real thing with Digital Twins, the best companies who do them, they always start with the end in mind. What am I going to track? Why is that important? What is the outcome that I'm looking for? And then we build to that outcome because digital twins can quickly... There's a lot of potential use cases, right? I want to track water. Hey, we want to track energy. Hey, I want to track occupancy. But which ones of those actually turn around and give you more dollars in your business, right? Improve your bottom line.

NICHOLAS CHILD So in the base factor, I've built digital twins on sheets. I've built digital twins in our product tandem. And it's really about understanding, okay, we're going to track these things and then building backwards. Okay. What does it mean to track those things? How often do we upgrade those things? What is the maintenance schedule? How many times a year should we update that drawing? And you're putting together the plan around upkeeping that digital twin, but it doesn't have to be perfect.

GREG SIMPSON I think one thing you told me the other day, automation can't help if your data is like a messy garage. So I guess now's really the time to unpack AI inside the space of maybe BIM, but also what's the starting spot?

NICHOLAS CHILD Awesome. Yeah. So Autodesk has been around in AI for a really, really long time. I believe our first papers came out in like 2017 and we were doing generative AI. Automation has to come before AI because the first thing you have to do is you have to move things in a smart way, or you have to put things into a place where everybody can find those things. But once those items, once that data is in a place, then you can start to ask it questions.

GREG SIMPSON Well, LLMs by definition are probabilistic, right? I mean, that's the nature of an LLM is to optimize on the moment and give you an answer.

NICHOLAS CHILD Exactly.

GREG SIMPSON So when we talk about determinism or deterministic systems, I guess what we're talking about is constraining the AI to a much more engineering-focused answer.

NICHOLAS CHILD I think the answer to that is yes. There are things that we're doing inside of Autodesk where we're training our models to respond better, to understand real world activities and understand the world at large.

NICHOLAS CHILD But on top of that, I also think that organizations have an opportunity right now to create spaces within their company that can leverage their IP to enhance what we're doing.

GREG SIMPSON So you got deterministic, that would be like standards and details and specs and lessons learned and stuff like that.

NICHOLAS CHILD Yeah.

GREG SIMPSON Versioning probabilistic shines in that.

NICHOLAS CHILD You know what? I think this is going to be... I think we're all on a ride for that, right? I wish I could say that we got it. We understand it. I think we understand what we understand.

NICHOLAS CHILD Some of the stuff that we do in the industry is just tasks and it's just heavy on the employee. It's like, okay, I've just got to get through it. I've got to do a thousand tasks and they got to be the exact same each time. Please, that needs to be automated.

GREG SIMPSON So where do we start as we start to say, “Hey Nick, I want to do AI. I'm ready.” Now what?

NICHOLAS CHILD It really does have to deal with those standards because you can also teach AI bad things about your business, right? You throw a bunch of bad models into AI and it then starts to think that those are the way that you model.

NICHOLAS CHILD It’s kind of going back to the basics a little bit. “Hey, this is an architect. This is what an architect does. This is how an architect thinks.”

GREG SIMPSON Deprecated. It does sound like something a dog does, but being able to actually deprecate something in my view is being able to, in a controlled way, know it's not in use anymore.

NICHOLAS CHILD There are a set of tools we call the interoperability tools. Within those tools there are tools like the model validation tool where basically you put in a check set and say, “Hey, take this model and check it against this and tell me what the report of that thing is.”

GREG SIMPSON So we've talked a lot about really the architect space. That's really not your space. You are cloud, but you're really more institutional cloud.

NICHOLAS CHILD That has been one of the funnest challenges I've experienced in my career is really working to understand what information does a person need and when do they need it?

NICHOLAS CHILD Especially in institutional clients, you have to go slow. You have to go really slow.

GREG SIMPSON But they don't want to go slow.

NICHOLAS CHILD Never do. They never want to go slow.

NICHOLAS CHILD What we're doing now is saying, “Okay, if you just need those two things, how would you like them presented?”

GREG SIMPSON What are we going to see as we move forward in this space? What are going to be the things that disrupt us?

NICHOLAS CHILD I think scanning and reality capture and all of that stuff is really important. I think the other thing that I'm starting to see is there's a lot more collaboration across organizations.

NICHOLAS CHILD Small to medium-sized companies are being able to really take on the big boys.

GREG SIMPSON Who's going to win out? The big guy who's bought something and makes the investment and has 200,000 people or the small guy who's more agile?

NICHOLAS CHILD Honestly, I would say it's probably the smaller company. Smaller companies can innovate quick. They can change quickly.

NICHOLAS CHILD Smaller companies have to think, “How do I solve that problem with the same people I have right now?” That mindset is the perfect mindset to be in right now.

GREG SIMPSON Do you see a change to the way we interact and the level that we interact with software?

NICHOLAS CHILD Oh yeah. Things are changing so rapidly.

NICHOLAS CHILD I don't really think today people engage in AI in kind of like a one-off sense. Okay, I need to answer this question. I'm going to go ask AI.

GREG SIMPSON I think the people who have a challenge with their job in this space are the ones that did the same thing 12 times today.

NICHOLAS CHILD Yes.

GREG SIMPSON There’s always going to be a job available for the person that can orchestrate.

NICHOLAS CHILD This is like the thing that Autodesk is aiming for and we call it outcome-based BIM.

NICHOLAS CHILD Rather than thinking about the problem and how you're going to solve the problem, we start asking it for the outcome.

NICHOLAS CHILD Organizations have to take the time and I know nobody wants to hear this. Sometimes it's okay to slow down to go faster.

GREG SIMPSON So what you just told me was AI is not here to make it cheap. It's here to make it good.

NICHOLAS CHILD You can have your cake and eat it too.

GREG SIMPSON Nick, a wonderful conversation.

NICHOLAS CHILD Thank you. Appreciate you all having me.